def do_json_request(self, jsonRequest=None, fullUrl=None, timeout=10, params=None, postData=None, returnFast=False,
cmd='get', extraComment=None, ignoreH2oError=False, noExtraErrorCheck=False, **kwargs):
# if url param is used, use it as full url. otherwise create from the jsonRequest
if fullUrl:
url = fullUrl
else:
url = self.url(jsonRequest)
# remove any params that are 'None'
# need to copy dictionary, since can't delete while iterating
if params is not None:
params2 = params.copy()
for k in params2:
if params2[k] is None:
del params[k]
paramsStr = '?' + '&'.join(['%s=%s' % (k, v) for (k, v) in params.items()])
else:
paramsStr = ''
extraComment2 = " " + str(postData)+";" if cmd=='post' else ""
extraComment2 += extraComment if extraComment else ""
if len(extraComment2) > 0:
log('Start ' + url + paramsStr, comment=extraComment2)
else:
log('Start ' + url + paramsStr)
# file get passed thru kwargs here
if h2o_args.no_timeout:
timeout = None # infinite
try:
if 'post' == cmd:
# NOTE == cmd: for now, since we don't have deserialization from JSON in h2o-dev, we use form-encoded POST.
# This is temporary.
#
# This following does application/json (aka, posting JSON in the body):
# r = requests.post(url, timeout=timeout, params=params, data=json.dumps(postData), **kwargs)
#
# This does form-encoded, which doesn't allow POST of nested structures
r = requests.post(url, timeout=timeout, params=params, data=postData, **kwargs)
elif 'delete' == cmd:
r = requests.delete(url, timeout=timeout, params=params, **kwargs)
elif 'get' == cmd:
r = requests.get(url, timeout=timeout, params=params, **kwargs)
else:
raise ValueError("Unknown HTTP command (expected 'get', 'post' or 'delete'): " + cmd)
except Exception, e:
# rethrow the exception after we've checked for stack trace from h2o
# out of memory errors maybe don't show up right away? so we should wait for h2o
# to get it out to h2o stdout. We don't want to rely on cloud teardown to check
# because there's no delay, and we don't want to delay all cloud teardowns by waiting.
exc_info = sys.exc_info()
# use this to ignore the initial connection errors during build cloud when h2o is coming up
if not noExtraErrorCheck:
h2p.red_print(
"ERROR: got exception on %s to h2o. \nGoing to check sandbox, then rethrow.." % (url + paramsStr))
time.sleep(2)
check_sandbox_for_errors(python_test_name=h2o_args.python_test_name);
raise exc_info[1], None, exc_info[2]
def do_json_request(self, jsonRequest=None, fullUrl=None, timeout=10, params=None, returnFast=False,
cmd='get', extraComment=None, ignoreH2oError=False, noExtraErrorCheck=False, **kwargs):
# if url param is used, use it as full url. otherwise crate from the jsonRequest
if fullUrl:
url = fullUrl
else:
url = self.url(jsonRequest)
# remove any params that are 'None'
# need to copy dictionary, since can't delete while iterating
if params is not None:
params2 = params.copy()
for k in params2:
if params2[k] is None:
del params[k]
paramsStr = '?' + '&'.join(['%s=%s' % (k, v) for (k, v) in params.items()])
else:
paramsStr = ''
if extraComment:
log('Start ' + url + paramsStr, comment=extraComment)
else:
log('Start ' + url + paramsStr)
log_rest("")
log_rest("----------------------------------------------------------------------\n")
if extraComment:
log_rest("# Extra comment info about this request: " + extraComment)
if cmd == 'get':
log_rest("GET")
else:
log_rest("POST")
log_rest(url + paramsStr)
# file get passed thru kwargs here
try:
if cmd == 'post':
r = requests.post(url, timeout=timeout, params=params, **kwargs)
else:
r = requests.get(url, timeout=timeout, params=params, **kwargs)
except Exception, e:
# rethrow the exception after we've checked for stack trace from h2o
# out of memory errors maybe don't show up right away? so we should wait for h2o
# to get it out to h2o stdout. We don't want to rely on cloud teardown to check
# because there's no delay, and we don't want to delay all cloud teardowns by waiting.
# (this is new/experimental)
exc_info = sys.exc_info()
# use this to ignore the initial connection errors during build cloud when h2o is coming up
if not noExtraErrorCheck:
h2p.red_print(
"ERROR: got exception on %s to h2o. \nGoing to check sandbox, then rethrow.." % (url + paramsStr))
time.sleep(2)
check_sandbox_for_errors(python_test_name=h2o_args.python_test_name);
log_rest("")
log_rest("EXCEPTION CAUGHT DOING REQUEST: " + str(e.message))
raise exc_info[1], None, exc_info[2]
def build_cloud_with_json(h2o_nodes_json='h2o-nodes.json'):
log("#*********************************************************************")
log("Starting new test: " + h2o_args.python_test_name + " at build_cloud_with_json()")
log("#*********************************************************************")
print "This only makes sense if h2o is running as defined by", h2o_nodes_json
print "For now, assuming it's a cloud on this machine, and here's info on h2o processes running here"
print "No output means no h2o here! Some other info about stuff on the system is printed first though."
import h2o_os_util
if not os.path.exists(h2o_nodes_json):
raise Exception("build_cloud_with_json: Can't find " + h2o_nodes_json + " file")
# h2o_os_util.show_h2o_processes()
with open(h2o_nodes_json, 'rb') as f:
cloneJson = json.load(f)
# These are supposed to be in the file.
# Just check the first one. if not there, the file must be wrong
if not 'cloud_start' in cloneJson:
raise Exception("Can't find 'cloud_start' in %s, wrong file? h2o-nodes.json?" % h2o_nodes_json)
else:
cs = cloneJson['cloud_start']
print "Info on the how the cloud we're cloning was started (info from %s)" % h2o_nodes_json
# required/legal values in 'cloud_start'. A robust check is good for easy debug when we add stuff
valList = ['time', 'cwd', 'python_test_name', 'python_cmd_line', 'config_json', 'username', 'ip']
for v in valList:
if v not in cs:
raise Exception("Can't find %s in %s, wrong file or version change?" % (v, h2o_nodes_json))
print "cloud_start['%s']: %s" % (v, cs[v])
# this is the internal node state for python..nodes rebuild
nodeStateList = cloneJson['h2o_nodes']
nodeList = []
if not nodeStateList:
raise Exception("nodeStateList is empty. %s file must be empty/corrupt" % h2o_nodes_json)
for nodeState in nodeStateList:
print "Cloning state for node", nodeState['node_id'], 'from', h2o_nodes_json
newNode = ExternalH2O(nodeState)
nodeList.append(newNode)
print ""
h2p.red_print("Ingested from json:", nodeList[0].java_heap_GB, "GB java heap(s) with",
len(nodeList), "total nodes")
print ""
# put the test start message in the h2o log, to create a marker
nodeList[0].h2o_log_msg()
# save it to a global copy, in case it's needed for tearDown
h2o_nodes.nodes[:] = nodeList
return nodeList
def test_build_for_clone(self):
# python gets confused about which 'start' if I used start here
elapsed = time.time() - beginning
print "\n%0.2f seconds to get here from start" % elapsed
# might as well open a browser on it? (because the ip/port will vary
# maybe just print the ip/port for now
## h2b.browseTheCloud()
maxTime = 4*3600
totalTime = 0
incrTime = 60
h2p.purple_print("\nSleeping for total of", (maxTime+0.0)/3600, "hours.")
print "Will check h2o logs every", incrTime, "seconds"
print "Should be able to run another test using h2o-nodes.json to clone cloud"
print "i.e. h2o.build_cloud_with_json()"
print "Bad test if a running test shuts down the cloud. I'm supposed to!\n"
h2p.green_print("To watch cloud in browser follow address:")
h2p.green_print(" http://{0}:{1}/Cloud.html".format(h2o.nodes[0].http_addr, h2o.nodes[0].port))
h2p.blue_print("You can start a test (or tests) now!")
h2p.blue_print("Will Check cloud status every %s secs and kill cloud if wrong or no answer" % incrTime)
if CHECK_WHILE_SLEEPING:
h2p.blue_print("Will also look at redirected stdout/stderr logs in sandbox every %s secs" % incrTime)
h2p.red_print("No checking of logs while sleeping, or check of cloud status")
h2p.yellow_print("So if H2O stack traces, it's up to you to kill me if 4 hours is too long")
h2p.yellow_print("ctrl-c will cause all jvms to die(thru psutil terminate, paramiko channel death or h2o shutdown...")
while (totalTime<maxTime): # die after 4 hours
time.sleep(incrTime)
totalTime += incrTime
# good to touch all the nodes to see if they're still responsive
# give them up to 120 secs to respond (each individually)
### h2o.verify_cloud_size(timeoutSecs=120)
if CHECK_WHILE_SLEEPING:
print "Checking sandbox log files"
h2o.check_sandbox_for_errors(cloudShutdownIsError=True)
else:
print str(datetime.datetime.now()), h2o_args.python_cmd_line, "still here", totalTime, maxTime, incrTime
# don't do this, as the cloud may be hung?
if 1==0:
print "Shutting down cloud, but first delete all keys"
start = time.time()
h2i.delete_keys_at_all_nodes()
elapsed = time.time() - start
print "delete_keys_at_all_nodes(): took", elapsed, "secs"
def test_build_for_clone(self):
# python gets confused about which 'start' if I used start here
elapsed = time.time() - beginning
print "\n%0.2f seconds to get here from start" % elapsed
# might as well open a browser on it? (because the ip/port will vary
# maybe just print the ip/port for now
## h2b.browseTheCloud()
maxTime = 4*3600
totalTime = 0
incrTime = 60
h2p.purple_print("\nSleeping for total of", (maxTime+0.0)/3600, "hours.")
print "Will check h2o logs every", incrTime, "seconds"
print "Should be able to run another test using h2o-nodes.json to clone cloud"
print "i.e. h2o.build_cloud_with_json()"
print "Bad test if a running test shuts down the cloud. I'm supposed to!\n"
h2p.green_print("To watch cloud in browser follow address:")
h2p.green_print(" http://{0}:{1}/Cloud.html".format(h2o.nodes[0].http_addr, h2o.nodes[0].port))
h2p.blue_print("You can start a test (or tests) now!")
h2p.blue_print("Will Check cloud status every %s secs and kill cloud if wrong or no answer" % incrTime)
if CHECK_WHILE_SLEEPING:
h2p.blue_print("Will also look at redirected stdout/stderr logs in sandbox every %s secs" % incrTime)
h2p.red_print("No checking of logs while sleeping, or check of cloud status")
h2p.yellow_print("So if H2O stack traces, it's up to you to kill me if 4 hours is too long")
h2p.yellow_print("ctrl-c will cause all jvms to die(thru psutil terminate, paramiko channel death or h2o shutdown...")
while (totalTime<maxTime): # die after 4 hours
h2o.sleep(incrTime)
totalTime += incrTime
# good to touch all the nodes to see if they're still responsive
# give them up to 120 secs to respond (each individually)
h2o.verify_cloud_size(timeoutSecs=120)
if CHECK_WHILE_SLEEPING:
print "Checking sandbox log files"
h2o.check_sandbox_for_errors(cloudShutdownIsError=True)
else:
print str(datetime.datetime.now()), h2o.python_cmd_line, "still here", totalTime, maxTime, incrTime
# don't do this, as the cloud may be hung?
if 1==0:
print "Shutting down cloud, but first delete all keys"
start = time.time()
h2i.delete_keys_at_all_nodes()
elapsed = time.time() - start
print "delete_keys_at_all_nodes(): took", elapsed, "secs"
def do_json_request(addr=None,
port=None,
jsonRequest=None,
params=None,
timeout=7,
**kwargs):
if params is not None:
paramsStr = '?' + '&'.join(
['%s=%s' % (k, v) for (k, v) in params.items()])
else:
paramsStr = ''
url = create_url(addr, port, jsonRequest)
print 'Start ' + url + paramsStr
try:
r = requests.get(url, timeout=timeout, params=params, **kwargs)
# the requests json decoder might fail if we didn't get something good
rjson = r.json()
emsg = "ERROR: Probing claimed existing cloud with Cloud.json"
if not isinstance(rjson, (list, dict)):
# probably good
raise Exception(emsg + "h2o json responses should always be lists or dicts. Got %s" %\
dump_json(rj))
elif r.status_code != requests.codes.ok:
rjson = None
raise Exception(emsg +
"Couldn't decode. Status: %s" % r.status_code)
except requests.ConnectionError, e:
rjson = None
emsg = "ERROR: json got ConnectionError or other exception"
# Rethrow the exception after we've checked for stack trace from h2o.
# Out of memory errors maybe don't show up right away?
# so we should wait for h2o to get it out to h2o stdout.
# Don't want to rely on cloud teardown to check because there's no delay,
# and we don't want to delay all cloud teardowns by waiting.
exc_info = sys.exc_info()
# we don't expect to have connection errors, so any exception is a bad thing.
h2p.red_print("%s\n %s\n %s\nGoing to check sandbox, then rethrow.." %
(emsg, exc_info, url + paramsStr))
time.sleep(2)
check_sandbox_for_errors()
raise exc_info[1], None, exc_info[2]
def test_build_for_clone(self):
# python gets confused about which 'start' if I used start here
elapsed = time.time() - beginning
print "\n%0.2f seconds to get here from start" % elapsed
# might as well open a browser on it? (because the ip/port will vary
# maybe just print the ip/port for now
## h2b.browseTheCloud()
maxTime = 4 * 3600
totalTime = 0
incrTime = 60
h2p.purple_print("\nSleeping for total of", (maxTime + 0.0) / 3600,
"hours.")
print "Will check h2o logs every", incrTime, "seconds"
print "Should be able to run another test using h2o-nodes.json to clone cloud"
print "i.e. h2o.build_cloud_with_json()"
print "Bad test if a running test shuts down the cloud. I'm supposed to!\n"
h2p.green_print("To watch cloud in browser follow address:")
h2p.green_print(" http://{0}:{1}/Cloud.html".format(
h2o.nodes[0].http_addr, h2o.nodes[0].port))
h2p.blue_print("You can start a test (or tests) now!")
h2p.blue_print(
"Will spin looking at redirected stdout/stderr logs in sandbox for h2o errors every %s secs"
% incrTime)
h2p.red_print("This is just for fun")
h2p.yellow_print("So is this")
while (totalTime < maxTime): # die after 4 hours
h2o.sleep(incrTime)
totalTime += incrTime
# good to touch all the nodes to see if they're still responsive
# give them up to 120 secs to respond (each individually)
h2o.verify_cloud_size(timeoutSecs=120)
print "Checking sandbox log files"
h2o.check_sandbox_for_errors(cloudShutdownIsError=True)
start = time.time()
h2i.delete_keys_at_all_nodes()
elapsed = time.time() - start
print "delete_keys_at_all_nodes(): took", elapsed, "secs"
def test_build_for_clone(self):
# python gets confused about which 'start' if I used start here
elapsed = time.time() - beginning
print "\n%0.2f seconds to get here from start" % elapsed
# might as well open a browser on it? (because the ip/port will vary
# maybe just print the ip/port for now
## h2b.browseTheCloud()
maxTime = 4*3600
totalTime = 0
incrTime = 60
h2p.purple_print("\nSleeping for total of", (maxTime+0.0)/3600, "hours.")
print "Will check h2o logs every", incrTime, "seconds"
print "Should be able to run another test using h2o-nodes.json to clone cloud"
print "i.e. h2o.build_cloud_with_json()"
print "Bad test if a running test shuts down the cloud. I'm supposed to!\n"
h2p.green_print("To watch cloud in browser follow address:")
h2p.green_print(" http://{0}:{1}/Cloud.html".format(h2o.nodes[0].http_addr, h2o.nodes[0].port))
h2p.blue_print("You can start a test (or tests) now!")
h2p.blue_print("Will spin looking at redirected stdout/stderr logs in sandbox for h2o errors every %s secs" % incrTime)
h2p.red_print("This is just for fun")
h2p.yellow_print("So is this")
while (totalTime<maxTime): # die after 4 hours
h2o.sleep(incrTime)
totalTime += incrTime
# good to touch all the nodes to see if they're still responsive
# give them up to 120 secs to respond (each individually)
h2o.verify_cloud_size(timeoutSecs=120)
print "Checking sandbox log files"
h2o.check_sandbox_for_errors(cloudShutdownIsError=True)
start = time.time()
h2i.delete_keys_at_all_nodes()
elapsed = time.time() - start
print "delete_keys_at_all_nodes(): took", elapsed, "secs"
def do_json_request(addr=None, port=None, jsonRequest=None, params=None, timeout=7, **kwargs):
if params is not None:
paramsStr = '?' + '&'.join(['%s=%s' % (k,v) for (k,v) in params.items()])
else:
paramsStr = ''
url = create_url(addr, port, jsonRequest)
print 'Start ' + url + paramsStr
try:
r = requests.get(url, timeout=timeout, params=params, **kwargs)
# the requests json decoder might fail if we didn't get something good
rjson = r.json()
emsg = "ERROR: Probing claimed existing cloud with Cloud.json"
if not isinstance(rjson, (list,dict)):
# probably good
raise Exception(emsg + "h2o json responses should always be lists or dicts. Got %s" %\
dump_json(rj))
elif r.status_code != requests.codes.ok:
rjson = None
raise Exception(emsg + "Couldn't decode. Status: %s" % r.status_code)
except requests.ConnectionError, e:
rjson = None
emsg = "ERROR: json got ConnectionError or other exception"
# Rethrow the exception after we've checked for stack trace from h2o.
# Out of memory errors maybe don't show up right away?
# so we should wait for h2o to get it out to h2o stdout.
# Don't want to rely on cloud teardown to check because there's no delay,
# and we don't want to delay all cloud teardowns by waiting.
exc_info = sys.exc_info()
# we don't expect to have connection errors, so any exception is a bad thing.
h2p.red_print(
"%s\n %s\n %s\nGoing to check sandbox, then rethrow.." % (emsg, exc_info, url + paramsStr))
time.sleep(2)
check_sandbox_for_errors()
raise exc_info[1], None, exc_info[2]
def do_h2o_glm(self, bucket, csvPathname, L, family="binomial"):
h2p.red_print("\nNow doing h2o")
h2o.beta_features = True
parseResult = h2i.import_parse(bucket="smalldata", path=csvPathname, schema="local", timeoutSecs=180)
# save the resolved pathname for use in the sklearn csv read below
inspect = h2o_cmd.runInspect(None, parseResult["destination_key"])
print inspect
print "\n" + csvPathname, " numRows:", "{:,}".format(inspect["numRows"]), " numCols:", "{:,}".format(
inspect["numCols"]
)
x = "ID"
y = "CAPSULE"
family = family
alpha = "0"
lambda_ = L
nfolds = "0"
f = "prostate"
modelKey = "GLM_" + f
kwargs = {
"response": y,
"ignored_cols": x,
"family": family,
"lambda": lambda_,
"alpha": alpha,
"n_folds": nfolds, # passes if 0, fails otherwise
"destination_key": modelKey,
}
timeoutSecs = 60
start = time.time()
glmResult = h2o_cmd.runGLM(parseResult=parseResult, timeoutSecs=timeoutSecs, **kwargs)
# this stuff was left over from when we got the result after polling the jobs list
# okay to do it again
# GLM2: when it redirects to the model view, we no longer have the job_key! (unlike the first response and polling)
(warnings, clist, intercept) = h2o_glm.simpleCheckGLM(self, glmResult, None, **kwargs)
cstring = "".join([("%.5e " % c) for c in clist])
h2p.green_print("h2o alpha ", alpha)
h2p.green_print("h2o lambda ", lambda_)
h2p.green_print("h2o coefficient list:", cstring)
h2p.green_print("h2o intercept", "%.5e " % intercept)
# other stuff in the json response
glm_model = glmResult["glm_model"]
_names = glm_model["_names"]
coefficients_names = glm_model["coefficients_names"]
# the first submodel is the right one, if onely one lambda is provided as a parameter above
submodels = glm_model["submodels"][0]
beta = submodels["beta"]
h2p.red_print("beta:", beta)
norm_beta = submodels["norm_beta"]
iteration = submodels["iteration"]
validation = submodels["validation"]
avg_err = validation["avg_err"]
auc = validation["auc"]
aic = validation["aic"]
null_deviance = validation["null_deviance"]
residual_deviance = validation["residual_deviance"]
print "_names", _names
print "coefficients_names", coefficients_names
# did beta get shortened? the simple check confirms names/beta/norm_beta are same length
print "beta", beta
print "iteration", iteration
print "avg_err", avg_err
print "auc", auc
def build_cloud(node_count=1, base_port=None, hosts=None,
timeoutSecs=30, retryDelaySecs=1, cleanup=True, rand_shuffle=True,
conservative=False, create_json=False, clone_cloud=None,
init_sandbox=True, usecloud=False, usecloud_size=None, **kwargs):
# expectedSize is only used if usecloud
# usecloud can be passed thru build_cloud param, or command line
# not in config json though so no build_cloud_with_hosts path.
# redirect to build_cloud_with_json if a command line arg
# wants to force a test to ignore it's build_cloud/build_cloud_with_hosts
# (both come thru here)
# clone_cloud is just another way to get the effect (maybe ec2 config file thru
# build_cloud_with_hosts?
global stdout_wrapped
if not h2o_args.disable_time_stamp and not stdout_wrapped:
sys.stdout = OutWrapper(sys.stdout)
stdout_wrapped = True
if h2o_args.usecloud or usecloud:
# for now, just have fixed name in local file. (think of this as a temp or debug file)
# eventually we'll pass the json object instead for speed?
nodesJsonPathname = "h2o_fc-nodes.json"
elif h2o_args.clone_cloud_json:
nodesJsonPathname = h2o_args.clone_cloud_json
elif clone_cloud:
nodesJsonPathname = clone_cloud
else:
# normal build_cloud() doesn't use
nodesJsonPathname = None
# usecloud dominates over all
if (h2o_args.clone_cloud_json or clone_cloud) or (h2o_args.usecloud or usecloud):
# then build_cloud_with_json with json object
# we don't need to specify these defaults, but leave here to show that we can pass
# I suppose kwargs will have it
if h2o_args.usecloud:
ip_port = h2o_args.usecloud
elif usecloud:
ip_port = usecloud
else:
ip_port = None
# h2o_args dominates
if h2o_args.usecloud_size:
# only used for expected size
useCloudExpectedSize = h2o_args.usecloud_size
else:
useCloudExpectedSize = usecloud_size
if (h2o_args.usecloud or usecloud):
nodesJsonObject = h2o_fc.find_cloud(ip_port=ip_port,
expectedSize=useCloudExpectedSize, nodesJsonPathname=nodesJsonPathname, **kwargs)
# potentially passed in kwargs
# hdfs_version='cdh4', hdfs_config=None, hdfs_name_node='172.16.1.176',
else:
if h2o_args.clone_cloud_json:
nodesJsonPathname = h2o_args.clone_cloud_json
else:
nodesJsonPathname = clone_cloud
nodeList = build_cloud_with_json(h2o_nodes_json=nodesJsonPathname)
return nodeList
# else
# moved to here from unit_main. so will run with nosetests too!
# Normally do this.
# Don't if build_cloud_with_hosts() did and put a flatfile in there already!
if init_sandbox:
clean_sandbox()
log("#*********************************************************************")
log("Starting new test: " + h2o_args.python_test_name + " at build_cloud() ")
log("#*********************************************************************")
# start up h2o to report the java version (once). output to python stdout
# only do this for regression testing
# temporarily disable this, to go a little faster
# if getpass.getuser() == 'jenkins':
# check_h2o_version()
ports_per_node = 2
nodeList = []
# shift the port used to run groups of tests on the same machine at the same time?
base_port = get_base_port(base_port)
try:
# if no hosts list, use psutil method on local host.
totalNodes = 0
# doing this list outside the loops so we can shuffle for better test variation
# this jvm startup shuffle is independent from the flatfile shuffle
portList = [base_port + ports_per_node * i for i in range(node_count)]
if hosts is None:
# if use_flatfile, we should create it
# because tests will just call build_cloud with use_flatfile=True
# best to just create it all the time..may or may not be used
write_flatfile(node_count=node_count, base_port=base_port)
hostCount = 1
if rand_shuffle:
random.shuffle(portList)
for p in portList:
verboseprint("psutil starting node", i)
newNode = LocalH2O(port=p, node_id=totalNodes, **kwargs)
nodeList.append(newNode)
totalNodes += 1
else:
# if hosts, the flatfile was created and uploaded to hosts already
# I guess don't recreate it, don't overwrite the one that was copied beforehand.
# we don't always use the flatfile (use_flatfile=False)
# Suppose we could dispatch from the flatfile to match it's contents
# but sometimes we want to test with a bad/different flatfile then we invoke h2o?
hostCount = len(hosts)
hostPortList = []
for h in hosts:
for port in portList:
hostPortList.append((h, port))
if rand_shuffle: random.shuffle(hostPortList)
for (h, p) in hostPortList:
verboseprint('ssh starting node', totalNodes, 'via', h)
newNode = h.remote_h2o(port=p, node_id=totalNodes, **kwargs)
nodeList.append(newNode)
totalNodes += 1
verboseprint("Attempting Cloud stabilize of", totalNodes, "nodes on", hostCount, "hosts")
start = time.time()
# UPDATE: best to stabilize on the last node!
# FIX! for now, always check sandbox, because h2oddev has TIME_WAIT port problems
stabilize_cloud(nodeList[0], nodeList,
timeoutSecs=timeoutSecs, retryDelaySecs=retryDelaySecs, noExtraErrorCheck=False)
stabilizeTime = time.time() - start
verboseprint(len(nodeList), "Last added node stabilized in ", stabilizeTime, " secs")
# assume all the heap sizes are the same as zero
if nodeList[0].java_heap_GB:
heapSize = str(nodeList[0].java_heap_GB) + " GB"
elif nodeList[0].java_heap_GB:
heapSize = str(nodeList[0].java_heap_MB) + " MB"
else:
heapSize = "(unknown)"
h2p.red_print("Built cloud: %s java heap(s) with %d nodes on %d hosts, stabilizing in %d secs" % \
(heapSize, len(nodeList), hostCount, stabilizeTime))
# FIX! using "consensus" in node[-1] should mean this is unnecessary?
# maybe there's a bug. For now do this. long term: don't want?
# UPDATE: do it for all cases now 2/14/13
if conservative: # still needed?
for n in nodeList:
# FIX! for now, always check sandbox, because h2oddev has TIME_WAIT port problems
stabilize_cloud(n, nodeList, timeoutSecs=timeoutSecs, noExtraErrorCheck=False)
# this does some extra checking now
# verifies cloud name too if param is not None
verify_cloud_size(nodeList, expectedCloudName=nodeList[0].cloud_name, expectedLocked=0)
# FIX! should probably check that the cloud's lock=0. It will go to 1 later.
# but if it's an existing cloud, it may already be locked.
# That will be in build_cloud_with_json, though
# best to check for any errors due to cloud building right away?
check_sandbox_for_errors(python_test_name=h2o_args.python_test_name)
# put the test start message in the h2o log, to create a marker
nodeList[0].h2o_log_msg()
except:
# nodeList might be empty in some exception cases?
# no shutdown issued first, though
if cleanup and nodeList:
for n in nodeList: n.terminate()
check_sandbox_for_errors(python_test_name=h2o_args.python_test_name)
raise
print len(nodeList), "total jvms in H2O cloud"
if h2o_args.config_json:
# like cp -p. Save the config file, to sandbox
print "Saving the ", h2o_args.config_json, "we used to", LOG_DIR
shutil.copy(h2o_args.config_json, LOG_DIR + "/" + os.path.basename(h2o_args.config_json))
if create_json:
# Figure out some stuff about how this test was run
cs_time = str(datetime.datetime.now())
cs_cwd = os.getcwd()
cs_python_cmd_line = "python %s %s" % (h2o_args.python_test_name, h2o_args.python_cmd_args)
cs_python_test_name = h2o_args.python_test_name
if h2o_args.config_json:
cs_config_json = os.path.abspath(h2o_args.config_json)
else:
cs_config_json = None
cs_username = h2o_args.python_username
cs_ip = h2o_args.python_cmd_ip
# dump the nodes state to a json file # include enough extra info to have someone
# rebuild the cloud if a test fails that was using that cloud.
q = {
'cloud_start':
{
'time': cs_time,
'cwd': cs_cwd,
'python_test_name': cs_python_test_name,
'python_cmd_line': cs_python_cmd_line,
'config_json': cs_config_json,
'username': cs_username,
'ip': cs_ip,
},
'h2o_nodes': h2o_util.json_repr(nodeList),
}
with open('h2o-nodes.json', 'w+') as f:
f.write(json.dumps(q, indent=4))
# save it to a local global copy, in case it's needed for tearDown
h2o_nodes.nodes[:] = nodeList
return nodeList
def do_scipy_glm(self, bucket, csvPathname, L, family='binomial'):
h2p.red_print("Now doing sklearn")
h2p.red_print("\nsee http://scikit-learn.org/0.11/modules/generated/sklearn.linear_model.LogisticRegression.html#sklearn.linear_model.LogisticRegression")
import numpy as np
import scipy as sp
from sklearn.linear_model import LogisticRegression
from numpy import loadtxt
csvPathnameFull = h2i.find_folder_and_filename(bucket, csvPathname, returnFullPath=True)
# make sure it does fp divide
C = 1/(L+0.0)
print "C regularization:", C
dataset = np.loadtxt(
open(csvPathnameFull,'r'),
skiprows=1, # skip the header
delimiter=',',
dtype='float');
print "\ncsv read for training, done"
n_features = len(dataset[0]) - 1;
print "n_features:", n_features
# don't want ID (col 0) or CAPSULE (col 1)
# get CAPSULE
target = [x[1] for x in dataset]
# slice off the first 2
train = np.array ( [x[2:] for x in dataset] )
n_samples, n_features = train.shape
print "n_samples:", n_samples, "n_features:", n_features
print "histogram of target"
print sp.histogram(target,3)
print "len(train):", len(train)
print "len(target):", len(target)
print "dataset shape:", dataset.shape
if family!='binomial':
raise Exception("Only have binomial logistic for scipy")
print "\nTrying l2"
clf2 = LogisticRegression(
C=C,
dual=False,
fit_intercept=True,
intercept_scaling=1,
penalty='l2',
tol=0.0001);
# train the classifier
start = time.time()
clf2.fit(train, target)
print "L2 fit took", time.time() - start, "seconds"
# print "coefficients:", clf2.coef_
cstring = "".join([("%.5e " % c) for c in clf2.coef_[0]])
h2p.green_print("sklearn L2 C", C)
h2p.green_print("sklearn coefficients:", cstring)
h2p.green_print("sklearn intercept:", "%.5e" % clf2.intercept_[0])
h2p.green_print("sklearn score:", clf2.score(train,target))
print "\nTrying l1"
clf1 = LogisticRegression(
C=C,
dual=False,
fit_intercept=True,
intercept_scaling=1,
penalty='l1',
tol=0.0001);
# train the classifier
start = time.time()
clf1.fit(train, target)
print "L1 fit took", time.time() - start, "seconds"
# print "coefficients:", clf1.coef_
cstring = "".join([("%.5e " % c) for c in clf1.coef_[0]])
h2p.green_print("sklearn L1 C", C)
h2p.green_print("sklearn coefficients:", cstring)
h2p.green_print("sklearn intercept:", "%.5e" % clf1.intercept_[0])
h2p.green_print("sklearn score:", clf1.score(train,target))
# attributes are accessed in the normal python way
dx = clf1.__dict__
dx.keys()
def quantile_comparisons(csvPathname, skipHeader=False, col=0, datatype='float',
h2oSummary2=None,
h2oSummary2MaxErr=None,
h2oQuantilesApprox=None, h2oQuantilesExact=None,
h2oExecQuantiles=None,
interpolate='linear', quantile=0.50, use_genfromtxt=False):
SCIPY_INSTALLED = False
try:
import scipy as sp
import numpy as np
print "Both numpy and scipy are installed. Will do extra checks"
except ImportError:
print "numpy or scipy is not installed. Will only do sort-based checking"
SCIPY_INSTALLED = False
if use_genfromtxt and SCIPY_INSTALLED:
print "Using numpy.genfromtxt. Better handling of null bytes"
target = np.genfromtxt(
open(csvPathname, 'r'),
delimiter=',',
skip_header=1 if skipHeader else 0,
dtype=None) # guess!
# print "shape:", target.shape()
else:
print "Using python csv reader"
target = h2o_util.file_read_csv_col(csvPathname, col=col, datatype=datatype,
skipHeader=skipHeader, preview=5)
if datatype=='float':
# to make irene's R runif files first col work (quoted row numbers, integers
#shouldn't hurt anyone else?
# strip " from left (ignore leading whitespace
# strip " from right (ignore leading whitespace
targetFP = map(float, target)
# targetFP= np.array(tFP, np.float)
if datatype=='int':
targetFP = map(int, target)
if SCIPY_INSTALLED:
# http://docs.scipy.org/doc/numpy-dev/reference/generated/numpy.percentile.html
# numpy.percentile has simple linear interpolate and midpoint
# need numpy 1.9 for interpolation. numpy 1.8 doesn't have
# p = np.percentile(targetFP, 50 if DO_MEDIAN else 99.9, interpolation='midpoint')
# 1.8
p = np.percentile(targetFP, quantile*100)
h2p.red_print("numpy.percentile", p)
# per = [100 * t for t in thresholds]
from scipy import stats
s1 = stats.scoreatpercentile(targetFP, quantile*100)
h2p.red_print("scipy stats.scoreatpercentile", s1)
# scipy apparently doesn't have the use of means (type 2)
# http://en.wikipedia.org/wiki/Quantile
# it has median (R-8) with 1/3, 1/3
if 1==0:
# type 6
alphap=0
betap=0
# type 5 okay but not perfect
alphap=0.5
betap=0.5
# type 8
alphap=1/3.0
betap=1/3.0
if interpolate=='mean':
# an approx? (was good when comparing to h2o type 2)
alphap=0.4
betap=0.4
if interpolate=='linear':
# this is type 7
alphap=1
betap=1
s2List = stats.mstats.mquantiles(targetFP, prob=quantile, alphap=alphap, betap=betap)
s2 = s2List[0]
# http://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.mstats.mquantiles.html
# type 7
# alphap=0.4, betap=0.4,
# type 2 not available? (mean)
# alphap=1/3.0, betap=1/3.0 is approx median?
h2p.red_print("scipy stats.mstats.mquantiles:", s2)
# also get the median with a painful sort (h2o_summ.percentileOnSortedlist()
# inplace sort
targetFP.sort()
# this matches scipy type 7 (linear)
# b = h2o_summ.percentileOnSortedList(targetFP, 0.50 if DO_MEDIAN else 0.999, interpolate='linear')
# this matches h2o type 2 (mean)
# b = h2o_summ.percentileOnSortedList(targetFP, 0.50 if DO_MEDIAN else 0.999, interpolate='mean')
b = percentileOnSortedList(targetFP, quantile, interpolate='linear')
label = str(quantile * 100) + '%'
h2p.blue_print(label, "from sort:", b)
if SCIPY_INSTALLED:
h2p.blue_print(label, "from numpy:", p)
h2p.blue_print(label, "from scipy 1:", s1)
h2p.blue_print(label, "from scipy 2:", s2)
h2p.blue_print(label, "from h2o summary:", h2oSummary2)
h2p.blue_print(label, "from h2o multipass:"******"from h2o singlepass:"******"from h2o exec:", h2oExecQuantiles)
# they should be identical. keep a tight absolute tolerance
# Note the comparisons have different tolerances, some are relative, some are absolute
if h2oQuantilesExact:
if math.isnan(float(h2oQuantilesExact)):
raise Exception("h2oQuantilesExact is unexpectedly NaN %s" % h2oQuantilesExact)
h2o_util.assertApproxEqual(h2oQuantilesExact, b, tol=0.0000002,
msg='h2o quantile multipass is not approx. same as sort algo')
if h2oQuantilesApprox:
# this can be NaN if we didn't calculate it. turn the NaN string into a float NaN
if math.isnan(float(h2oQuantilesApprox)):
raise Exception("h2oQuantilesApprox is unexpectedly NaN %s" % h2oQuantilesApprox)
if h2oSummary2MaxErr:
h2o_util.assertApproxEqual(h2oQuantilesApprox, b, tol=h2oSummary2MaxErr,
msg='h2o quantile singlepass is not approx. same as sort algo')
else:
h2o_util.assertApproxEqual(h2oQuantilesApprox, b, rel=0.1,
msg='h2o quantile singlepass is not approx. same as sort algo')
if h2oSummary2:
if math.isnan(float(h2oSummary2)):
raise Exception("h2oSummary2 is unexpectedly NaN %s" % h2oSummary2)
if h2oSummary2MaxErr:
# maxErr absolute was calculated in the test from 0.5*(max-min/(max_qbins-2))
h2o_util.assertApproxEqual(h2oSummary2, b, tol=h2oSummary2MaxErr,
msg='h2o summary2 is not approx. same as sort algo (calculated expected max error)')
else:
# bounds are way off, since it depends on the min/max of the col, not the expected value
h2o_util.assertApproxEqual(h2oSummary2, b, rel=1.0,
msg='h2o summary2 is not approx. same as sort algo (sloppy compare)')
if h2oQuantilesApprox and h2oSummary2:
# they should both get the same answer. Currently they have different code, but same algo
# FIX! ...changing to a relative tolerance, since we're getting a miscompare in some cases.
# not sure why..maybe some subtle algo diff.
h2o_util.assertApproxEqual(h2oSummary2, h2oQuantilesApprox, rel=0.04,
msg='h2o summary2 is not approx. same as h2o singlepass.'+\
' Check that max_qbins is 1000 (summary2 is fixed) and type 7 interpolation')
if h2oExecQuantiles:
if math.isnan(float(h2oExecQuantiles)):
raise Exception("h2oExecQuantiles is unexpectedly NaN %s" % h2oExecQuantiles)
# bounds are way off
h2o_util.assertApproxEqual(h2oExecQuantiles, b, rel=1.0,
msg='h2o summary2 is not approx. same as sort algo')
if SCIPY_INSTALLED:
if h2oQuantilesExact:
h2o_util.assertApproxEqual(h2oQuantilesExact, p, tol=0.0000002,
msg='h2o quantile multipass is not same as numpy.percentile')
h2o_util.assertApproxEqual(h2oQuantilesExact, s1, tol=0.0000002,
msg='h2o quantile multipass is not same as scipy stats.scoreatpercentile')
# give us some slack compared to the scipy use of median (instead of desired mean)
# since we don't have bounds here like above, just stop this test for now
if h2oQuantilesApprox and 1==0:
if interpolate=='mean':
h2o_util.assertApproxEqual(h2oQuantilesApprox, s2, rel=0.5,
msg='h2o quantile singlepass is not approx. same as scipy stats.mstats.mquantiles')
else:
h2o_util.assertApproxEqual(h2oQuantilesApprox, s2, rel=0.5,
msg='h2o quantile singlepass is not same as scipy stats.mstats.mquantiles')
# see if scipy changes. nope. it doesn't
if 1==0:
a = stats.mstats.mquantiles(targetFP, prob=quantile, alphap=alphap, betap=betap)
h2p.red_print("after sort")
h2p.red_print("scipy stats.mstats.mquantiles:", s3)
def generate_scipy_comparison(csvPathname, col=0, h2oMedian=None, h2oMedian2=None):
# this is some hack code for reading the csv and doing some percentile stuff in scipy
# from numpy import loadtxt, genfromtxt, savetxt
import numpy as np
import scipy as sp
dataset = np.genfromtxt(
open(csvPathname, 'r'),
delimiter=',',
# skip_header=1,
dtype=None); # guess!
print "csv read for training, done"
# we're going to strip just the last column for percentile work
# used below
NUMCLASSES = 10
print "csv read for training, done"
# data is last column
# drop the output
print dataset.shape
if len(dataset.shape) > 1:
target = [x[col] for x in dataset]
else:
target = dataset
# we may have read it in as a string. coerce to number
targetFP = np.array(target, np.float)
if 1==0:
n_features = len(dataset[0]) - 1;
print "n_features:", n_features
# get the end
# target = [x[-1] for x in dataset]
# get the 2nd col
print "histogram of target"
print target
print sp.histogram(target, bins=NUMCLASSES)
print target[0]
print target[1]
thresholds = [0.001, 0.01, 0.1, 0.25, 0.33, 0.5, 0.66, 0.75, 0.9, 0.99, 0.999]
print "scipy per:", thresholds
from scipy import stats
# a = stats.scoreatpercentile(target, per=per)
a = stats.mstats.mquantiles(targetFP, prob=thresholds)
a2 = ["%.2f" % v for v in a]
h2p.red_print("scipy stats.mstats.mquantiles:", a2)
# also get the median with a painful sort (h2o_summ.percentileOnSortedlist()
# inplace sort
targetFP.sort()
b = h2o_summ.percentileOnSortedList(targetFP, 0.50 if DO_MEDIAN else 0.999, interpolate='linear')
label = '50%' if DO_MEDIAN else '99.9%'
h2p.blue_print(label, "from sort:", b)
s = a[5 if DO_MEDIAN else 10]
h2p.blue_print(label, "from scipy:", s)
h2p.blue_print(label, "from h2o summary2:", h2oMedian)
h2p.blue_print(label, "from h2o quantile multipass:"******"%.2f" % v for v in a]
h2p.red_print("after sort")
h2p.red_print("scipy stats.mstats.mquantiles:", a2)
def test_exec_enums_rand_cut(self):
SYNDATASETS_DIR = h2o.make_syn_dir()
tryList = [
(ROWS, 3, 2, 'cE', 300),
]
# create key names to use for exec
eKeys = ['e%s' % i for i in range(10)]
# h2b.browseTheCloud()
trial = 0
for (rowCount, iColCount, oColCount, hex_key, timeoutSecs) in tryList:
colCount = iColCount + oColCount
hex_key = 'p'
colEnumList = create_col_enum_list(iColCount)
# create 100 possible cut expressions here, so we don't waste time below
rowExprList = []
for j in range(CUT_EXPR_CNT):
print "Creating", CUT_EXPR_CNT, 'cut expressions'
# init cutValue. None means no compare
cutValue = [None for i in range(iColCount)]
# build up a random cut expression
cols = random.sample(range(iColCount), random.randint(1,iColCount))
for c in cols:
# possible choices within the column
# cel = colEnumList[c]
cel = colEnumList
# for now the cutValues are numbers for the enum mappings
if 1==1:
# FIX! hack. don't use encoding 0, maps to NA here? h2o doesn't like
celChoice = str(random.choice(range(len(cel))))
else:
celChoice = random.choice(cel)
cutValue[c] = celChoice
cutExprList = []
for i,c in enumerate(cutValue):
if c is None:
continue
else:
# new ...ability to reference cols
# src[ src$age<17 && src$zip=95120 && ... , ]
cutExprList.append('p$C'+str(i+1)+'=='+c)
cutExpr = ' && '.join(cutExprList)
print "cutExpr:", cutExpr
# should be two different keys in the sample
e = random.sample(eKeys,2)
fKey = e[0]
eKey = e[1]
rowExpr = '%s[%s,];' % (hex_key, cutExpr)
print "rowExpr:", rowExpr
rowExprList.append(rowExpr)
print "j:", j
# CREATE DATASET*******************************************
SEEDPERFILE = random.randint(0, sys.maxint)
csvFilename = 'syn_enums_' + str(rowCount) + 'x' + str(colCount) + '.csv'
csvPathname = SYNDATASETS_DIR + '/' + csvFilename
print "Creating random", csvPathname
write_syn_dataset(csvPathname, rowCount, iColCount, oColCount, SEEDPERFILE, colEnumList=colEnumList)
# PARSE*******************************************************
parseResult = h2i.import_parse(path=csvPathname, schema='put', hex_key=hex_key, timeoutSecs=30, doSummary=False, header=0)
print "Parse result['destination_key']:", parseResult['destination_key']
inspect = h2o_cmd.runInspect(key=parseResult['destination_key'])
h2o_cmd.infoFromInspect(inspect, csvPathname)
# print h2o.dump_json(inspect)
rSummary = h2o_cmd.runSummary(key=parseResult['destination_key'])
h2o_cmd.infoFromSummary(rSummary)
(missingValuesDict, constantValuesDict, enumSizeDict, colTypeDict, colNameDict) = \
h2o_cmd.columnInfoFromInspect(parseResult['destination_key'], exceptionOnMissingValues=False)
# error if any col has constant values
if len(constantValuesDict) != 0:
raise Exception("Probably got a col NA'ed and constant values as a result %s" % constantValuesDict)
# INIT all possible key names used***************************
# remember. 1 indexing!
# is this needed?
if 1==1:
a = 'a=c(1,2,3);' + ';'.join(['a[,%s]=a[,%s-1]'% (i,i) for i in range(2,colCount)])
print a
for eKey in eKeys:
# build up the columns
e = h2o.nodes[0].exec_query(str='%s;%s=a' % (a, eKey), print_params=False)
## print h2o.dump_json(e)
xList = []
eList = []
fList = []
for repeat in range(200):
# EXEC*******************************************************
# don't use exec_expr to avoid issues with Inspect following etc.
randICol = random.randint(0,iColCount-1)
randOCol = random.randint(iColCount, iColCount+oColCount-1)
# should be two different keys in the sample
e = random.sample(eKeys,2)
fKey = e[0]
eKey = e[1]
if 1==0:
start = time.time()
e = h2o.nodes[0].exec_query(str='%s=%s[,%s]' % (fKey, hex_key, randOCol+1))
elapsed = time.time() - start
print "exec 1 took", elapsed, "seconds."
execTime = elapsed
if 1==1:
start = time.time()
h2o.nodes[0].exec_query(str="%s=%s" % (fKey, random.choice(rowExprList)))
elapsed = time.time() - start
execTime = elapsed
print "exec 2 took", elapsed, "seconds."
if 1==0:
gKey = random.choice(eKeys)
# do a 2nd random to see if things blow up
start = time.time()
h2o.nodes[0].exec_query(str="%s=%s" % (gKey, fKey))
elapsed = time.time() - start
print "exec 3 took", elapsed, "seconds."
if 1==1:
inspect = h2o_cmd.runInspect(key=fKey)
h2o_cmd.infoFromInspect(inspect, fKey)
numRows = inspect['numRows']
numCols = inspect['numCols']
if numRows==0 or numCols!=colCount:
h2p.red_print("Warning: Cut resulted in", numRows, "rows and", numCols, "cols. Quantile will abort")
# QUANTILE*******************************************************
quantile = 0.5 if DO_MEDIAN else .999
# first output col. always fed by an exec cut, so 0?
column = iColCount
start = time.time()
q = h2o.nodes[0].quantiles(source_key=fKey, column=column,
quantile=quantile, max_qbins=MAX_QBINS, multiple_pass=MULTI_PASS)
h2p.red_print("quantile", quantile, q['result'])
elapsed = time.time() - start
print "quantile end on ", fKey, 'took', elapsed, 'seconds.'
quantileTime = elapsed
# remove all keys*******************************************************
# what about hex_key?
if 1==0:
start = time.time()
h2o.nodes[0].remove_all_keys()
elapsed = time.time() - start
print "remove all keys end on ", csvFilename, 'took', elapsed, 'seconds.'
trial += 1
xList.append(trial)
eList.append(execTime)
fList.append(quantileTime)
# just get a plot of the last one (biggest)
if DO_PLOT:
xLabel = 'trial'
eLabel = 'exec cut time'
fLabel = 'quantile time'
eListTitle = ""
fListTitle = ""
h2o_gbm.plotLists(xList, xLabel, eListTitle, eList, eLabel, fListTitle, fList, fLabel)
def build_cloud(node_count=1, base_port=None, hosts=None,
timeoutSecs=30, retryDelaySecs=1, cleanup=True, rand_shuffle=True,
conservative=False, create_json=False, clone_cloud=None,
init_sandbox=True, usecloud=False, usecloud_size=None, **kwargs):
# expectedSize is only used if usecloud
# usecloud can be passed thru build_cloud param, or command line
# not in config json though so no build_cloud_with_hosts path.
# redirect to build_cloud_with_json if a command line arg
# wants to force a test to ignore it's build_cloud/build_cloud_with_hosts
# (both come thru here)
# clone_cloud is just another way to get the effect (maybe ec2 config file thru
# build_cloud_with_hosts?
global stdout_wrapped
if not h2o_args.disable_time_stamp and not stdout_wrapped:
sys.stdout = OutWrapper(sys.stdout)
stdout_wrapped = True
if h2o_args.usecloud or usecloud:
# for now, just have fixed name in local file. (think of this as a temp or debug file)
# eventually we'll pass the json object instead for speed?
nodesJsonPathname = "h2o_fc-nodes.json"
elif h2o_args.clone_cloud_json:
nodesJsonPathname = h2o_args.clone_cloud_json
elif clone_cloud:
nodesJsonPathname = clone_cloud
else:
# normal build_cloud() doesn't use
nodesJsonPathname = None
# usecloud dominates over all
if (h2o_args.clone_cloud_json or clone_cloud) or (h2o_args.usecloud or usecloud):
# then build_cloud_with_json with json object
# we don't need to specify these defaults, but leave here to show that we can pass
# I suppose kwargs will have it
if h2o_args.usecloud:
ip_port = h2o_args.usecloud
elif usecloud:
ip_port = usecloud
else:
ip_port = None
# h2o_args dominates
if h2o_args.usecloud_size:
# only used for expected size
useCloudExpectedSize = h2o_args.usecloud_size
else:
useCloudExpectedSize = usecloud_size
nodesJsonObject = h2o_fc.find_cloud(ip_port=ip_port,
expectedSize=useCloudExpectedSize, nodesJsonPathname=nodesJsonPathname, **kwargs)
# potentially passed in kwargs
# hdfs_version='cdh4', hdfs_config=None, hdfs_name_node='172.16.1.176',
nodeList = build_cloud_with_json(h2o_nodes_json=nodesJsonPathname)
return nodeList
# else
# moved to here from unit_main. so will run with nosetests too!
# Normally do this.
# Don't if build_cloud_with_hosts() did and put a flatfile in there already!
if init_sandbox:
clean_sandbox()
log("#*********************************************************************")
log("Starting new test: " + h2o_args.python_test_name + " at build_cloud() ")
log("#*********************************************************************")
# start up h2o to report the java version (once). output to python stdout
# only do this for regression testing
# temporarily disable this, to go a little faster
# if getpass.getuser() == 'jenkins':
# check_h2o_version()
ports_per_node = 2
nodeList = []
# shift the port used to run groups of tests on the same machine at the same time?
base_port = get_base_port(base_port)
try:
# if no hosts list, use psutil method on local host.
totalNodes = 0
# doing this list outside the loops so we can shuffle for better test variation
# this jvm startup shuffle is independent from the flatfile shuffle
portList = [base_port + ports_per_node * i for i in range(node_count)]
if hosts is None:
# if use_flatfile, we should create it
# because tests will just call build_cloud with use_flatfile=True
# best to just create it all the time..may or may not be used
write_flatfile(node_count=node_count, base_port=base_port)
hostCount = 1
if rand_shuffle:
random.shuffle(portList)
for p in portList:
verboseprint("psutil starting node", i)
newNode = LocalH2O(port=p, node_id=totalNodes, **kwargs)
nodeList.append(newNode)
totalNodes += 1
else:
# if hosts, the flatfile was created and uploaded to hosts already
# I guess don't recreate it, don't overwrite the one that was copied beforehand.
# we don't always use the flatfile (use_flatfile=False)
# Suppose we could dispatch from the flatfile to match it's contents
# but sometimes we want to test with a bad/different flatfile then we invoke h2o?
hostCount = len(hosts)
hostPortList = []
for h in hosts:
for port in portList:
hostPortList.append((h, port))
if rand_shuffle: random.shuffle(hostPortList)
for (h, p) in hostPortList:
verboseprint('ssh starting node', totalNodes, 'via', h)
newNode = h.remote_h2o(port=p, node_id=totalNodes, **kwargs)
nodeList.append(newNode)
totalNodes += 1
verboseprint("Attempting Cloud stabilize of", totalNodes, "nodes on", hostCount, "hosts")
start = time.time()
# UPDATE: best to stabilize on the last node!
stabilize_cloud(nodeList[0], nodeList,
timeoutSecs=timeoutSecs, retryDelaySecs=retryDelaySecs, noSandboxErrorCheck=True)
verboseprint(len(nodeList), "Last added node stabilized in ", time.time() - start, " secs")
verboseprint("Built cloud: %d nodes on %d hosts, in %d s" % \
(len(nodeList), hostCount, (time.time() - start)))
h2p.red_print("Built cloud:", nodeList[0].java_heap_GB, "GB java heap(s) with",
len(nodeList), "total nodes")
# FIX! using "consensus" in node[-1] should mean this is unnecessary?
# maybe there's a bug. For now do this. long term: don't want?
# UPDATE: do it for all cases now 2/14/13
if conservative: # still needed?
for n in nodeList:
stabilize_cloud(n, nodeList, timeoutSecs=timeoutSecs, noSandboxErrorCheck=True)
# this does some extra checking now
# verifies cloud name too if param is not None
verify_cloud_size(nodeList, expectedCloudName=nodeList[0].cloud_name)
# best to check for any errors due to cloud building right away?
check_sandbox_for_errors(python_test_name=h2o_args.python_test_name)
except:
# nodeList might be empty in some exception cases?
# no shutdown issued first, though
if cleanup and nodeList:
for n in nodeList: n.terminate()
check_sandbox_for_errors(python_test_name=h2o_args.python_test_name)
raise
print len(nodeList), "total jvms in H2O cloud"
# put the test start message in the h2o log, to create a marker
nodeList[0].h2o_log_msg()
if h2o_args.config_json:
LOG_DIR = get_sandbox_name()
# like cp -p. Save the config file, to sandbox
print "Saving the ", h2o_args.config_json, "we used to", LOG_DIR
shutil.copy(h2o_args.config_json, LOG_DIR + "/" + os.path.basename(h2o_args.config_json))
# Figure out some stuff about how this test was run
cs_time = str(datetime.datetime.now())
cs_cwd = os.getcwd()
cs_python_cmd_line = "python %s %s" % (h2o_args.python_test_name, h2o_args.python_cmd_args)
cs_python_test_name = h2o_args.python_test_name
if h2o_args.config_json:
cs_config_json = os.path.abspath(h2o_args.config_json)
else:
cs_config_json = None
cs_username = h2o_args.python_username
cs_ip = h2o_args.python_cmd_ip
# dump the nodes state to a json file # include enough extra info to have someone
# rebuild the cloud if a test fails that was using that cloud.
if create_json:
q = {
'cloud_start':
{
'time': cs_time,
'cwd': cs_cwd,
'python_test_name': cs_python_test_name,
'python_cmd_line': cs_python_cmd_line,
'config_json': cs_config_json,
'username': cs_username,
'ip': cs_ip,
},
'h2o_nodes': h2o_util.json_repr(nodeList),
}
with open('h2o-nodes.json', 'w+') as f:
f.write(json.dumps(q, indent=4))
# save it to a local global copy, in case it's needed for tearDown
h2o_nodes.nodes[:] = nodeList
return nodeList
def test_exec2_enums_rand_cut(self):
SYNDATASETS_DIR = h2o.make_syn_dir()
n = ROWS
tryList = [
(n, 10, 9, 'cE', 300),
]
# create key names to use for exec
eKeys = ['e%s' % i for i in range(10)]
# h2b.browseTheCloud()
trial = 0
for (rowCount, iColCount, oColCount, hex_key, timeoutSecs) in tryList:
colCount = iColCount + oColCount
hex_key = 'p'
colEnumList = create_col_enum_list(iColCount)
# create 100 possible cut expressions here, so we don't waste time below
rowExprList = []
print "Creating", CUT_EXPR_CNT, 'cut expressions'
for j in range(CUT_EXPR_CNT):
# init cutValue. None means no compare
cutValue = [None for i in range(iColCount)]
# build up a random cut expression
cols = random.sample(range(iColCount),
random.randint(1, iColCount))
for c in cols:
# possible choices within the column
cel = colEnumList[c]
# for now the cutValues are numbers for the enum mappings
# FIX! hack. don't use encoding 0, maps to NA here? h2o doesn't like
# celChoice = str(random.choice(range(len(cel))))
celChoice = random.choice(range(len(cel)))
cutValue[c] = celChoice
cutExprList = []
pKey = Key('p')
for i, c in enumerate(cutValue):
if c is None:
continue
else:
# new ...ability to reference cols
# src[ src$age<17 && src$zip=95120 && ... , ]
# cutExprList.append('p$C'+str(i+1)+'=='+c)
# all column indexing in h2o-dev is with number
e = Fcn('==', c, pKey[:, i])
cutExprList.append(e)
cutExpr = None
for ce in cutExprList:
if cutExpr:
cutExpr = Fcn('&', cutExpr, ce)
else:
cutExpr = ce
print "cutExpr:", cutExpr
# should be two different keys in the sample
e = random.sample(eKeys, 2)
fKey = e[0]
eKey = e[1]
# rowExpr = '%s[%s,];' % (hex_key, cutExpr)
hKey = Key(hex_key)
rowExpr = hKey[cutExpr, :]
print "rowExpr:", rowExpr
rowExprList.append(rowExpr)
# CREATE DATASET*******************************************
SEEDPERFILE = random.randint(0, sys.maxint)
csvFilename = 'syn_enums_' + str(rowCount) + 'x' + str(
colCount) + '.csv'
csvPathname = SYNDATASETS_DIR + '/' + csvFilename
print "Creating random", csvPathname
write_syn_dataset(csvPathname,
rowCount,
iColCount,
oColCount,
SEEDPERFILE,
colEnumList=colEnumList)
# PARSE*******************************************************
parseResult = h2i.import_parse(path=csvPathname,
schema='put',
hex_key=hex_key,
timeoutSecs=30)
numRows, numCols, parse_key = h2o_cmd.infoFromParse(parseResult)
inspect = h2o_cmd.runInspect(key=parse_key)
missingList, valueList, numRows, numCols = h2o_cmd.infoFromInspect(
inspect)
# print h2o.dump_json(inspect)
# (missingValuesDict, constantValuesDict, enumSizeDict, colTypeDict, colNameDict) = \
# h2o_cmd.columnInfoFromInspect(parse_key, exceptionOnMissingValues=False)
# error if any col has constant values
# if len(constantValuesDict) != 0:
# raise Exception("Probably got a col NA'ed and constant values as a result %s" % constantValuesDict)
# INIT all possible key names used***************************
# remember. 1 indexing!
# build up the columns
Assign('b', [1, 2, 3])
# could also append 1 col at a time, by assigning to the next col number?
Assign('a', Cbind(['b' for i in range(colCount)]))
for eKey in eKeys:
Assign(eKey, 'a')
## print h2o.dump_json(e)
xList = []
eList = []
fList = []
for repeat in range(200):
# EXEC*******************************************************
# don't use exec_expr to avoid issues with Inspect following etc.
randICol = random.randint(0, iColCount - 1)
randOCol = random.randint(iColCount, iColCount + oColCount - 1)
# should be two different keys in the sample
e = random.sample(eKeys, 2)
fKey = e[0]
eKey = e[1]
if 1 == 1:
start = time.time()
Assign(fKey, random.choice(rowExprList)).do()
elapsed = time.time() - start
execTime = elapsed
print "exec 2 took", elapsed, "seconds."
inspect = h2o_cmd.runInspect(key=fKey)
missingList, valueList, numRows, numCols = h2o_cmd.infoFromInspect(
inspect)
if numRows == 0 or numCols != colCount:
h2p.red_print("Warning: Cut resulted in", numRows,
"rows and", numCols,
"cols. Quantile will abort")
# FIX! put quantile back in?
quantileTime = 0
# remove all keys*******************************************************
# what about hex_key?
if 1 == 0:
start = time.time()
h2o.nodes[0].remove_all_keys()
elapsed = time.time() - start
print "remove all keys end on ", csvFilename, 'took', elapsed, 'seconds.'
trial += 1
xList.append(trial)
eList.append(execTime)
fList.append(quantileTime)
# just get a plot of the last one (biggest)
if DO_PLOT:
xLabel = 'trial'
eLabel = 'exec cut time'
fLabel = 'quantile time'
eListTitle = ""
fListTitle = ""
h2o_gbm.plotLists(xList, xLabel, eListTitle, eList, eLabel,
fListTitle, fList, fLabel)
def test_quant_cols(self):
h2o.beta_features = True
SYNDATASETS_DIR = h2o.make_syn_dir()
if getpass.getuser()=='kevin':
tryList = [
('home-0xdiag-datasets', 'airlines/year2013.csv', None, None, 'cE', 300),
(None, '/home/kevin/Downloads/t.csv', 15, 11, 'cE', 300),
]
else:
tryList = [
('home-0xdiag-datasets', 'airlines/year2013.csv', None, None, 'cE', 300),
]
# h2b.browseTheCloud()
trial = 0
for (bucket, csvPathname, iColCount, oColCount, hex_key, timeoutSecs) in tryList:
xList = []
eList = []
fList = []
# PARSE*******************************************************
parseResult = h2i.import_parse(bucket=bucket, path=csvPathname, schema='put', hex_key=hex_key, timeoutSecs=200, doSummary=False)
csvPathnameFull = h2i.find_folder_and_filename(bucket, csvPathname, returnFullPath=True)
print "Parse result['destination_key']:", parseResult['destination_key']
inspect = h2o_cmd.runInspect(key=parseResult['destination_key'])
h2o_cmd.infoFromInspect(inspect, csvPathname)
numRows = inspect['numRows']
numCols = inspect['numCols']
if not oColCount:
iColCount = 0
if not oColCount:
oColCount = numCols
colCount = iColCount + oColCount
for i in range (0,numCols):
print "Column", i, "summary"
h2o_cmd.runSummary(key=hex_key, max_qbins=1, cols=i);
# print h2o.dump_json(inspect)
levels = h2o.nodes[0].levels(source=hex_key)
print "levels result:", h2o.dump_json(levels)
(missingValuesDict, constantValuesDict, enumSizeDict, colTypeDict, colNameDict) = \
h2o_cmd.columnInfoFromInspect(parseResult['destination_key'], exceptionOnMissingValues=False)
# error if any col has constant values
if len(constantValuesDict) != 0:
# raise Exception("Probably got a col NA'ed and constant values as a result %s" % constantValuesDict)
print "Probably got a col NA'ed and constant values as a result %s" % constantValuesDict
# start after the last input col
levels = h2o.nodes[0].levels(source=hex_key);
l = levels['levels']
for column in range(iColCount, iColCount+oColCount):
if l[column]:
print "Skipping", column, "because it's enum (says levels)"
continue
# QUANTILE*******************************************************
quantile = 0.5 if DO_MEDIAN else .999
# first output col. always fed by an exec cut, so 0?
start = time.time()
# file has headers. use col index
q = h2o.nodes[0].quantiles(source_key=hex_key, column=column,
quantile=quantile, max_qbins=MAX_QBINS, multiple_pass=1)
qresult = q['result']
h2p.red_print("result:", q['result'], "quantile", quantile,
"interpolated:", q['interpolated'], "iterations", q['iterations'])
elapsed = time.time() - start
print "quantile end on ", hex_key, 'took', elapsed, 'seconds.'
quantileTime = elapsed
# don't do for enums
# also get the median with a sort (h2o_summ.percentileOnSortedlist()
if 1==0:
h2o_summ.quantile_comparisons(
csvPathnameFull,
skipHeader=True,
col=column, # what col to extract from the csv
datatype='float',
quantile=0.5 if DO_MEDIAN else 0.999,
# h2oSummary2=pctile[5 if DO_MEDIAN else 10],
# h2oQuantilesApprox=qresult_single,
h2oQuantilesExact=qresult,
use_genfromtxt=True,
)
trial += 1
execTime = 0
xList.append(column)
eList.append(execTime)
fList.append(quantileTime)
# remove all keys*******************************************************
# what about hex_key?
if 1==0:
start = time.time()
h2o.nodes[0].remove_all_keys()
elapsed = time.time() - start
print "remove all keys end on took", elapsed, 'seconds.'
#****************************************************************
# PLOTS. look for eplot.jpg and fplot.jpg in local dir?
if DO_PLOT:
xLabel = 'column (0 is first)'
eLabel = 'exec cut time'
fLabel = 'quantile time'
eListTitle = ""
fListTitle = ""
h2o_gbm.plotLists(xList, xLabel, eListTitle, eList, eLabel, fListTitle, fList, fLabel, server=True)
def find_folder_and_filename(bucket, pathWithRegex, schema='put', returnFullPath=False):
checkPath = True
# strip the common mistake of leading "/" in path, if bucket is specified too
giveUpAndSearchLocally = False
if bucket is not None and re.match("/", pathWithRegex):
h2o.verboseprint("You said bucket:", bucket, "so stripping incorrect leading '/' from", pathWithRegex)
pathWithRegex = pathWithRegex.lstrip('/')
if bucket is None: # good for absolute path name
bucketPath = ""
elif bucket == ".":
bucketPath = os.getcwd()
# only use if the build_cloud was for remote H2O
# Never use the var for remote, if you're doing a put! (which always sources local)
elif h2o.nodes[0].remoteH2O and schema!='put' and \
(os.environ.get('H2O_REMOTE_BUCKETS_ROOT') or h2o.nodes[0].h2o_remote_buckets_root):
if (bucket=='smalldata' or bucket=='datasets') and schema=='local':
msg1 = "\nWARNING: you're using remote nodes, and 'smalldata' or 'datasets' git buckets, with schema!=put"
msg2 = "\nThose aren't git pull'ed by the test. Since they are user-maintained, not globally-maintained-by-0xdata,"
msg3 = "\nthey may be out of date at those remote nodes?"
msg4 = "\nGoing to assume we find a path to them locally, and remote path will be the same"
h2p.red_print(msg1, msg2, msg3, msg4)
giveUpAndSearchLocally = True
else:
if os.environ.get('H2O_REMOTE_BUCKETS_ROOT'):
rootPath = os.environ.get('H2O_REMOTE_BUCKETS_ROOT')
print "Found H2O_REMOTE_BUCKETS_ROOT:", rootPath
else:
rootPath = h2o.nodes[0].h2o_remote_buckets_root
print "Found h2o_nodes[0].h2o_remote_buckets_root:", rootPath
bucketPath = os.path.join(rootPath, bucket)
checkPath = False
# does it work to use bucket "." to get current directory
# this covers reote with put too
elif os.environ.get('H2O_BUCKETS_ROOT'):
rootPath = os.environ.get('H2O_BUCKETS_ROOT')
print "Using H2O_BUCKETS_ROOT environment variable:", rootPath
if not (os.path.exists(rootPath)):
raise Exception("H2O_BUCKETS_ROOT in env but %s doesn't exist." % rootPath)
bucketPath = os.path.join(rootPath, bucket)
if not (os.path.exists(bucketPath)):
raise Exception("H2O_BUCKETS_ROOT and path used to form %s which doesn't exist." % bucketPath)
else:
giveUpAndSearchLocally = True
#******************************************************************************************
if giveUpAndSearchLocally:
# if we run remotely, we're assuming the import folder path on the remote machine
# matches what we find on our local machine. But maybe the local user doesn't exist remotely
# so using his path won't work.
# Resolve by looking for special state in the config. If user = 0xdiag, just force the bucket location
# This is a lot like knowing about fixed paths with s3 and hdfs
# Otherwise the remote path needs to match the local discovered path.
# want to check the username being used remotely first. should exist here too if going to use
username = getpass.getuser()
h2oUsername = h2o.nodes[0].username
h2o.verboseprint("username:"******"h2oUsername:"******"datasets" is special. Don't want to find it in /home/0xdiag/datasets
# needs to be the git clone 'datasets'. Find it by walking upwards below
# disable it from this looking in home dir. Could change priority order?
# resolved in order, looking for bucket (ln -s will work) in these home dirs.
if bucket=='datasets': # special case
possibleUsers = []
elif h2oUsername != username:
possibleUsers = [username, h2oUsername, "0xdiag"]
else:
possibleUsers = [username, "0xdiag"]
for u in possibleUsers:
rootPath = os.path.expanduser("~" + u)
bucketPath = os.path.join(rootPath, bucket)
h2o.verboseprint("Checking bucketPath:", bucketPath, 'assuming home is', rootPath)
if os.path.exists(bucketPath):
h2o.verboseprint("search A did find", bucket, "at", rootPath)
break
else:
# last chance to find it by snooping around
rootPath = os.getcwd()
h2o.verboseprint("find_bucket looking upwards from", rootPath, "for", bucket)
# don't spin forever
levels = 0
while not (os.path.exists(os.path.join(rootPath, bucket))):
h2o.verboseprint("Didn't find", bucket, "at", rootPath)
rootPath = os.path.split(rootPath)[0]
levels += 1
if (levels==6):
raise Exception("unable to find bucket: %s. Maybe missing link in /home/0xdiag or /home/0xcustomer or jenkins ~? or whatever user is running the python or the h2o?" % bucket)
h2o.verboseprint("search B did find", bucket, "at", rootPath)
bucketPath = os.path.join(rootPath, bucket)
#******************************************************************************************
# if there's no path, just return the bucketPath
# but what about cases with a header in the folder too? (not putfile)
if pathWithRegex is None:
if returnFullPath:
return bucketPath
else:
return (bucketPath, None)
# if there is a "/" in the path, that means it's not just a pattern
# split it
# otherwise it is a pattern. use it to search for files in python first?
# FIX! do that later
elif "/" in pathWithRegex:
(head, tail) = os.path.split(pathWithRegex)
folderPath = os.path.abspath(os.path.join(bucketPath, head))
# accept all 0xcustomer-datasets without checking..since the current python user
# may not have permission, but h2o will
# try a couple times with os.stat in between, in case it's not automounting
if '/mnt/0xcustomer-datasets' in folderPath:
pass
else:
retry = 0
while checkPath and (not os.path.exists(folderPath)) and retry<5:
# we can't stat an actual file, because we could have a regex at the end of the pathname
print "Retrying", folderPath, "in case there's a autofs mount problem"
os.stat(folderPath)
retry += 1
time.sleep(1)
if checkPath and not os.path.exists(folderPath):
raise Exception("%s doesn't exist. %s under %s may be wrong?" % (folderPath, head, bucketPath))
else:
folderPath = bucketPath
tail = pathWithRegex
h2o.verboseprint("folderPath:", folderPath, "tail:", tail)
if returnFullPath:
return os.path.join(folderPath, tail)
else:
return (folderPath, tail)
def __do_json_request(self, jsonRequest=None, fullUrl=None, timeout=10, params=None, postData=None, returnFast=False,
cmd='get', extraComment=None, ignoreH2oError=False, noExtraErrorCheck=False, raiseIfNon200=True, suppressErrorMsg=False, **kwargs):
H2O.verboseprint("__do_json_request, timeout: " + str(timeout))
# if url param is used, use it as full url. otherwise crate from the jsonRequest
if fullUrl:
url = fullUrl
else:
url = self.__url(jsonRequest)
# remove any params that are 'None'
# need to copy dictionary, since can't delete while iterating
if params is not None:
params_serialized = params.copy()
for k in params_serialized:
if params_serialized[k] is None:
del params[k]
paramsStr = '?' + '&'.join(['%s=%s' % (k, v) for (k, v) in params.items()])
else:
paramsStr = ''
# The requests package takes array parameters and explodes them: ['f00', 'b4r'] becomes "f00,b4r".
# NOTE: this handles 1D arrays only; if we need ND this needs to be recursive.
# NOTE: we currently don't need to do this for GET, so that's not implemented.
if postData is not None:
munged_postData = {}
for k, v in postData.iteritems():
if type(v) is list:
if len(v) == 0:
munged_postData[k] = '[]'
else:
first = True
array_str = '['
for val in v:
if not first: array_str += ', '
if val is None:
array_str += 'null'
elif isinstance(val, basestring):
array_str += "\"" + str(val) + "\""
else:
array_str += str(val)
first = False
array_str += ']'
munged_postData[k] = array_str
elif type(v) is dict:
if len(v) == 0:
munged_postData[k] = '{}'
else:
first = True
map_str = '{'
for key, val in v.iteritems():
if not first: map_str += ', '
if val is None:
map_str += "\"" + key + "\"" + ': null'
elif isinstance(val, basestring):
map_str += "\"" + str(key) + "\"" + ":" + "\"" + str(val) + "\""
else:
map_str += "\"" + key + "\"" + ':' + str(val)
first = False
map_str += '}'
munged_postData[k] = map_str
else:
# not list:
munged_postData[k] = v
else:
# None
munged_postData = postData
# print("munged_postData: " + repr(munged_postData))
if extraComment:
log('Start ' + url + paramsStr, comment=extraComment)
else:
log('Start ' + url + paramsStr)
log_rest("")
log_rest("----------------------------------------------------------------------\n")
if extraComment:
log_rest("# Extra comment info about this request: " + extraComment)
if cmd == 'get':
log_rest("GET")
else:
log_rest("POST")
log_rest(url + paramsStr)
# file get passed thru kwargs here
try:
if 'post' == cmd:
# NOTE == cmd: for now, since we don't have deserialization from JSON in h2o-dev, we use form-encoded POST.
# This is temporary.
#
# This following does application/json (aka, posting JSON in the body):
# r = requests.post(url, timeout=timeout, params=params, data=json.dumps(munged_postData), **kwargs)
#
# This does form-encoded, which doesn't allow POST of nested structures
r = requests.post(url, timeout=timeout, params=params, data=munged_postData, **kwargs)
elif 'delete' == cmd:
r = requests.delete(url, timeout=timeout, params=params, **kwargs)
elif 'get' == cmd:
r = requests.get(url, timeout=timeout, params=params, **kwargs)
else:
raise ValueError("Unknown HTTP command (expected 'get', 'post' or 'delete'): " + cmd)
except Exception as e:
# rethrow the exception after we've checked for stack trace from h2o
# out of memory errors maybe don't show up right away? so we should wait for h2o
# to get it out to h2o stdout. We don't want to rely on cloud teardown to check
# because there's no delay, and we don't want to delay all cloud teardowns by waiting.
# (this is new/experimental)
exc_info = sys.exc_info()
# use this to ignore the initial connection errors during build cloud when h2o is coming up
if not noExtraErrorCheck:
h2p.red_print(
"ERROR: got exception on %s to h2o. \nGoing to check sandbox, then rethrow.." % (url + paramsStr))
time.sleep(2)
H2O.check_sandbox_for_errors(python_test_name=H2O.python_test_name);
log_rest("")
log_rest("EXCEPTION CAUGHT DOING REQUEST: " + str(e.message))
raise (exc_info[1], None, exc_info[2])
H2O.verboseprint("r: " + repr(r))
if 200 != r.status_code:
pp = pprint.PrettyPrinter(indent=4)
msg = "JSON call returned non-200 status: " + url
json = r.json()
if None != json and 'dev_msg' in json:
msg += "\ndev_msg: "
msg += str(json['dev_msg'])
msg += "\nr.status_code: " + str(r.status_code)
msg += "\nr.headers: " + repr(r.headers)
if None == json:
msg += '\nERROR: the error output from H2O is not JSON!'
msg += "\nr.text: " + r.text
else:
msg += "\nr.json: "
msg += pp.pformat(json)
if raiseIfNon200:
pass # we'll pass msg up with the exception
elif not suppressErrorMsg:
print(msg)
log_rest(msg)
log_rest("")
try:
if r is None:
log_rest("r is None")
else:
log_rest("HTTP status code: " + str(r.status_code))
# The following accesses to r.text were taking most of the runtime:
log_text = False
if log_text:
if hasattr(r, 'text'):
if r.text is None:
log_rest("r.text is None")
else:
log_rest(r.text)
else:
log_rest("r does not have attr text")
except Exception as e:
# Paranoid exception catch.
# Ignore logging exceptions in the case that the above error checking isn't sufficient.
print("Caught exception from result logging: ", e, "; result: ", repr(r))
# fatal if no response
if raiseIfNon200 and not r:
raise Exception("Maybe bad url? no r in __do_json_request in %s:" % inspect.stack()[1][3] + "\n\n" + msg)
# this is used to open a browser on results, or to redo the operation in the browser
# we don't' have that may urls flying around, so let's keep them all
H2O.json_url_history.append(r.url)
# if r.json():
# raise Exception("Maybe bad url? no r.json in __do_json_request in %s:" % inspect.stack()[1][3])
rjson = None
if returnFast:
return
try:
rjson = r.json()
except:
print(h2o_test_utils.dump_json(r.text))
if not isinstance(r, (list, dict)):
raise Exception("h2o json responses should always be lists or dicts, see previous for text")
raise Exception("Could not decode any json from the request.")
# TODO
# TODO
# TODO
# TODO: we should really only look in the response object. This check
# prevents us from having a field called "error" (e.g., for a scoring result).
for e in ['error', 'Error', 'errors', 'Errors']:
# error can be null (python None). This happens in exec2
if e in rjson and rjson[e]:
H2O.verboseprint("rjson:" + h2o_test_utils.dump_json(rjson))
emsg = 'rjson %s in %s: %s' % (e, inspect.stack()[1][3], rjson[e])
if ignoreH2oError:
# well, we print it..so not totally ignore. test can look at rjson returned
print(emsg)
else:
print(emsg)
raise Exception(emsg)
for w in ['warning', 'Warning', 'warnings', 'Warnings']:
# warning can be null (python None).
if w in rjson and rjson[w]:
H2O.verboseprint(dump_json(rjson))
print('rjson %s in %s: %s' % (w, inspect.stack()[1][3], rjson[w]))
# Allow the caller to check things like __http_request.status_code.
# The response object is not JSON-serializable, so we capture the fields we want here:
response = {}
# response['headers'] = r.headers
response['url'] = r.url
response['status_code'] = r.status_code
response['text'] = r.text
rjson['__http_response'] = response
return rjson
def do_statsmodels_glm(self, bucket, csvPathname, L, family='gaussian'):
h2p.red_print("Now doing statsmodels")
h2p.red_print(
"http://statsmodels.sourceforge.net/devel/glm.html#module-reference")
h2p.red_print(
"http://statsmodels.sourceforge.net/devel/generated/statsmodels.genmod.generalized_linear_model.GLM.html"
)
import numpy as np
import scipy as sp
from numpy import loadtxt
import statsmodels as sm
csvPathnameFull = h2i.find_folder_and_filename(bucket,
csvPathname,
returnFullPath=True)
if 1 == 1:
dataset = np.loadtxt(
open(csvPathnameFull, 'r'),
skiprows=1, # skip the header
delimiter=',',
dtype='float')
# skipping cols from the begining... (ID is col 1)
# In newer versions of Numpy, np.genfromtxt can take an iterable argument,
# so you can wrap the file you're reading in a generator that generates lines,
# skipping the first N columns. If your numbers are comma-separated, that's something like
if 1 == 0:
f = open(csvPathnameFull, 'r'),
np.genfromtxt(
(",".join(ln.split()[1:]) for ln in f),
skiprows=1, # skip the header
delimiter=',',
dtype='float')
print "\ncsv read for training, done"
# data is last column
# drop the output
n_features = len(dataset[0]) - 1
print "n_features:", n_features
# don't want ID (col 0) or CAPSULE (col 1)
# get CAPSULE
target = [x[1] for x in dataset]
# slice off the first 2
train = np.array([x[2:] for x in dataset])
n_samples, n_features = train.shape
print "n_samples:", n_samples, "n_features:", n_features
print "histogram of target"
print sp.histogram(target, 3)
print "len(train):", len(train)
print "len(target):", len(target)
print "dataset shape:", dataset.shape
if family != 'gaussian':
raise Exception("Only have gaussian logistic for scipy")
# train the classifier
gauss_log = sm_api.GLM(target,
train,
family=sm_api.families.Gaussian(
sm_api.families.links.log))
start = time.time()
gauss_log_results = gauss_log.fit()
print "sm_api.GLM took", time.time() - start, "seconds"
print gauss_log_results.summary()
def test_quant_cols(self):
h2o.beta_features = True
SYNDATASETS_DIR = h2o.make_syn_dir()
if getpass.getuser() == 'kevin':
tryList = [
(None, '/home/kevin/Downloads/t.csv', 15, 11, 'cE', 300),
('home-0xdiag-datasets', 'airlines/year2013.csv', None, None,
'cE', 300),
]
else:
tryList = [
('home-0xdiag-datasets', 'airlines/year2013.csv', None, None,
'cE', 300),
]
# h2b.browseTheCloud()
trial = 0
for (bucket, csvPathname, iColCount, oColCount, hex_key,
timeoutSecs) in tryList:
xList = []
eList = []
fList = []
# PARSE*******************************************************
parseResult = h2i.import_parse(bucket=bucket,
path=csvPathname,
schema='put',
hex_key=hex_key,
timeoutSecs=200,
doSummary=False)
csvPathnameFull = h2i.find_folder_and_filename(bucket,
csvPathname,
returnFullPath=True)
print "Parse result['destination_key']:", parseResult[
'destination_key']
inspect = h2o_cmd.runInspect(key=parseResult['destination_key'])
h2o_cmd.infoFromInspect(inspect, csvPathname)
numRows = inspect['numRows']
numCols = inspect['numCols']
if not oColCount:
iColCount = 0
if not oColCount:
oColCount = numCols
colCount = iColCount + oColCount
for i in range(0, numCols):
print "Column", i, "summary"
h2o_cmd.runSummary(key=hex_key, max_qbins=1, cols=i)
# print h2o.dump_json(inspect)
levels = h2o.nodes[0].levels(source=hex_key)
# print "levels result:", h2o.dump_json(levels)
(missingValuesDict, constantValuesDict, enumSizeDict, colTypeDict, colNameDict) = \
h2o_cmd.columnInfoFromInspect(parseResult['destination_key'], exceptionOnMissingValues=False)
# error if any col has constant values
if len(constantValuesDict) != 0:
# raise Exception("Probably got a col NA'ed and constant values as a result %s" % constantValuesDict)
print "Probably got a col NA'ed and constant values as a result %s" % constantValuesDict
# start after the last input col
levels = h2o.nodes[0].levels(source=hex_key)
l = levels['levels']
for column in range(iColCount, iColCount + oColCount):
if l[column]:
print "Skipping", column, "because it's enum (says levels)"
continue
# QUANTILE*******************************************************
quantile = 0.5 if DO_MEDIAN else .999
# first output col. always fed by an exec cut, so 0?
start = time.time()
# file has headers. use col index
q = h2o.nodes[0].quantiles(source_key=hex_key,
column=column,
quantile=quantile,
max_qbins=MAX_QBINS,
multiple_pass=1)
qresult = q['result']
h2p.red_print("result:", q['result'], "quantile", quantile,
"interpolated:", q['interpolated'], "iterations",
q['iterations'])
elapsed = time.time() - start
print "quantile end on ", hex_key, 'took', elapsed, 'seconds.'
quantileTime = elapsed
# don't do for enums
# also get the median with a sort (h2o_summ.percentileOnSortedlist()
if 1 == 1:
h2o_summ.quantile_comparisons(
csvPathnameFull,
skipHeader=True,
col=column, # what col to extract from the csv
datatype='float',
quantile=0.5 if DO_MEDIAN else 0.999,
# h2oSummary2=pctile[5 if DO_MEDIAN else 10],
# h2oQuantilesApprox=qresult_single,
h2oQuantilesExact=qresult,
use_genfromtxt=True,
)
trial += 1
execTime = 0
xList.append(column)
eList.append(execTime)
fList.append(quantileTime)
# remove all keys*******************************************************
# what about hex_key?
if 1 == 0:
start = time.time()
h2o.nodes[0].remove_all_keys()
elapsed = time.time() - start
print "remove all keys end on took", elapsed, 'seconds.'
#****************************************************************
# PLOTS. look for eplot.jpg and fplot.jpg in local dir?
if DO_PLOT:
xLabel = 'column (0 is first)'
eLabel = 'exec cut time'
fLabel = 'quantile time'
eListTitle = ""
fListTitle = ""
h2o_gbm.plotLists(xList,
xLabel,
eListTitle,
eList,
eLabel,
fListTitle,
fList,
fLabel,
server=True)
def build_cloud_with_json(h2o_nodes_json="h2o-nodes.json"):
# local sandbox may not exist. Don't clean if it does, just append
if not os.path.exists(LOG_DIR):
os.mkdir(LOG_DIR)
log("#*********************************************************************")
log("Starting new test: " + h2o_args.python_test_name + " at build_cloud_with_json()")
log("#*********************************************************************")
print "This only makes sense if h2o is running as defined by", h2o_nodes_json
print "For now, assuming it's a cloud on this machine, and here's info on h2o processes running here"
print "No output means no h2o here! Some other info about stuff on the system is printed first though."
import h2o_os_util
if not os.path.exists(h2o_nodes_json):
raise Exception("build_cloud_with_json: Can't find " + h2o_nodes_json + " file")
## h2o_os_util.show_h2o_processes()
with open(h2o_nodes_json, "rb") as f:
cloneJson = json.load(f)
# These are supposed to be in the file.
# Just check the first one. if not there, the file must be wrong
if not "cloud_start" in cloneJson:
raise Exception("Can't find 'cloud_start' in %s, wrong file? h2o-nodes.json?" % h2o_nodes_json)
else:
cs = cloneJson["cloud_start"]
print "Info on the how the cloud we're cloning was started (info from %s)" % h2o_nodes_json
# required/legal values in 'cloud_start'. A robust check is good for easy debug when we add stuff
valList = ["time", "cwd", "python_test_name", "python_cmd_line", "config_json", "username", "ip"]
for v in valList:
if v not in cs:
raise Exception("Can't find %s in %s, wrong file or version change?" % (v, h2o_nodes_json))
print "cloud_start['%s']: %s" % (v, cs[v])
# this is the internal node state for python..nodes rebuild
nodeStateList = cloneJson["h2o_nodes"]
nodeList = []
if not nodeStateList:
raise Exception("nodeStateList is empty. %s file must be empty/corrupt" % h2o_nodes_json)
try:
for nodeState in nodeStateList:
print "Cloning state for node", nodeState["node_id"], "from", h2o_nodes_json
newNode = ExternalH2O(nodeState)
nodeList.append(newNode)
# If it's an existing cloud, it may already be locked. so never check.
# we don't have the cloud name in the -ccj since it may change (and the file be static?)
# so don't check expectedCloudName
verify_cloud_size(nodeList, expectedCloudName=None, expectedLocked=None)
# best to check for any errors right away?
# (we won't report errors from prior tests due to marker stuff?
## check_sandbox_for_errors(python_test_name=h2o_args.python_test_name)
# put the test start message in the h2o log, to create a marker
nodeList[0].h2o_log_msg()
except:
# nodeList might be empty in some exception cases?
# no shutdown issued first, though
## if cleanup and nodeList:
## for n in nodeList: n.terminate()
check_sandbox_for_errors(python_test_name=h2o_args.python_test_name)
raise
# like cp -p. Save the config file, to sandbox
print "Saving the ", h2o_nodes_json, "we used to", LOG_DIR
shutil.copy(h2o_nodes_json, LOG_DIR + "/" + os.path.basename(h2o_nodes_json))
print ""
h2p.red_print("Ingested from json:", nodeList[0].java_heap_GB, "GB java heap(s) with", len(nodeList), "total nodes")
print ""
# save it to a global copy, in case it's needed for tearDown
h2o_nodes.nodes[:] = nodeList
return nodeList
def quantile_comparisons(csvPathname,
skipHeader=False,
col=0,
datatype='float',
h2oSummary2=None,
h2oSummary2MaxErr=None,
h2oQuantilesApprox=None,
h2oQuantilesExact=None,
h2oExecQuantiles=None,
interpolate='linear',
quantile=0.50,
use_genfromtxt=False):
SCIPY_INSTALLED = True
try:
import scipy as sp
import numpy as np
print "Both numpy and scipy are installed. Will do extra checks"
except ImportError:
print "numpy or scipy is not installed. Will only do sort-based checking"
SCIPY_INSTALLED = False
if not SCIPY_INSTALLED:
return
if use_genfromtxt:
print "Using numpy.genfromtxt. Better handling of null bytes"
target = np.genfromtxt(open(csvPathname, 'r'),
delimiter=',',
skip_header=1 if skipHeader else 0,
dtype=None) # guess!
# print "shape:", target.shape()
else:
print "Using python csv reader"
target = h2o_util.file_read_csv_col(csvPathname,
col=col,
datatype=datatype,
skipHeader=skipHeader,
preview=5)
if datatype == 'float':
# to make irene's R runif files first col work (quoted row numbers, integers
#shouldn't hurt anyone else?
# strip " from left (ignore leading whitespace
# strip " from right (ignore leading whitespace
targetFP = map(float, target)
# targetFP= np.array(tFP, np.float)
if datatype == 'int':
targetFP = map(int, target)
# http://docs.scipy.org/doc/numpy-dev/reference/generated/numpy.percentile.html
# numpy.percentile has simple linear interpolate and midpoint
# need numpy 1.9 for interpolation. numpy 1.8 doesn't have
# p = np.percentile(targetFP, 50 if DO_MEDIAN else 99.9, interpolation='midpoint')
# 1.8
p = np.percentile(targetFP, quantile * 100)
h2p.red_print("numpy.percentile", p)
# per = [100 * t for t in thresholds]
from scipy import stats
s1 = stats.scoreatpercentile(targetFP, quantile * 100)
h2p.red_print("scipy stats.scoreatpercentile", s1)
# scipy apparently doesn't have the use of means (type 2)
# http://en.wikipedia.org/wiki/Quantile
# it has median (R-8) with 1/3, 1/3
if 1 == 0:
# type 6
alphap = 0
betap = 0
# type 5 okay but not perfect
alphap = 0.5
betap = 0.5
# type 8
alphap = 1 / 3.0
betap = 1 / 3.0
if interpolate == 'mean':
# an approx? (was good when comparing to h2o type 2)
alphap = 0.4
betap = 0.4
if interpolate == 'linear':
# this is type 7
alphap = 1
betap = 1
s2List = stats.mstats.mquantiles(targetFP,
prob=quantile,
alphap=alphap,
betap=betap)
s2 = s2List[0]
# http://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.mstats.mquantiles.html
# type 7
# alphap=0.4, betap=0.4,
# type 2 not available? (mean)
# alphap=1/3.0, betap=1/3.0 is approx median?
h2p.red_print("scipy stats.mstats.mquantiles:", s2)
# also get the median with a painful sort (h2o_summ.percentileOnSortedlist()
# inplace sort
targetFP.sort()
# this matches scipy type 7 (linear)
# b = h2o_summ.percentileOnSortedList(targetFP, 0.50 if DO_MEDIAN else 0.999, interpolate='linear')
# this matches h2o type 2 (mean)
# b = h2o_summ.percentileOnSortedList(targetFP, 0.50 if DO_MEDIAN else 0.999, interpolate='mean')
b = percentileOnSortedList(targetFP, quantile, interpolate='linear')
label = str(quantile * 100) + '%'
h2p.blue_print(label, "from sort:", b)
h2p.blue_print(label, "from numpy:", p)
h2p.blue_print(label, "from scipy 1:", s1)
h2p.blue_print(label, "from scipy 2:", s2)
h2p.blue_print(label, "from h2o summary:", h2oSummary2)
h2p.blue_print(label, "from h2o multipass:"******"from h2o singlepass:"******"from h2o exec:", h2oExecQuantiles)
# they should be identical. keep a tight absolute tolerance
# Note the comparisons have different tolerances, some are relative, some are absolute
if h2oQuantilesExact:
if math.isnan(float(h2oQuantilesExact)):
raise Exception("h2oQuantilesExact is unexpectedly NaN %s" %
h2oQuantilesExact)
h2o_util.assertApproxEqual(
h2oQuantilesExact,
b,
tol=0.0000002,
msg='h2o quantile multipass is not approx. same as sort algo')
if h2oQuantilesApprox:
# this can be NaN if we didn't calculate it. turn the NaN string into a float NaN
if math.isnan(float(h2oQuantilesApprox)):
raise Exception("h2oQuantilesApprox is unexpectedly NaN %s" %
h2oQuantilesApprox)
if h2oSummary2MaxErr:
h2o_util.assertApproxEqual(
h2oQuantilesApprox,
b,
tol=h2oSummary2MaxErr,
msg='h2o quantile singlepass is not approx. same as sort algo')
else:
h2o_util.assertApproxEqual(
h2oQuantilesApprox,
b,
rel=0.1,
msg='h2o quantile singlepass is not approx. same as sort algo')
if h2oSummary2:
if math.isnan(float(h2oSummary2)):
raise Exception("h2oSummary2 is unexpectedly NaN %s" % h2oSummary2)
if h2oSummary2MaxErr:
# maxErr absolute was calculated in the test from 0.5*(max-min/(max_qbins-2))
h2o_util.assertApproxEqual(
h2oSummary2,
b,
tol=h2oSummary2MaxErr,
msg=
'h2o summary2 is not approx. same as sort algo (calculated expected max error)'
)
else:
# bounds are way off, since it depends on the min/max of the col, not the expected value
h2o_util.assertApproxEqual(
h2oSummary2,
b,
rel=1.0,
msg=
'h2o summary2 is not approx. same as sort algo (sloppy compare)'
)
if h2oQuantilesApprox and h2oSummary2:
# they should both get the same answer. Currently they have different code, but same algo
# FIX! ...changing to a relative tolerance, since we're getting a miscompare in some cases.
# not sure why..maybe some subtle algo diff.
h2o_util.assertApproxEqual(h2oSummary2, h2oQuantilesApprox, rel=0.04,
msg='h2o summary2 is not approx. same as h2o singlepass.'+\
' Check that max_qbins is 1000 (summary2 is fixed) and type 7 interpolation')
if h2oExecQuantiles:
if math.isnan(float(h2oExecQuantiles)):
raise Exception("h2oExecQuantiles is unexpectedly NaN %s" %
h2oExecQuantiles)
# bounds are way off
h2o_util.assertApproxEqual(
h2oExecQuantiles,
b,
rel=1.0,
msg='h2o summary2 is not approx. same as sort algo')
if SCIPY_INSTALLED:
if h2oQuantilesExact:
h2o_util.assertApproxEqual(
h2oQuantilesExact,
p,
tol=0.0000002,
msg='h2o quantile multipass is not same as numpy.percentile')
h2o_util.assertApproxEqual(
h2oQuantilesExact,
s1,
tol=0.0000002,
msg=
'h2o quantile multipass is not same as scipy stats.scoreatpercentile'
)
# give us some slack compared to the scipy use of median (instead of desired mean)
# since we don't have bounds here like above, just stop this test for now
if h2oQuantilesApprox and 1 == 0:
if interpolate == 'mean':
h2o_util.assertApproxEqual(
h2oQuantilesApprox,
s2,
rel=0.5,
msg=
'h2o quantile singlepass is not approx. same as scipy stats.mstats.mquantiles'
)
else:
h2o_util.assertApproxEqual(
h2oQuantilesApprox,
s2,
rel=0.5,
msg=
'h2o quantile singlepass is not same as scipy stats.mstats.mquantiles'
)
# see if scipy changes. nope. it doesn't
if 1 == 0:
a = stats.mstats.mquantiles(targetFP,
prob=quantile,
alphap=alphap,
betap=betap)
h2p.red_print("after sort")
h2p.red_print("scipy stats.mstats.mquantiles:", s3)
thresholds = [0.001, 0.01, 0.1, 0.25, 0.33, 0.5, 0.66, 0.75, 0.9, 0.99, 0.999]
#*****************************************************************
# h2o
#*****************************************************************
d = target
# target = dataset
dmin = min(d)
dmax = max(d)
thresholdList = [OTHER_T]
quantiles = findQuantileList(d, dmin, dmax, thresholdList)
h2p.red_print('\nthis b result:', quantiles)
#*****************************************************************
# for comparison
#*****************************************************************
# perPrint = ["%.2f" % v for v in a]
# scipy apparently doesn't have the use of means (type 2)
# http://en.wikipedia.org/wiki/Quantile
# it has median (R-8) with 1/3, 1/3
# type 6
alphap=0
betap=0
# type 5 okay but not perfect
alphap=0.5
betap=0.5
def findQuantile(d, dmin, dmax, threshold):
# return the value at the threshold, or the mean of the two rows that bound it.
# fixed bin count per pass. Stops at maxIterations if not resolved to one true answer
maxIterations = 30
# totalRows should be cleansed of NAs. assume d doesn't have NAs (cleaned elsewhere)
totalRows = len(d)
# Used to have
desiredBinCnt = BIN_COUNT
maxBinCnt = desiredBinCnt + 1 # might go one over due to FP issues
# initialize
newValStart = dmin
newValEnd = dmax
newValRange = newValEnd - newValStart
desiredBinCnt = BIN_COUNT # Could do per-pass adjustment, but fixed works fine.
newValBinSize = newValRange / (desiredBinCnt + 0.0)
newLowCount = 0 # count of rows below the bins
# yes there is no newHighCount. Created during the pass, though.
# state shared by each pass
assert maxBinCnt > 0
hcnt2 = [None for b in range(maxBinCnt)]
hcnt2_min = [None for b in range(maxBinCnt)]
hcnt2_max = [None for b in range(maxBinCnt)]
hcnt2_low = 0
hcnt2_high = 0
assert newValBinSize != 0 # can be negative
assert newValEnd > newValStart
assert newValRange > 0
# break out on stopping condition
# reuse the histogram array hcnt2[]
iteration = 0
done = False
# append to a list of best guesses per pass
best_result = []
def htot2():
return sum(hcnt2) + hcnt2_low + hcnt2_high
while iteration <= maxIterations and not done:
h2p.green_print("newValStart", newValStart)
h2p.green_print("newValEnd", newValEnd)
h2p.green_print("newValRange", newValRange)
h2p.green_print("newValBinSize", newValBinSize)
h2p.green_print("newLowCount", newLowCount)
h2p.green_print("threshold", threshold)
valStart = newValStart
valEnd = newValEnd
valRange = newValRange
valBinSize = newValBinSize
lowCount = newLowCount
desiredBinCnt = BIN_COUNT
maxBinCnt = desiredBinCnt + 1 # might go one over due to FP issues
# playing with creating relative NUDGE values to make sure bin range
# is always inclusive of target.
# ratio it down from valBinSize.
# It doesn't need to be as big as valBinSize.
# implicitly, it shouldn't need to be as large as valBinSize
# can't seem to make it work yet. leave NUDGE=0
NUDGE = 0
# init to zero for each pass
for b in range(maxBinCnt):
hcnt2[b] = 0.0
# Init counts outside of the bins
hcnt2_low = 0
hcnt2_high = 0
# minimum value for higher than the bin. Needed for interpolation
hcnt2_high_min = None
for val in d:
# Need to count the stuff outside the bin-gathering,
# since threshold compare is based on total row compare
# on first pass, shouldn't see anything exceed the start/end bounds
# since those are min/max for the column? (shouldn't be any fp precision issue? or ??)
# oh wait, this valOffset math creates possible precision issue?
# maybe we should address it with the NUDGE value below? but what about first pass?
valOffset = val - valStart
# where are we zeroing in? (start)
binIdx2 = int(math.floor(valOffset / (valBinSize + 0.0))) # make sure it's always an fp divide?
# do some close looking for possible fp arith issues
cA = valOffset < 0
cB = binIdx2 < 0
t = {True: 1, False: 0}
# we get the 10 case
if ((cA and not cB) or (not cA and cB)):
h2p.red_print("AB Interesting lower bin edge case %s%s" % (t[cA], t[cB]), "cA", cA, "cB", cB, "valOffSet", valOffSet, \
"binIdx2", binIdx2)
cC = val > valEnd
cD = binIdx2 >= (maxBinCnt-1) # tighten the compare for printing
if ((cC and not cD) or (not cC and cD)):
h2p.red_print("CD Interesting upper bin edge case %s%s" % (t[cC], t[cD]), "cC", cC, "cB", cD, "val", val, "valEnd", valEnd, \
"binIdx2", binIdx2, "maxBinCnt", maxBinCnt)
# example hits this case..i.e. the max value
# CD Interesting upper bin edge case 01 cC False cB True val 100.995097486 valEnd 100.995097486 binIdx2 2 maxBinCnt 3
if valOffset < 0 or binIdx2<0:
# if valOffset < 0:
# if binIdx2<0:
hcnt2_low += 1
# prevent the extra bin from being used..i.e. eliminate the fuzziness for sure!
# have to use both compares, since can wrap the index (due to start/end shift)
# elif val > valEnd or binIdx2>=(maxBinCnt-1):
# should this really be a valOffset compare?
elif val > valEnd or binIdx2 >= maxBinCnt:
# elif val > valEnd:
# elif binIdx2>=(maxBinCnt-1):
if (hcnt2_high==0) or (val < hcnt2_high_min):
hcnt2_high_min = val;
print "hcnt2_high_min update:", hcnt2_high_min, valOffset, val, valStart, hcnt2_high, val, valEnd
hcnt2_high += 1
else:
# print "(multi) val: ",val," valOffset: ",valOffset," valBinSize: ",valBinSize
assert binIdx2 >=0 and binIdx2<=(maxBinCnt-1), "val %s %s %s %s binIdx2: %s maxBinCnt: %s valBinSize: %s" % \
(val, valStart, valEnd, valOffset, binIdx2, maxBinCnt, valBinSize)
if hcnt2[binIdx2]==0 or (val < hcnt2_min[binIdx2]):
hcnt2_min[binIdx2] = val;
if hcnt2[binIdx2]==0 or (val > hcnt2_max[binIdx2]):
hcnt2_max[binIdx2] = val;
hcnt2[binIdx2] += 1
# check if we went into the magic extra bin
if binIdx2 == (maxBinCnt-1):
print "\nFP! val went into the extra maxBinCnt bin:", \
binIdx2, hcnt2_high_min, valOffset, val, valStart, hcnt2_high, val, valEnd,"\n"
# check the legal states for these two
# we don't have None for checking hcnt2_high_min in java
assert hcnt2_high==0 or (hcnt2_high_min is not None)
assert (hcnt2_high_min is None) or hcnt2_high!=0
# everything should either be in low, the bins, or high
totalBinnedRows = htot2()
print "totalRows check: %s htot2(): %s should be equal. hcnt2_low: %s hcnt2_high: %s" % \
(totalRows, totalBinnedRows, hcnt2_low, hcnt2_high)
assert totalRows==totalBinnedRows, "totalRows: %s htot2() %s not equal. hcnt2_low: %s hcnt2_high: %s" % \
(totalRows, totalBinnedRows, hcnt2_low, hcnt2_high)
# now walk thru and find out what bin to look inside
currentCnt = hcnt2_low
targetCntFull = threshold * (totalRows-1) # zero based indexing
targetCntInt = int(math.floor(threshold * (totalRows-1)))
targetCntFract = targetCntFull - targetCntInt
assert targetCntFract>=0 and targetCntFract<=1
print "targetCntInt:", targetCntInt, "targetCntFract", targetCntFract
k = 0
while ((currentCnt + hcnt2[k]) <= targetCntInt):
# print "looping for k (multi): ",k," ",currentCnt," ",targetCntInt," ",totalRows," ",hcnt2[k]," ",hcnt2_min[k]," ",hcnt2_max[k]
currentCnt += hcnt2[k]
# ugly but have to break out if we'd cycle along with == adding h0's until we go too far
# are we supposed to advance to a none zero bin?
k += 1 # goes over in the equal case?
# if currentCnt >= targetCntInt:
# break
if k==maxBinCnt:
break
assert k<maxBinCnt, "k too large, k: %s maxBinCnt %s %s %s %s" % (k, maxBinCnt, currentCnt, targetCntInt, hcnt2[k-1])
# format string to match java Log.info() in Quantiles.java
print "Found k (multi): ",k," ",currentCnt," ",targetCntInt," ",totalRows," ",hcnt2[k]," ",hcnt2_min[k]," ",hcnt2_max[k]
assert hcnt2[k]!=1 or hcnt2_min[k]==hcnt2_max[k]
# some possibily interpolating guesses first, in guess we have to iterate (best guess)
done = False
guess = (hcnt2_max[k] - hcnt2_min[k]) / 2
if currentCnt==targetCntInt:
if hcnt2[k]>2 and (hcnt2_min[k]==hcnt2_max[k]):
guess = hcnt2_min[k]
print "Guess A", guess, k, hcnt2[k]
if hcnt2[k]==2:
print "\nTwo values in this bin but we could be aligned to the 2nd. so can't stop"
# no mattter what size the fraction it would be on this number
guess = (hcnt2_max[k] + hcnt2_min[k]) / 2.0
# no mattter what size the fraction it would be on this number
if INTERPOLATION_TYPE==2: # type 2 (mean)
guess = (hcnt2_max[k] + hcnt2_min[k]) / 2.0
else: # default to type 7 (linear interpolation)
# Unlike mean, which just depends on two adjacent values, this adjustment
# adds possible errors related to the arithmetic on the total # of rows.
dDiff = hcnt2_max[k] - hcnt2_min[k] # two adjacent..as if sorted!
pctDiff = targetCntFract # This is the fraction of total rows
guess = hcnt2_min[k] + (pctDiff * dDiff)
done = False
print "Guess B", guess
if hcnt2[k]==1 and targetCntFract==0:
assert hcnt2_min[k]==hcnt2_max[k]
guess = hcnt2_min[k]
done = True
print "k", k
print "Guess C", guess
if hcnt2[k]==1 and targetCntFract!=0:
assert hcnt2_min[k]==hcnt2_max[k]
print "\nSingle value in this bin, but fractional means we need to interpolate to next non-zero"
if k<maxBinCnt:
nextK = k + 1 # could put it over maxBinCnt
else:
nextK = k
while nextK<maxBinCnt and hcnt2[nextK]==0:
nextK += 1
# have the "extra bin" for this
if nextK >= maxBinCnt:
assert hcnt2_high!=0
print "Using hcnt2_high_min for interpolate:", hcnt2_high_min
nextVal = hcnt2_high_min
else:
print "Using nextK for interpolate:", nextK
assert hcnt2[nextK]!=0
nextVal = hcnt2_min[nextK]
guess = (hcnt2_max[k] + nextVal) / 2.0
# OH! fixed bin as opposed to sort. Of course there are gaps between k and nextK
if INTERPOLATION_TYPE==2: # type 2 (mean)
guess = (hcnt2_max[k] + nextVal) / 2.0
pctDiff = 0.5
else: # default to type 7 (linear interpolation)
dDiff = nextVal - hcnt2_max[k] # two adjacent, as if sorted!
pctDiff = targetCntFract # This is the fraction of total rows
guess = hcnt2_max[k] + (pctDiff * dDiff)
done = True # has to be one above us when needed. (or we're at end)
print 'k', 'hcnt2_max[k]', 'nextVal'
print "hello3:", k, hcnt2_max[k], nextVal
print "\nInterpolating result using nextK: %s nextVal: %s" % (nextK, nextVal)
print "Guess D", guess
if not done:
print "Not done, setting new range",\
"k: ", k,\
"currentCnt: ", currentCnt,\
"hcnt2_min[k]: ", hcnt2_min[k],\
"hcnt2_max[k]: ", hcnt2_max[k]
# possible bin leakage at start/end edges due to fp arith.
# the bin index arith may resolve OVER the boundary created by the compare for hcnt2_high compare
# rather than using NUDGE, see if there's a non-zero bin below (min) or above (max) you.
# Just need to check the one bin below and above k, if they exist.
if k > 0 and hcnt2[k-1]>0 and (hcnt2_max[k-1]<hcnt2_min[k]):
newValStart = hcnt2_max[k-1]
else:
newValStart = hcnt2_min[k]
# subtle. we do put stuff in the extra end bin (see the print above that happens)
# k might be pointing to one less than that (like k=0 for 1 bin case)
if k < maxBinCnt and hcnt2[k+1]>0 and (hcnt2_min[k+1]>hcnt2_max[k]):
print "hello"
newValEnd = hcnt2_min[k+1]
else:
newValEnd = hcnt2_max[k]
newValRange = newValEnd - newValStart
# maxBinCnt is always binCount + 1, since we might cover over due to rounding/fp issues?
newValBinSize = newValRange / (desiredBinCnt + 0.0)
# the start/end should never change if we're just using one bin
# this is a bin leakage test, if you use one bin. (we should never resolve exactly stop at max iterations
# assumes NUDGE is 0
if NUDGE == 0.0:
assert desiredBinCnt>1 or (valStart==newValStart and valEnd==newValEnd),\
"if 1 bin, should be no per-pass edge leakage %s %s %s %s %s %s" % (k, hcnt2_high, valStart, newValStart, valEnd, newValEnd)
newLowCount = currentCnt
if newValBinSize==0:
# assert done or newValBinSize!=0 and live with current guess
print "Assuming done because newValBinSize is 0."
print "newValRange: %s, hcnt2[k]: %s hcnt2_min[k]: %s hcnt2_max[k]: %s" %\
(newValRange, hcnt2[k], hcnt2_min[k], hcnt2_max[k])
guess = newValStart
print "Guess E", guess
done = True
# if we have to interpolate
# if it falls into this bin, interpolate to this bin means one answer?
# cover the case above with multiple entris in a bin, all the same value
# will be zero on the last pass?
# assert newValBinSize != 0 or done
# need the count up to but not including newValStart
best_result.append(guess)
iteration += 1
h2p.blue_print("Ending Pass", iteration)
h2p.blue_print("best_result:", best_result, "done:", done, "hcnt2[k]", hcnt2[k])
print "currentCnt", currentCnt, "targetCntInt", targetCntInt, "hcnt2_low", hcnt2_low, "hcnt2_high", hcnt2_high
print "was", valStart, valEnd, valRange, valBinSize
print "next", newValStart, newValEnd, newValRange, newValBinSize
return best_result[-1]
def do_statsmodels_glm(self, bucket, csvPathname, L, family='gaussian'):
h2p.red_print("Now doing statsmodels")
h2p.red_print("http://statsmodels.sourceforge.net/devel/glm.html#module-reference")
h2p.red_print("http://statsmodels.sourceforge.net/devel/generated/statsmodels.genmod.generalized_linear_model.GLM.html")
import numpy as np
import scipy as sp
from numpy import loadtxt
import statsmodels as sm
csvPathnameFull = h2i.find_folder_and_filename(bucket, csvPathname, returnFullPath=True)
if 1==1:
dataset = np.loadtxt(
open(csvPathnameFull,'r'),
skiprows=1, # skip the header
delimiter=',',
dtype='float');
# skipping cols from the begining... (ID is col 1)
# In newer versions of Numpy, np.genfromtxt can take an iterable argument,
# so you can wrap the file you're reading in a generator that generates lines,
# skipping the first N columns. If your numbers are comma-separated, that's something like
if 1==0:
f = open(csvPathnameFull,'r'),
np.genfromtxt(
(",".join(ln.split()[1:]) for ln in f),
skiprows=1, # skip the header
delimiter=',',
dtype='float');
print "\ncsv read for training, done"
# data is last column
# drop the output
n_features = len(dataset[0]) - 1;
print "n_features:", n_features
# don't want ID (col 0) or CAPSULE (col 1)
# get CAPSULE
target = [x[1] for x in dataset]
# slice off the first 2
train = np.array ( [x[2:] for x in dataset] )
n_samples, n_features = train.shape
print "n_samples:", n_samples, "n_features:", n_features
print "histogram of target"
print sp.histogram(target,3)
print "len(train):", len(train)
print "len(target):", len(target)
print "dataset shape:", dataset.shape
if family!='gaussian':
raise Exception("Only have gaussian logistic for scipy")
# train the classifier
gauss_log = sm_api.GLM(target, train, family=sm_api.families.Gaussian(sm_api.families.links.log))
start = time.time()
gauss_log_results = gauss_log.fit()
print "sm_api.GLM took", time.time() - start, "seconds"
print gauss_log_results.summary()
def test_exec_enums_rand_cut2(self):
SYNDATASETS_DIR = h2o.make_syn_dir()
n = ROWS
tryList = [
# (n, 10, 9, 'cE', 300),
(n, 1, 1, 'cE', 300),
]
# create key names to use for exec
eKeys = ['e%s' % i for i in range(10)]
# h2b.browseTheCloud()
trial = 0
for (rowCount, iColCount, oColCount, hex_key, timeoutSecs) in tryList:
colCount = iColCount + oColCount
hex_key = 'p'
colEnumList = create_col_enum_list(iColCount)
# create 100 possible cut expressions here, so we don't waste time below
rowExprList = []
print "Creating", CUT_EXPR_CNT, 'cut expressions'
for j in range(CUT_EXPR_CNT):
# init cutValue. None means no compare
cutValue = [None for i in range(iColCount)]
# build up a random cut expression
MAX_COLS_IN_EXPR = iColCount
cols = random.sample(range(MAX_COLS_IN_EXPR), random.randint(1,MAX_COLS_IN_EXPR))
for c in cols:
# possible choices within the column
cel = colEnumList[c]
# for now the cutValues are numbers for the enum mappings
if 1==1:
# FIX! hack. don't use encoding 0, maps to NA here? h2o doesn't like
celChoice = str(random.choice(range(len(cel))))
else:
celChoice = random.choice(cel)
cutValue[c] = celChoice
cutExprList = []
for i,c in enumerate(cutValue):
if c is None:
continue
else:
# new ...ability to reference cols
# src[ src$age<17 && src$zip=95120 && ... , ]
# randomly pick == or !=
if random.randint(0,1)==0:
cutExprList.append('p$C'+str(i+1)+'!='+c)
else:
cutExprList.append('p$C'+str(i+1)+'=='+c)
cutExpr = ' & '.join(cutExprList)
# print "cutExpr:", cutExpr
# just extract one output col (the first one)
rowExpr = '%s[%s,%s];' % (hex_key, cutExpr, iColCount+1)
# print "rowExpr:", rowExpr
print rowExpr
rowExprList.append(rowExpr)
# CREATE DATASET*******************************************
SEEDPERFILE = random.randint(0, sys.maxint)
csvFilename = 'syn_enums_' + str(rowCount) + 'x' + str(colCount) + '.csv'
csvPathname = SYNDATASETS_DIR + '/' + csvFilename
print "Creating random", csvPathname
write_syn_dataset(csvPathname, rowCount, iColCount, oColCount, SEEDPERFILE, colEnumList=colEnumList)
# PARSE*******************************************************
src_key = csvFilename
parseResult = h2i.import_only(path=csvPathname, schema='put', src_key='A'+src_key, timeoutSecs=200)
parseResult = h2i.import_only(path=csvPathname, schema='put', src_key='B'+src_key, timeoutSecs=200)
parseResult = h2i.import_only(path=csvPathname, schema='put', src_key='C'+src_key, timeoutSecs=200)
parseResult = h2i.import_only(path=csvPathname, schema='put', src_key='D'+src_key, timeoutSecs=200)
parseResult = h2i.import_only(path=csvPathname, schema='put', src_key='E'+src_key, timeoutSecs=200)
parseResult = h2i.import_only(path=csvPathname, schema='put', src_key='F'+src_key, timeoutSecs=200)
parseResult = h2i.import_only(path=csvPathname, schema='put', src_key='G'+src_key, timeoutSecs=200)
parseResult = h2i.import_only(path=csvPathname, schema='put', src_key='H'+src_key, timeoutSecs=200)
parseResult = h2i.import_only(path=csvPathname, schema='put', src_key='I'+src_key, timeoutSecs=200)
parseResult = h2i.import_only(path=csvPathname, schema='put', src_key='J'+src_key, timeoutSecs=200)
parseResult = h2i.parse_only(pattern='*'+src_key, hex_key=hex_key, timeoutSecs=800)
print "Parse result['destination_key']:", parseResult['destination_key']
inspect = h2o_cmd.runInspect(key=parseResult['destination_key'])
h2o_cmd.infoFromInspect(inspect, csvPathname)
pNumRows = inspect['numRows']
pNumCols = inspect['numCols']
# print h2o.dump_json(inspect)
levels = h2o.nodes[0].levels(source=hex_key)
print "levels result:", h2o.dump_json(levels)
(missingValuesDict, constantValuesDict, enumSizeDict, colTypeDict, colNameDict) = \
h2o_cmd.columnInfoFromInspect(parseResult['destination_key'], exceptionOnMissingValues=False)
# error if any col has constant values
if len(constantValuesDict) != 0:
raise Exception("Probably got a col NA'ed and constant values as a result %s" % constantValuesDict)
# INIT all possible key names used***************************
# remember. 1 indexing!
# is this needed?
if 1==1:
a = 'a=c(1,2,3);' + ';'.join(['a[,%s]=a[,%s-1]'% (i,i) for i in range(2,colCount)])
print a
for eKey in eKeys:
# build up the columns
e = h2o.nodes[0].exec_query(str='%s;%s=a' % (a, eKey), print_params=False)
## print h2o.dump_json(e)
xList = []
eList = []
fList = []
for repeat in range(CUT_LOOP_CNT):
# EXEC*******************************************************
# don't use exec_expr to avoid issues with Inspect following etc.
randICol = random.randint(0,iColCount-1)
randOCol = random.randint(iColCount, iColCount+oColCount-1)
# should be two different keys in the sample
e = random.sample(eKeys,2)
fKey = e[0]
eKey = e[1]
start = time.time()
h2o.nodes[0].exec_query(str="%s=%s" % (fKey, random.choice(rowExprList)))
elapsed = time.time() - start
execTime = elapsed
print "exec 2 took", elapsed, "seconds."
inspect = h2o_cmd.runInspect(key=fKey)
h2o_cmd.infoFromInspect(inspect, fKey)
numRows = inspect['numRows']
numCols = inspect['numCols']
if numRows==0 or numCols!=colCount:
h2p.red_print("Warning: Cut resulted in", numRows, "rows and", numCols, "cols. Quantile will abort")
# QUANTILE*******************************************************
quantile = 0.5 if DO_MEDIAN else .999
# first output col. always fed by an exec cut, so 0?
column = iColCount
column = 0
start = time.time()
q = h2o.nodes[0].quantiles(source_key=fKey, column=column,
quantile=quantile, max_qbins=MAX_QBINS, multiple_pass=MULTI_PASS)
h2p.red_print("quantile", quantile, q['result'])
elapsed = time.time() - start
print "quantile end on ", fKey, 'took', elapsed, 'seconds.'
quantileTime = elapsed
# remove all keys*******************************************************
# what about hex_key?
if 1==0:
start = time.time()
h2o.nodes[0].remove_all_keys()
elapsed = time.time() - start
print "remove all keys end on ", csvFilename, 'took', elapsed, 'seconds.'
trial += 1
xList.append(trial)
eList.append(execTime)
fList.append(quantileTime)
#****************************************************************
# QUANTILE APPROX. BASELINE FOR SINGLE COL WALK FULL DATASET
print "QUANTILE APPROX. BASELINE FOR SINGLE COL WALK FULL DATASET. Although it's a real col, not an enum col"
quantile = 0.5 if DO_MEDIAN else .999
# first output col. always fed by an exec cut, so 0?
column = iColCount
start = time.time()
q = h2o.nodes[0].quantiles(source_key=hex_key, column='C'+str(iColCount+1),
quantile=quantile, max_qbins=MAX_QBINS, multiple_pass=0)
elapsed = time.time() - start
h2p.red_print(hex_key, pNumRows, "rows Baseline: quantile single col (C" + str(iColCount+1) + ")", "one iteration", elapsed, "secs. threshold:", quantile, q['result'])
print "quantile single col 1 iteration end on", hex_key, "took", elapsed, 'seconds.'
quantileTime = elapsed
#****************************************************************
# PLOTS. look for eplot.jpg and fplot.jpg in local dir?
if DO_PLOT:
xLabel = 'trial'
eLabel = 'exec cut time'
fLabel = 'quantile time'
eListTitle = ""
fListTitle = ""
h2o_gbm.plotLists(xList, xLabel, eListTitle, eList, eLabel, fListTitle, fList, fLabel, server=True)
def test_exec2_enums_rand_cut(self):
SYNDATASETS_DIR = h2o.make_syn_dir()
n = ROWS
tryList = [
(n, 10, 9, 'cE', 300),
]
# create key names to use for exec
eKeys = ['e%s' % i for i in range(10)]
# h2b.browseTheCloud()
trial = 0
for (rowCount, iColCount, oColCount, hex_key, timeoutSecs) in tryList:
colCount = iColCount + oColCount
hex_key = 'p'
colEnumList = create_col_enum_list(iColCount)
# create 100 possible cut expressions here, so we don't waste time below
rowExprList = []
print "Creating", CUT_EXPR_CNT, 'cut expressions'
for j in range(CUT_EXPR_CNT):
# init cutValue. None means no compare
cutValue = [None for i in range(iColCount)]
# build up a random cut expression
cols = random.sample(range(iColCount), random.randint(1,iColCount))
for c in cols:
# possible choices within the column
cel = colEnumList[c]
# for now the cutValues are numbers for the enum mappings
# FIX! hack. don't use encoding 0, maps to NA here? h2o doesn't like
# celChoice = str(random.choice(range(len(cel))))
celChoice = random.choice(range(len(cel)))
cutValue[c] = celChoice
cutExprList = []
pKey = Key('p')
for i,c in enumerate(cutValue):
if c is None:
continue
else:
# new ...ability to reference cols
# src[ src$age<17 && src$zip=95120 && ... , ]
# cutExprList.append('p$C'+str(i+1)+'=='+c)
# all column indexing in h2o-dev is with number
e = Fcn('==', c, pKey[:,i])
cutExprList.append(e)
cutExpr = None
for ce in cutExprList:
if cutExpr:
cutExpr = Fcn('&', cutExpr, ce)
else:
cutExpr = ce
print "cutExpr:", cutExpr
# should be two different keys in the sample
e = random.sample(eKeys,2)
fKey = e[0]
eKey = e[1]
# rowExpr = '%s[%s,];' % (hex_key, cutExpr)
hKey = Key(hex_key)
rowExpr = hKey[cutExpr, :]
print "rowExpr:", rowExpr
rowExprList.append(rowExpr)
# CREATE DATASET*******************************************
SEEDPERFILE = random.randint(0, sys.maxint)
csvFilename = 'syn_enums_' + str(rowCount) + 'x' + str(colCount) + '.csv'
csvPathname = SYNDATASETS_DIR + '/' + csvFilename
print "Creating random", csvPathname
write_syn_dataset(csvPathname, rowCount, iColCount, oColCount, SEEDPERFILE, colEnumList=colEnumList)
# PARSE*******************************************************
parseResult = h2i.import_parse(path=csvPathname, schema='put', hex_key=hex_key, timeoutSecs=30)
numRows, numCols, parse_key = h2o_cmd.infoFromParse(parseResult)
inspect = h2o_cmd.runInspect(key=parse_key)
missingList, valueList, numRows, numCols = h2o_cmd.infoFromInspect(inspect)
# print h2o.dump_json(inspect)
# (missingValuesDict, constantValuesDict, enumSizeDict, colTypeDict, colNameDict) = \
# h2o_cmd.columnInfoFromInspect(parse_key, exceptionOnMissingValues=False)
# error if any col has constant values
# if len(constantValuesDict) != 0:
# raise Exception("Probably got a col NA'ed and constant values as a result %s" % constantValuesDict)
# INIT all possible key names used***************************
# remember. 1 indexing!
# build up the columns
Assign('b', [1,2,3])
# could also append 1 col at a time, by assigning to the next col number?
Assign('a', Cbind(['b' for i in range(colCount)]))
for eKey in eKeys:
Assign(eKey, 'a')
## print h2o.dump_json(e)
xList = []
eList = []
fList = []
for repeat in range(200):
# EXEC*******************************************************
# don't use exec_expr to avoid issues with Inspect following etc.
randICol = random.randint(0,iColCount-1)
randOCol = random.randint(iColCount, iColCount+oColCount-1)
# should be two different keys in the sample
e = random.sample(eKeys,2)
fKey = e[0]
eKey = e[1]
if 1==1:
start = time.time()
Assign(fKey, random.choice(rowExprList)).do()
elapsed = time.time() - start
execTime = elapsed
print "exec 2 took", elapsed, "seconds."
inspect = h2o_cmd.runInspect(key=fKey)
missingList, valueList, numRows, numCols = h2o_cmd.infoFromInspect(inspect)
if numRows==0 or numCols!=colCount:
h2p.red_print("Warning: Cut resulted in", numRows, "rows and", numCols, "cols. Quantile will abort")
# FIX! put quantile back in?
quantileTime = 0
# remove all keys*******************************************************
# what about hex_key?
if 1==0:
start = time.time()
h2o.nodes[0].remove_all_keys()
elapsed = time.time() - start
print "remove all keys end on ", csvFilename, 'took', elapsed, 'seconds.'
trial += 1
xList.append(trial)
eList.append(execTime)
fList.append(quantileTime)
# just get a plot of the last one (biggest)
if DO_PLOT:
xLabel = 'trial'
eLabel = 'exec cut time'
fLabel = 'quantile time'
eListTitle = ""
fListTitle = ""
h2o_gbm.plotLists(xList, xLabel, eListTitle, eList, eLabel, fListTitle, fList, fLabel)
thresholds = [0.001, 0.01, 0.1, 0.25, 0.33, 0.5, 0.66, 0.75, 0.9, 0.99, 0.999]
#*****************************************************************
# h2o
#*****************************************************************
d = target
# target = dataset
dmin = min(d)
dmax = max(d)
thresholdList = [OTHER_T]
quantiles = findQuantileList(d, dmin, dmax, thresholdList)
h2p.red_print('\nthis b result:', quantiles)
#*****************************************************************
# for comparison
#*****************************************************************
# perPrint = ["%.2f" % v for v in a]
# scipy apparently doesn't have the use of means (type 2)
# http://en.wikipedia.org/wiki/Quantile
# it has median (R-8) with 1/3, 1/3
# type 6
alphap = 0
betap = 0
# type 5 okay but not perfect
alphap = 0.5
betap = 0.5
def find_folder_and_filename(bucket, pathWithRegex, schema='put', returnFullPath=False):
checkPath = True
# strip the common mistake of leading "/" in path, if bucket is specified too
giveUpAndSearchLocally = False
if bucket is not None and re.match("/", pathWithRegex):
h2o.verboseprint("You said bucket:", bucket, "so stripping incorrect leading '/' from", pathWithRegex)
pathWithRegex = pathWithRegex.lstrip('/')
if bucket is None: # good for absolute path name
bucketPath = ""
elif bucket == ".":
bucketPath = os.getcwd()
# only use if the build_cloud was for remote H2O
# Never use the var for remote, if you're doing a put! (which always sources local)
elif h2o.nodes[0].remoteH2O and schema!='put' and \
(os.environ.get('H2O_REMOTE_BUCKETS_ROOT') or h2o.nodes[0].h2o_remote_buckets_root):
if (bucket=='smalldata' or bucket=='datasets') and schema=='local':
msg1 = "\nWARNING: you're using remote nodes, and 'smalldata' or 'datasets' git buckets, with schema!=put"
msg2 = "\nThose aren't git pull'ed by the test. Since they are user-maintained, not globally-maintained-by-0xdata,"
msg3 = "\nthey may be out of date at those remote nodes?"
msg4 = "\nGoing to assume we find a path to them locally, and remote path will be the same"
h2p.red_print(msg1, msg2, msg3, msg4)
giveUpAndSearchLocally = True
else:
if os.environ.get('H2O_REMOTE_BUCKETS_ROOT'):
rootPath = os.environ.get('H2O_REMOTE_BUCKETS_ROOT')
print "Found H2O_REMOTE_BUCKETS_ROOT:", rootPath
else:
rootPath = h2o.nodes[0].h2o_remote_buckets_root
print "Found h2o_nodes[0].h2o_remote_buckets_root:", rootPath
bucketPath = os.path.join(rootPath, bucket)
checkPath = False
# does it work to use bucket "." to get current directory
# this covers reote with put too
elif os.environ.get('H2O_BUCKETS_ROOT'):
rootPath = os.environ.get('H2O_BUCKETS_ROOT')
print "Using H2O_BUCKETS_ROOT environment variable:", rootPath
if not (os.path.exists(rootPath)):
raise Exception("H2O_BUCKETS_ROOT in env but %s doesn't exist." % rootPath)
bucketPath = os.path.join(rootPath, bucket)
if not (os.path.exists(bucketPath)):
raise Exception("H2O_BUCKETS_ROOT and path used to form %s which doesn't exist." % bucketPath)
else:
giveUpAndSearchLocally = True
#******************************************************************************************
if giveUpAndSearchLocally:
# if we run remotely, we're assuming the import folder path on the remote machine
# matches what we find on our local machine. But maybe the local user doesn't exist remotely
# so using his path won't work.
# Resolve by looking for special state in the config. If user = 0xdiag, just force the bucket location
# This is a lot like knowing about fixed paths with s3 and hdfs
# Otherwise the remote path needs to match the local discovered path.
# want to check the username being used remotely first. should exist here too if going to use
username = getpass.getuser()
h2oUsername = h2o.nodes[0].username
h2o.verboseprint("username:"******"h2oUsername:"******"datasets" is special. Don't want to find it in /home/0xdiag/datasets
# needs to be the git clone 'datasets'. Find it by walking upwards below
# disable it from this looking in home dir. Could change priority order?
# resolved in order, looking for bucket (ln -s will work) in these home dirs.
if bucket=='datasets': # special case
possibleUsers = []
elif h2oUsername != username:
possibleUsers = [username, h2oUsername, "0xdiag"]
else:
possibleUsers = [username, "0xdiag"]
for u in possibleUsers:
rootPath = os.path.expanduser("~" + u)
bucketPath = os.path.join(rootPath, bucket)
h2o.verboseprint("Checking bucketPath:", bucketPath, 'assuming home is', rootPath)
if os.path.exists(bucketPath):
h2o.verboseprint("search A did find", bucket, "at", rootPath)
break
else:
# last chance to find it by snooping around
rootPath = os.getcwd()
h2o.verboseprint("find_bucket looking upwards from", rootPath, "for", bucket)
# don't spin forever
levels = 0
while not (os.path.exists(os.path.join(rootPath, bucket))):
h2o.verboseprint("Didn't find", bucket, "at", rootPath)
rootPath = os.path.split(rootPath)[0]
levels += 1
if (levels==6):
raise Exception("unable to find bucket: %s" % bucket)
h2o.verboseprint("search B did find", bucket, "at", rootPath)
bucketPath = os.path.join(rootPath, bucket)
#******************************************************************************************
# if there's no path, just return the bucketPath
# but what about cases with a header in the folder too? (not putfile)
if pathWithRegex is None:
if returnFullPath:
return bucketPath
else:
return (bucketPath, None)
# if there is a "/" in the path, that means it's not just a pattern
# split it
# otherwise it is a pattern. use it to search for files in python first?
# FIX! do that later
elif "/" in pathWithRegex:
(head, tail) = os.path.split(pathWithRegex)
folderPath = os.path.abspath(os.path.join(bucketPath, head))
# accept all 0xcustomer-datasets without checking..since the current python user
# may not have permission, but h2o will
# try a couple times with os.stat in between, in case it's not automounting
if '/mnt/0xcustomer-datasets' in folderPath:
pass
else:
retry = 0
while checkPath and (not os.path.exists(folderPath)) and retry<5:
# we can't stat an actual file, because we could have a regex at the end of the pathname
print "Retrying", folderPath, "in case there's a autofs mount problem"
os.stat(folderPath)
retry += 1
time.sleep(1)
if checkPath and not os.path.exists(folderPath):
raise Exception("%s doesn't exist. %s under %s may be wrong?" % (folderPath, head, bucketPath))
else:
folderPath = bucketPath
tail = pathWithRegex
h2o.verboseprint("folderPath:", folderPath, "tail:", tail)
if returnFullPath:
return os.path.join(folderPath, tail)
else:
return (folderPath, tail)
class H2O(object):
def __init__(self,
use_this_ip_addr=None, port=54321, capture_output=True,
force_ip=False, network=None,
use_debugger=None, classpath=None,
use_hdfs=False, use_maprfs=False,
hdfs_version=None, hdfs_name_node=None, hdfs_config=None,
aws_credentials=None,
use_flatfile=False, java_heap_GB=None, java_heap_MB=None, java_extra_args=None,
use_home_for_ice=False, node_id=None, username=None,
random_udp_drop=False, force_tcp=False,
redirect_import_folder_to_s3_path=None,
redirect_import_folder_to_s3n_path=None,
disable_h2o_log=False,
enable_benchmark_log=False,
h2o_remote_buckets_root=None,
delete_keys_at_teardown=False,
cloud_name=None,
disable_assertions=None,
sandbox_ignore_errors=False,
):
if use_hdfs:
# see if we can touch a 0xdata machine
try:
# long timeout in ec2...bad
a = requests.get('http://172.16.2.176:80', timeout=1)
hdfs_0xdata_visible = True
except:
hdfs_0xdata_visible = False
# different defaults, depending on where we're running
if hdfs_name_node is None:
if hdfs_0xdata_visible:
hdfs_name_node = "172.16.2.176"
else: # ec2
hdfs_name_node = "10.78.14.235:9000"
if hdfs_version is None:
if hdfs_0xdata_visible:
hdfs_version = "cdh4"
else: # ec2
hdfs_version = "0.20.2"
self.redirect_import_folder_to_s3_path = redirect_import_folder_to_s3_path
self.redirect_import_folder_to_s3n_path = redirect_import_folder_to_s3n_path
self.aws_credentials = aws_credentials
self.port = port
# None is legal for self.h2o_addr.
# means we won't give an ip to the jar when we start.
# Or we can say use use_this_ip_addr=127.0.0.1, or the known address
# if use_this_addr is None, use 127.0.0.1 for urls and json
# Command line arg 'ip_from_cmd_line' dominates:
# ip_from_cmd_line and use_this_ip_addr shouldn't be used for mutli-node
if h2o_args.ip_from_cmd_line:
self.h2o_addr = h2o_args.ip_from_cmd_line
else:
self.h2o_addr = use_this_ip_addr
self.force_ip = force_ip or (self.h2o_addr!=None)
if self.h2o_addr:
self.http_addr = self.h2o_addr
else:
self.http_addr = h2o_args.python_cmd_ip
if h2o_args.network_from_cmd_line:
self.network = h2o_args.network_from_cmd_line
else:
self.network = network
# command line should always dominate for enabling
if h2o_args.debugger: use_debugger = True
self.use_debugger = use_debugger
self.classpath = classpath
self.capture_output = capture_output
self.use_hdfs = use_hdfs
self.use_maprfs = use_maprfs
self.hdfs_name_node = hdfs_name_node
self.hdfs_version = hdfs_version
self.hdfs_config = hdfs_config
self.use_flatfile = use_flatfile
self.java_heap_GB = java_heap_GB
self.java_heap_MB = java_heap_MB
self.java_extra_args = java_extra_args
self.use_home_for_ice = use_home_for_ice
self.node_id = node_id
if username:
self.username = username
else:
self.username = getpass.getuser()
# don't want multiple reports from tearDown and tearDownClass
# have nodes[0] remember (0 always exists)
self.sandbox_error_was_reported = False
self.sandbox_ignore_errors = sandbox_ignore_errors
self.random_udp_drop = random_udp_drop
self.force_tcp = force_tcp
self.disable_h2o_log = disable_h2o_log
# this dumps stats from tests, and perf stats while polling to benchmark.log
self.enable_benchmark_log = enable_benchmark_log
self.h2o_remote_buckets_root = h2o_remote_buckets_root
self.delete_keys_at_teardown = delete_keys_at_teardown
self.disable_assertions = disable_assertions
if cloud_name:
self.cloud_name = cloud_name
else:
self.cloud_name = 'pytest-%s-%s' % (getpass.getuser(), os.getpid())
def __str__(self):
return '%s - http://%s:%d/' % (type(self), self.http_addr, self.port)
def url(self, loc, port=None):
# always use the new api port
if port is None: port = self.port
if loc.startswith('/'):
delim = ''
else:
delim = '/'
u = 'http://%s:%d%s%s' % (self.http_addr, port, delim, loc)
return u
def do_json_request(self, jsonRequest=None, fullUrl=None, timeout=10, params=None, postData=None, returnFast=False,
cmd='get', extraComment=None, ignoreH2oError=False, noExtraErrorCheck=False, **kwargs):
# if url param is used, use it as full url. otherwise create from the jsonRequest
if fullUrl:
url = fullUrl
else:
url = self.url(jsonRequest)
# remove any params that are 'None'
# need to copy dictionary, since can't delete while iterating
if params is not None:
params2 = params.copy()
for k in params2:
if params2[k] is None:
del params[k]
paramsStr = '?' + '&'.join(['%s=%s' % (k, v) for (k, v) in params.items()])
else:
paramsStr = ''
extraComment2 = " " + str(postData)+";" if cmd=='post' else ""
extraComment2 += extraComment if extraComment else ""
if len(extraComment2) > 0:
log('Start ' + url + paramsStr, comment=extraComment2)
else:
log('Start ' + url + paramsStr)
# file get passed thru kwargs here
if h2o_args.no_timeout:
timeout = None # infinite
try:
if 'post' == cmd:
# NOTE == cmd: for now, since we don't have deserialization from JSON in h2o-dev, we use form-encoded POST.
# This is temporary.
#
# This following does application/json (aka, posting JSON in the body):
# r = requests.post(url, timeout=timeout, params=params, data=json.dumps(postData), **kwargs)
#
# This does form-encoded, which doesn't allow POST of nested structures
r = requests.post(url, timeout=timeout, params=params, data=postData, **kwargs)
elif 'delete' == cmd:
r = requests.delete(url, timeout=timeout, params=params, **kwargs)
elif 'get' == cmd:
r = requests.get(url, timeout=timeout, params=params, **kwargs)
else:
raise ValueError("Unknown HTTP command (expected 'get', 'post' or 'delete'): " + cmd)
except Exception, e:
# rethrow the exception after we've checked for stack trace from h2o
# out of memory errors maybe don't show up right away? so we should wait for h2o
# to get it out to h2o stdout. We don't want to rely on cloud teardown to check
# because there's no delay, and we don't want to delay all cloud teardowns by waiting.
exc_info = sys.exc_info()
# use this to ignore the initial connection errors during build cloud when h2o is coming up
if not noExtraErrorCheck:
h2p.red_print(
"ERROR: got exception on %s to h2o. \nGoing to check sandbox, then rethrow.." % (url + paramsStr))
time.sleep(2)
check_sandbox_for_errors(python_test_name=h2o_args.python_test_name);
raise exc_info[1], None, exc_info[2]
if 200 != r.status_code:
print "JSON call returned non-200 status with ", (url + paramsStr)
print "r.status_code: " + str(r.status_code)
print "r.headers: " + repr(r.headers)
print "r.text: " + r.text
# fatal if no response
# FIX! why is this not working on bad response to GLM
# if not r:
# raise Exception("Maybe bad url? no r in do_json_request in %s:" % inspect.stack()[1][3])
# this is used to open a browser on results, or to redo the operation in the browser
# we don't' have that may urls flying around, so let's keep them all
# FIX! this doesn't work now with all the extra post data required?
h2o_nodes.json_url_history.append(r.url)
# if r.json():
# raise Exception("Maybe bad url? no r.json in do_json_request in %s:" % inspect.stack()[1][3])
rjson = None
if returnFast:
return
try:
# h2o-dev sometimes is returning ISO-8859-2, Latin-2?
## print "apparent_coding", r.apparent_encoding
r.encoding = 'utf-8'
rjson = r.json()
except:
h2p.red_print("r.text:", r.text.encode('utf8'))
try:
# try to decode the r.text?
if not isinstance(json.loads(r.text), (list, dict)):
raise Exception("h2o json responses should always be lists or dicts, see previous for text")
except:
raise Exception("Could not decode any json from the request %s." % r.text)
# TODO: we should really only look in the response object. This check
# prevents us from having a field called "error" (e.g., for a scoring result).
for e in ['error', 'Error', 'errors', 'Errors']:
# error can be null (python None). This happens in exec2
if e in rjson and rjson[e]:
print "rjson:", dump_json(rjson)
emsg = 'rjson %s in %s: %s' % (e, inspect.stack()[1][3], rjson[e])
if ignoreH2oError:
# well, we print it..so not totally ignore. test can look at rjson returned
print emsg
else:
print emsg
raise Exception(emsg)
for w in ['warning', 'Warning', 'warnings', 'Warnings']:
# warning can be null (python None).
if w in rjson and rjson[w]:
verboseprint(dump_json(rjson))
print 'rjson %s in %s: %s' % (w, inspect.stack()[1][3], rjson[w])
return rjson
def do_h2o_glm(self, bucket, csvPathname, L, family='binomial'):
h2p.red_print("\nNow doing h2o")
parseResult = h2i.import_parse(bucket='smalldata', path=csvPathname, schema='local', timeoutSecs=180)
# save the resolved pathname for use in the sklearn csv read below
inspect = h2o_cmd.runInspect(None, parseResult['destination_key'])
print inspect
print "\n" + csvPathname, \
" numRows:", "{:,}".format(inspect['numRows']), \
" numCols:", "{:,}".format(inspect['numCols'])
x = 'ID'
y = 'CAPSULE'
family = family
alpha = '0'
lambda_ = L
nfolds = '0'
f = 'prostate'
modelKey = 'GLM_' + f
kwargs = {
'response' : y,
'ignored_cols' : x,
'family' : family,
'lambda' : lambda_,
'alpha' : alpha,
'n_folds' : nfolds, # passes if 0, fails otherwise
'destination_key' : modelKey,
}
timeoutSecs = 60
start = time.time()
glmResult = h2o_cmd.runGLM(parseResult=parseResult, timeoutSecs=timeoutSecs, **kwargs)
# this stuff was left over from when we got the result after polling the jobs list
# okay to do it again
# GLM2: when it redirects to the model view, we no longer have the job_key! (unlike the first response and polling)
(warnings, clist, intercept) = h2o_glm.simpleCheckGLM(self, glmResult, None, **kwargs)
cstring = "".join([("%.5e " % c) for c in clist])
h2p.green_print("h2o alpha ", alpha)
h2p.green_print("h2o lambda ", lambda_)
h2p.green_print("h2o coefficient list:", cstring)
h2p.green_print("h2o intercept", "%.5e " % intercept)
# other stuff in the json response
glm_model = glmResult['glm_model']
_names = glm_model['_names']
coefficients_names = glm_model['coefficients_names']
# the first submodel is the right one, if onely one lambda is provided as a parameter above
submodels = glm_model['submodels'][0]
beta = submodels['beta']
h2p.red_print("beta:", beta)
norm_beta = submodels['norm_beta']
iteration = submodels['iteration']
validation = submodels['validation']
auc = validation['auc']
aic = validation['aic']
null_deviance = validation['null_deviance']
residual_deviance = validation['residual_deviance']
print '_names', _names
print 'coefficients_names', coefficients_names
# did beta get shortened? the simple check confirms names/beta/norm_beta are same length
print 'beta', beta
print 'iteration', iteration
print 'auc', auc
def __do_json_request(self, jsonRequest=None, fullUrl=None, timeout=10, params=None, postData=None, returnFast=False,
cmd='get', extraComment=None, ignoreH2oError=False, noExtraErrorCheck=False, raiseIfNon200=True, **kwargs):
H2O.verboseprint("__do_json_request, timeout: " + str(timeout))
# if url param is used, use it as full url. otherwise crate from the jsonRequest
if fullUrl:
url = fullUrl
else:
url = self.__url(jsonRequest)
# remove any params that are 'None'
# need to copy dictionary, since can't delete while iterating
if params is not None:
params_serialized = params.copy()
for k in params_serialized:
if params_serialized[k] is None:
del params[k]
paramsStr = '?' + '&'.join(['%s=%s' % (k, v) for (k, v) in params.items()])
else:
paramsStr = ''
# The requests package takes array parameters and explodes them: ['f00', 'b4r'] becomes "f00,b4r".
# NOTE: this handles 1D arrays only; if we need ND this needs to be recursive.
# NOTE: we currently don't need to do this for GET, so that's not implemented.
if postData is not None:
munged_postData = {}
for k, v in postData.iteritems():
if type(v) is list:
if len(v) == 0:
munged_postData[k] = '[]'
else:
first = True
array_str = '['
for val in v:
if not first: array_str += ', '
if val is None:
array_str += 'null'
elif isinstance(val, basestring):
array_str += "\"" + str(val) + "\""
else:
array_str += str(val)
first = False
array_str += ']'
munged_postData[k] = array_str
else:
# not list:
munged_postData[k] = v
else:
# None
munged_postData = postData
if extraComment:
log('Start ' + url + paramsStr, comment=extraComment)
else:
log('Start ' + url + paramsStr)
log_rest("")
log_rest("----------------------------------------------------------------------\n")
if extraComment:
log_rest("# Extra comment info about this request: " + extraComment)
if cmd == 'get':
log_rest("GET")
else:
log_rest("POST")
log_rest(url + paramsStr)
# file get passed thru kwargs here
try:
if 'post' == cmd:
# NOTE == cmd: for now, since we don't have deserialization from JSON in h2o-dev, we use form-encoded POST.
# This is temporary.
#
# This following does application/json (aka, posting JSON in the body):
# r = requests.post(url, timeout=timeout, params=params, data=json.dumps(munged_postData), **kwargs)
#
# This does form-encoded, which doesn't allow POST of nested structures
r = requests.post(url, timeout=timeout, params=params, data=munged_postData, **kwargs)
elif 'delete' == cmd:
r = requests.delete(url, timeout=timeout, params=params, **kwargs)
elif 'get' == cmd:
r = requests.get(url, timeout=timeout, params=params, **kwargs)
else:
raise ValueError("Unknown HTTP command (expected 'get', 'post' or 'delete'): " + cmd)
except Exception, e:
# rethrow the exception after we've checked for stack trace from h2o
# out of memory errors maybe don't show up right away? so we should wait for h2o
# to get it out to h2o stdout. We don't want to rely on cloud teardown to check
# because there's no delay, and we don't want to delay all cloud teardowns by waiting.
# (this is new/experimental)
exc_info = sys.exc_info()
# use this to ignore the initial connection errors during build cloud when h2o is coming up
if not noExtraErrorCheck:
h2p.red_print(
"ERROR: got exception on %s to h2o. \nGoing to check sandbox, then rethrow.." % (url + paramsStr))
time.sleep(2)
H2O.check_sandbox_for_errors(python_test_name=H2O.python_test_name);
log_rest("")
log_rest("EXCEPTION CAUGHT DOING REQUEST: " + str(e.message))
raise exc_info[1], None, exc_info[2]
H2O.verboseprint("r: " + repr(r))
def do_json_request(self,
jsonRequest=None,
fullUrl=None,
timeout=10,
params=None,
returnFast=False,
cmd='get',
extraComment=None,
ignoreH2oError=False,
noExtraErrorCheck=False,
**kwargs):
# if url param is used, use it as full url. otherwise crate from the jsonRequest
if fullUrl:
url = fullUrl
else:
url = self.url(jsonRequest)
# remove any params that are 'None'
# need to copy dictionary, since can't delete while iterating
if params is not None:
params2 = params.copy()
for k in params2:
if params2[k] is None:
del params[k]
paramsStr = '?' + '&'.join(
['%s=%s' % (k, v) for (k, v) in params.items()])
else:
paramsStr = ''
if extraComment:
log('Start ' + url + paramsStr, comment=extraComment)
else:
log('Start ' + url + paramsStr)
log_rest("")
log_rest(
"----------------------------------------------------------------------\n"
)
if extraComment:
log_rest("# Extra comment info about this request: " +
extraComment)
if cmd == 'get':
log_rest("GET")
else:
log_rest("POST")
log_rest(url + paramsStr)
# file get passed thru kwargs here
try:
if cmd == 'post':
r = requests.post(url,
timeout=timeout,
params=params,
**kwargs)
else:
r = requests.get(url, timeout=timeout, params=params, **kwargs)
except Exception, e:
# rethrow the exception after we've checked for stack trace from h2o
# out of memory errors maybe don't show up right away? so we should wait for h2o
# to get it out to h2o stdout. We don't want to rely on cloud teardown to check
# because there's no delay, and we don't want to delay all cloud teardowns by waiting.
exc_info = sys.exc_info()
# use this to ignore the initial connection errors during build cloud when h2o is coming up
if not noExtraErrorCheck:
h2p.red_print(
"ERROR: got exception on %s to h2o. \nGoing to check sandbox, then rethrow.."
% (url + paramsStr))
time.sleep(2)
check_sandbox_for_errors(
python_test_name=h2o_args.python_test_name)
log_rest("")
log_rest("EXCEPTION CAUGHT DOING REQUEST: " + str(e.message))
raise exc_info[1], None, exc_info[2]
def build_cloud_with_json(h2o_nodes_json='h2o-nodes.json'):
# local sandbox may not exist. Don't clean if it does, just append
if not os.path.exists(LOG_DIR):
os.mkdir(LOG_DIR)
log("#*********************************************************************")
log("Starting new test: " + h2o_args.python_test_name + " at build_cloud_with_json()")
log("#*********************************************************************")
print "This only makes sense if h2o is running as defined by", h2o_nodes_json
print "For now, assuming it's a cloud on this machine, and here's info on h2o processes running here"
print "No output means no h2o here! Some other info about stuff on the system is printed first though."
import h2o_os_util
if not os.path.exists(h2o_nodes_json):
raise Exception("build_cloud_with_json: Can't find " + h2o_nodes_json + " file")
## h2o_os_util.show_h2o_processes()
with open(h2o_nodes_json, 'rb') as f:
cloneJson = json.load(f)
# These are supposed to be in the file.
# Just check the first one. if not there, the file must be wrong
if not 'cloud_start' in cloneJson:
raise Exception("Can't find 'cloud_start' in %s, wrong file? h2o-nodes.json?" % h2o_nodes_json)
else:
cs = cloneJson['cloud_start']
print "Info on the how the cloud we're cloning was started (info from %s)" % h2o_nodes_json
# required/legal values in 'cloud_start'. A robust check is good for easy debug when we add stuff
valList = ['time', 'cwd', 'python_test_name', 'python_cmd_line', 'config_json', 'username', 'ip']
for v in valList:
if v not in cs:
raise Exception("Can't find %s in %s, wrong file or version change?" % (v, h2o_nodes_json))
print "cloud_start['%s']: %s" % (v, cs[v])
# this is the internal node state for python..nodes rebuild
nodeStateList = cloneJson['h2o_nodes']
nodeList = []
if not nodeStateList:
raise Exception("nodeStateList is empty. %s file must be empty/corrupt" % h2o_nodes_json)
try:
for nodeState in nodeStateList:
print "Cloning state for node", nodeState['node_id'], 'from', h2o_nodes_json
newNode = ExternalH2O(nodeState)
nodeList.append(newNode)
# If it's an existing cloud, it may already be locked. so never check.
# we don't have the cloud name in the -ccj since it may change (and the file be static?)
# so don't check expectedCloudName
verify_cloud_size(nodeList, expectedCloudName=None, expectedLocked=None)
# best to check for any errors right away?
# (we won't report errors from prior tests due to marker stuff?
## check_sandbox_for_errors(python_test_name=h2o_args.python_test_name)
# put the test start message in the h2o log, to create a marker
nodeList[0].h2o_log_msg()
except:
# nodeList might be empty in some exception cases?
# no shutdown issued first, though
## if cleanup and nodeList:
## for n in nodeList: n.terminate()
check_sandbox_for_errors(python_test_name=h2o_args.python_test_name)
raise
# like cp -p. Save the config file, to sandbox
print "Saving the ", h2o_nodes_json, "we used to", LOG_DIR
shutil.copy(h2o_nodes_json, LOG_DIR + "/" + os.path.basename(h2o_nodes_json))
print ""
h2p.red_print("Ingested from json:",
nodeList[0].java_heap_GB, "GB java heap(s) with",
len(nodeList), "total nodes")
print ""
# save it to a global copy, in case it's needed for tearDown
h2o_nodes.nodes[:] = nodeList
return nodeList
def findQuantile(d, dmin, dmax, threshold):
# return the value at the threshold, or the mean of the two rows that bound it.
# fixed bin count per pass. Stops at maxIterations if not resolved to one true answer
maxIterations = 30
# totalRows should be cleansed of NAs. assume d doesn't have NAs (cleaned elsewhere)
totalRows = len(d)
# Used to have
desiredBinCnt = BIN_COUNT
maxBinCnt = desiredBinCnt + 1 # might go one over due to FP issues
# initialize
newValStart = dmin
newValEnd = dmax
newValRange = newValEnd - newValStart
desiredBinCnt = BIN_COUNT # Could do per-pass adjustment, but fixed works fine.
newValBinSize = newValRange / (desiredBinCnt + 0.0)
newLowCount = 0 # count of rows below the bins
# yes there is no newHighCount. Created during the pass, though.
# state shared by each pass
assert maxBinCnt > 0
hcnt2 = [None for b in range(maxBinCnt)]
hcnt2_min = [None for b in range(maxBinCnt)]
hcnt2_max = [None for b in range(maxBinCnt)]
hcnt2_low = 0
hcnt2_high = 0
assert newValBinSize != 0 # can be negative
assert newValEnd > newValStart
assert newValRange > 0
# break out on stopping condition
# reuse the histogram array hcnt2[]
iteration = 0
done = False
# append to a list of best guesses per pass
best_result = []
def htot2():
return sum(hcnt2) + hcnt2_low + hcnt2_high
while iteration <= maxIterations and not done:
h2p.green_print("newValStart", newValStart)
h2p.green_print("newValEnd", newValEnd)
h2p.green_print("newValRange", newValRange)
h2p.green_print("newValBinSize", newValBinSize)
h2p.green_print("newLowCount", newLowCount)
h2p.green_print("threshold", threshold)
valStart = newValStart
valEnd = newValEnd
valRange = newValRange
valBinSize = newValBinSize
lowCount = newLowCount
desiredBinCnt = BIN_COUNT
maxBinCnt = desiredBinCnt + 1 # might go one over due to FP issues
# playing with creating relative NUDGE values to make sure bin range
# is always inclusive of target.
# ratio it down from valBinSize.
# It doesn't need to be as big as valBinSize.
# implicitly, it shouldn't need to be as large as valBinSize
# can't seem to make it work yet. leave NUDGE=0
NUDGE = 0
# init to zero for each pass
for b in range(maxBinCnt):
hcnt2[b] = 0.0
# Init counts outside of the bins
hcnt2_low = 0
hcnt2_high = 0
# minimum value for higher than the bin. Needed for interpolation
hcnt2_high_min = None
for val in d:
# Need to count the stuff outside the bin-gathering,
# since threshold compare is based on total row compare
# on first pass, shouldn't see anything exceed the start/end bounds
# since those are min/max for the column? (shouldn't be any fp precision issue? or ??)
# oh wait, this valOffset math creates possible precision issue?
# maybe we should address it with the NUDGE value below? but what about first pass?
valOffset = val - valStart
# where are we zeroing in? (start)
binIdx2 = int(math.floor(
valOffset /
(valBinSize + 0.0))) # make sure it's always an fp divide?
# do some close looking for possible fp arith issues
cA = valOffset < 0
cB = binIdx2 < 0
t = {True: 1, False: 0}
# we get the 10 case
if ((cA and not cB) or (not cA and cB)):
h2p.red_print("AB Interesting lower bin edge case %s%s" % (t[cA], t[cB]), "cA", cA, "cB", cB, "valOffSet", valOffSet, \
"binIdx2", binIdx2)
cC = val > valEnd
cD = binIdx2 >= (maxBinCnt - 1) # tighten the compare for printing
if ((cC and not cD) or (not cC and cD)):
h2p.red_print("CD Interesting upper bin edge case %s%s" % (t[cC], t[cD]), "cC", cC, "cB", cD, "val", val, "valEnd", valEnd, \
"binIdx2", binIdx2, "maxBinCnt", maxBinCnt)
# example hits this case..i.e. the max value
# CD Interesting upper bin edge case 01 cC False cB True val 100.995097486 valEnd 100.995097486 binIdx2 2 maxBinCnt 3
if valOffset < 0 or binIdx2 < 0:
# if valOffset < 0:
# if binIdx2<0:
hcnt2_low += 1
# prevent the extra bin from being used..i.e. eliminate the fuzziness for sure!
# have to use both compares, since can wrap the index (due to start/end shift)
# elif val > valEnd or binIdx2>=(maxBinCnt-1):
# should this really be a valOffset compare?
elif val > valEnd or binIdx2 >= maxBinCnt:
# elif val > valEnd:
# elif binIdx2>=(maxBinCnt-1):
if (hcnt2_high == 0) or (val < hcnt2_high_min):
hcnt2_high_min = val
print "hcnt2_high_min update:", hcnt2_high_min, valOffset, val, valStart, hcnt2_high, val, valEnd
hcnt2_high += 1
else:
# print "(multi) val: ",val," valOffset: ",valOffset," valBinSize: ",valBinSize
assert binIdx2 >=0 and binIdx2<=(maxBinCnt-1), "val %s %s %s %s binIdx2: %s maxBinCnt: %s valBinSize: %s" % \
(val, valStart, valEnd, valOffset, binIdx2, maxBinCnt, valBinSize)
if hcnt2[binIdx2] == 0 or (val < hcnt2_min[binIdx2]):
hcnt2_min[binIdx2] = val
if hcnt2[binIdx2] == 0 or (val > hcnt2_max[binIdx2]):
hcnt2_max[binIdx2] = val
hcnt2[binIdx2] += 1
# check if we went into the magic extra bin
if binIdx2 == (maxBinCnt - 1):
print "\nFP! val went into the extra maxBinCnt bin:", \
binIdx2, hcnt2_high_min, valOffset, val, valStart, hcnt2_high, val, valEnd,"\n"
# check the legal states for these two
# we don't have None for checking hcnt2_high_min in java
assert hcnt2_high == 0 or (hcnt2_high_min is not None)
assert (hcnt2_high_min is None) or hcnt2_high != 0
# everything should either be in low, the bins, or high
totalBinnedRows = htot2()
print "totalRows check: %s htot2(): %s should be equal. hcnt2_low: %s hcnt2_high: %s" % \
(totalRows, totalBinnedRows, hcnt2_low, hcnt2_high)
assert totalRows==totalBinnedRows, "totalRows: %s htot2() %s not equal. hcnt2_low: %s hcnt2_high: %s" % \
(totalRows, totalBinnedRows, hcnt2_low, hcnt2_high)
# now walk thru and find out what bin to look inside
currentCnt = hcnt2_low
targetCntFull = threshold * (totalRows - 1) # zero based indexing
targetCntInt = int(math.floor(threshold * (totalRows - 1)))
targetCntFract = targetCntFull - targetCntInt
assert targetCntFract >= 0 and targetCntFract <= 1
print "targetCntInt:", targetCntInt, "targetCntFract", targetCntFract
k = 0
while ((currentCnt + hcnt2[k]) <= targetCntInt):
# print "looping for k (multi): ",k," ",currentCnt," ",targetCntInt," ",totalRows," ",hcnt2[k]," ",hcnt2_min[k]," ",hcnt2_max[k]
currentCnt += hcnt2[k]
# ugly but have to break out if we'd cycle along with == adding h0's until we go too far
# are we supposed to advance to a none zero bin?
k += 1 # goes over in the equal case?
# if currentCnt >= targetCntInt:
# break
if k == maxBinCnt:
break
assert k < maxBinCnt, "k too large, k: %s maxBinCnt %s %s %s %s" % (
k, maxBinCnt, currentCnt, targetCntInt, hcnt2[k - 1])
# format string to match java Log.info() in Quantiles.java
print "Found k (multi): ", k, " ", currentCnt, " ", targetCntInt, " ", totalRows, " ", hcnt2[
k], " ", hcnt2_min[k], " ", hcnt2_max[k]
assert hcnt2[k] != 1 or hcnt2_min[k] == hcnt2_max[k]
# some possibily interpolating guesses first, in guess we have to iterate (best guess)
done = False
guess = (hcnt2_max[k] - hcnt2_min[k]) / 2
# we maight not have gottent all the way
if currentCnt == targetCntInt:
if hcnt2[k] > 2 and (hcnt2_min[k] == hcnt2_max[k]):
guess = hcnt2_min[k]
print "Guess A", guess, k, hcnt2[k]
if hcnt2[k] == 2:
print "hello"
print "\nTwo values in this bin but we could be aligned to the 2nd. so can't stop"
# no mattter what size the fraction it would be on this number
guess = (hcnt2_max[k] + hcnt2_min[k]) / 2.0
# no mattter what size the fraction it would be on this number
if INTERPOLATION_TYPE == 2: # type 2 (mean)
guess = (hcnt2_max[k] + hcnt2_min[k]) / 2.0
else: # default to type 7 (linear interpolation)
# Unlike mean, which just depends on two adjacent values, this adjustment
# adds possible errors related to the arithmetic on the total # of rows.
dDiff = hcnt2_max[k] - hcnt2_min[
k] # two adjacent..as if sorted!
pctDiff = targetCntFract # This is the fraction of total rows
guess = hcnt2_min[k] + (pctDiff * dDiff)
done = False
print "Guess B", guess
if hcnt2[k] == 1 and targetCntFract == 0:
assert hcnt2_min[k] == hcnt2_max[k]
guess = hcnt2_min[k]
done = True
print "k", k
print "Guess C", guess
if hcnt2[k] == 1 and targetCntFract != 0:
assert hcnt2_min[k] == hcnt2_max[k]
print "\nSingle value in this bin, but fractional means we need to interpolate to next non-zero"
if k < maxBinCnt:
nextK = k + 1 # could put it over maxBinCnt
else:
nextK = k
while nextK < maxBinCnt and hcnt2[nextK] == 0:
nextK += 1
# have the "extra bin" for this
if nextK >= maxBinCnt:
assert hcnt2_high != 0
print "Using hcnt2_high_min for interpolate:", hcnt2_high_min
nextVal = hcnt2_high_min
else:
print "Using nextK for interpolate:", nextK
assert hcnt2[nextK] != 0
nextVal = hcnt2_min[nextK]
guess = (hcnt2_max[k] + nextVal) / 2.0
# OH! fixed bin as opposed to sort. Of course there are gaps between k and nextK
if INTERPOLATION_TYPE == 2: # type 2 (mean)
guess = (hcnt2_max[k] + nextVal) / 2.0
pctDiff = 0.5
else: # default to type 7 (linear interpolation)
dDiff = nextVal - hcnt2_max[
k] # two adjacent, as if sorted!
pctDiff = targetCntFract # This is the fraction of total rows
guess = hcnt2_max[k] + (pctDiff * dDiff)
done = True # has to be one above us when needed. (or we're at end)
print 'k', 'hcnt2_max[k]', 'nextVal'
print "hello3:", k, hcnt2_max[k], nextVal
print "\nInterpolating result using nextK: %s nextVal: %s" % (
nextK, nextVal)
print "Guess D", guess
if not done:
print "%s %s %s %s Not done, setting new range" % (hcnt2[k], currentCnt, targetCntInt, targetCntFract),\
"k: ", k,\
"currentCnt: ", currentCnt,\
"hcnt2_min[k]: ", hcnt2_min[k],\
"hcnt2_max[k]: ", hcnt2_max[k]
# possible bin leakage at start/end edges due to fp arith.
# the bin index arith may resolve OVER the boundary created by the compare for hcnt2_high compare
# rather than using NUDGE, see if there's a non-zero bin below (min) or above (max) you.
# Just need to check the one bin below and above k, if they exist.
if k > 0 and hcnt2[k - 1] > 0 and (hcnt2_max[k - 1] <
hcnt2_min[k]):
print "1"
newValStart = hcnt2_max[k - 1]
else:
print "2"
newValStart = hcnt2_min[k]
# subtle. we do put stuff in the extra end bin (see the print above that happens)
# k might be pointing to one less than that (like k=0 for 1 bin case)
if k < maxBinCnt and hcnt2[k + 1] > 0 and (hcnt2_min[k + 1] >
hcnt2_max[k]):
print "3"
newValEnd = hcnt2_min[k + 1]
else:
print "4"
newValEnd = hcnt2_max[k]
newValRange = newValEnd - newValStart
# maxBinCnt is always binCount + 1, since we might cover over due to rounding/fp issues?
newValBinSize = newValRange / (desiredBinCnt + 0.0)
# the start/end should never change if we're just using one bin
# this is a bin leakage test, if you use one bin. (we should never resolve exactly stop at max iterations
# assumes NUDGE is 0
if NUDGE == 0.0:
assert desiredBinCnt>1 or (valStart==newValStart and valEnd==newValEnd),\
"if 1 bin, should be no per-pass edge leakage %s %s %s %s %s %s" % (k, hcnt2_high, valStart, newValStart, valEnd, newValEnd)
newLowCount = currentCnt
if newValBinSize == 0:
# assert done or newValBinSize!=0 and live with current guess
print "Assuming done because newValBinSize is 0."
print "newValRange: %s, hcnt2[k]: %s hcnt2_min[k]: %s hcnt2_max[k]: %s" %\
(newValRange, hcnt2[k], hcnt2_min[k], hcnt2_max[k])
guess = newValStart
print "Guess E", guess
# was done = True 3/20/14
done = True
# if we have to interpolate
# if it falls into this bin, interpolate to this bin means one answer?
# cover the case above with multiple entries in a bin, all the same value
# will be zero on the last pass?
# assert newValBinSize != 0 or done
# need the count up to but not including newValStart
best_result.append(guess)
iteration += 1
h2p.blue_print("Ending Pass", iteration)
h2p.blue_print("best_result:", best_result, "done:", done, "hcnt2[k]",
hcnt2[k])
print "currentCnt", currentCnt, "targetCntInt", targetCntInt, "hcnt2_low", hcnt2_low, "hcnt2_high", hcnt2_high
print "was", valStart, valEnd, valRange, valBinSize
print "next", newValStart, newValEnd, newValRange, newValBinSize
return best_result[-1]
def quantile_comparisons(csvPathname, skipHeader=False, col=0, datatype='float', h2oSummary2=None,
h2oQuantilesApprox=None, h2oQuantilesExact=None, interpolate='linear', quantile=0.50):
SCIPY_INSTALLED = True
try:
import scipy as sp
import numpy as np
print "Both numpy and scipy are installed. Will do extra checks"
except ImportError:
print "numpy or scipy is not installed. Will only do sort-based checking"
SCIPY_INSTALLED = false
target = h2o_util.file_read_csv_col(csvPathname, col=col, datatype=datatype,
skipHeader=skipHeader, preview=5)
if datatype=='float':
# to make irene's R runif files first col work (quoted row numbers, integers
#shouldn't hurt anyone else?
# strip " from left (ignore leading whitespace
# strip " from right (ignore leading whitespace
targetFP= map(float, target)
# targetFP= np.array(tFP, np.float)
if datatype=='int':
targetFP= map(int, target)
# http://docs.scipy.org/doc/numpy-dev/reference/generated/numpy.percentile.html
# numpy.percentile has simple linear interpolate and midpoint
# need numpy 1.9 for interpolation. numpy 1.8 doesn't have
# p = np.percentile(targetFP, 50 if DO_MEDIAN else 99.9, interpolation='midpoint')
# 1.8
if SCIPY_INSTALLED:
p = np.percentile(targetFP, quantile*100)
h2p.red_print("numpy.percentile", p)
# per = [100 * t for t in thresholds]
from scipy import stats
s1 = stats.scoreatpercentile(targetFP, quantile*100)
h2p.red_print("scipy stats.scoreatpercentile", s1)
# scipy apparently doesn't have the use of means (type 2)
# http://en.wikipedia.org/wiki/Quantile
# it has median (R-8) with 1/3, 1/3
if 1==0:
# type 6
alphap=0
betap=0
# type 5 okay but not perfect
alphap=0.5
betap=0.5
# type 8
alphap=1/3.0
betap=1/3.0
if interpolate=='mean':
# an approx? (was good when comparing to h2o type 2)
alphap=0.4
betap=0.4
if interpolate=='linear':
# this is type 7
alphap=1
betap=1
s2List = stats.mstats.mquantiles(targetFP, prob=quantile, alphap=alphap, betap=betap)
s2 = s2List[0]
# http://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.mstats.mquantiles.html
# type 7
# alphap=0.4, betap=0.4,
# type 2 not available? (mean)
# alphap=1/3.0, betap=1/3.0 is approx median?
h2p.red_print("scipy stats.mstats.mquantiles:", s2)
# also get the median with a painful sort (h2o_summ.percentileOnSortedlist()
# inplace sort
targetFP.sort()
# this matches scipy type 7 (linear)
# b = h2o_summ.percentileOnSortedList(targetFP, 0.50 if DO_MEDIAN else 0.999, interpolate='linear')
# this matches h2o type 2 (mean)
# b = h2o_summ.percentileOnSortedList(targetFP, 0.50 if DO_MEDIAN else 0.999, interpolate='mean')
b = percentileOnSortedList(targetFP, quantile, interpolate='linear')
label = str(quantile * 100) + '%'
h2p.blue_print(label, "from sort:", b)
if SCIPY_INSTALLED:
h2p.blue_print(label, "from numpy:", p)
h2p.blue_print(label, "from scipy 1:", s1)
h2p.blue_print(label, "from scipy 2:", s2)
h2p.blue_print(label, "from h2o summary:", h2oSummary2)
h2p.blue_print(label, "from h2o multipass:"******"from h2o singlepass:"******"h2oQuantilesApprox is unexpectedly NaN %s" % h2oQuantilesApprox)
h2o_util.assertApproxEqual(h2oQuantilesApprox, b, rel=0.5,
msg='h2o quantile singlepass is not approx. same as sort algo')
if h2oQuantilesExact:
if math.isnan(float(h2oQuantilesExact)):
raise Exception("h2oQuantilesExact is unexpectedly NaN %s" % h2oQuantilesExact)
h2o_util.assertApproxEqual(h2oQuantilesExact, b, tol=0.0000002,
msg='h2o quantile multipass is not approx. same as sort algo')
if h2oSummary2:
if math.isnan(float(h2oSummary2)):
raise Exception("h2oSummary2 is unexpectedly NaN %s" % h2oSummary2)
h2o_util.assertApproxEqual(h2oSummary2, b, rel=0.5,
msg='h2o summary2 is not approx. same as sort algo')
if SCIPY_INSTALLED:
if h2oQuantilesApprox:
h2o_util.assertApproxEqual(h2oQuantilesExact, p, tol=0.0000002,
msg='h2o quantile multipass is not same as numpy.percentile')
h2o_util.assertApproxEqual(h2oQuantilesExact, s1, tol=0.0000002,
msg='h2o quantile multipass is not same as scipy stats.scoreatpercentile')
# give us some slack compared to the scipy use of median (instead of desired mean)
if h2oQuantilesExact:
if interpolate=='mean':
h2o_util.assertApproxEqual(h2oQuantilesExact, s2, rel=0.01,
msg='h2o quantile multipass is not approx. same as scipy stats.mstats.mquantiles')
else:
h2o_util.assertApproxEqual(h2oQuantilesExact, s2, tol=0.0000002,
msg='h2o quantile multipass is not same as scipy stats.mstats.mquantiles')
# see if scipy changes. nope. it doesn't
if 1==0:
a = stats.mstats.mquantiles(targetFP, prob=quantile, alphap=alphap, betap=betap)
h2p.red_print("after sort")
h2p.red_print("scipy stats.mstats.mquantiles:", s3)
def test_exec_enums_rand_cut2(self):
h2o.beta_features = True
SYNDATASETS_DIR = h2o.make_syn_dir()
n = ROWS
tryList = [
# (n, 10, 9, 'cE', 300),
(n, 1, 1, 'cE', 300),
]
# create key names to use for exec
eKeys = ['e%s' % i for i in range(10)]
# h2b.browseTheCloud()
trial = 0
for (rowCount, iColCount, oColCount, hex_key, timeoutSecs) in tryList:
colCount = iColCount + oColCount
hex_key = 'p'
colEnumList = create_col_enum_list(iColCount)
# create 100 possible cut expressions here, so we don't waste time below
rowExprList = []
print "Creating", CUT_EXPR_CNT, 'cut expressions'
for j in range(CUT_EXPR_CNT):
# init cutValue. None means no compare
cutValue = [None for i in range(iColCount)]
# build up a random cut expression
MAX_COLS_IN_EXPR = iColCount
cols = random.sample(range(MAX_COLS_IN_EXPR),
random.randint(1, MAX_COLS_IN_EXPR))
for c in cols:
# possible choices within the column
cel = colEnumList[c]
# for now the cutValues are numbers for the enum mappings
if 1 == 1:
# FIX! hack. don't use encoding 0, maps to NA here? h2o doesn't like
celChoice = str(random.choice(range(len(cel))))
else:
celChoice = random.choice(cel)
cutValue[c] = celChoice
cutExprList = []
for i, c in enumerate(cutValue):
if c is None:
continue
else:
# new ...ability to reference cols
# src[ src$age<17 && src$zip=95120 && ... , ]
# randomly pick == or !=
if random.randint(0, 1) == 0:
cutExprList.append('p$C' + str(i + 1) + '!=' + c)
else:
cutExprList.append('p$C' + str(i + 1) + '==' + c)
cutExpr = ' & '.join(cutExprList)
# print "cutExpr:", cutExpr
# just extract one output col (the first one)
rowExpr = '%s[%s,%s];' % (hex_key, cutExpr, iColCount + 1)
# print "rowExpr:", rowExpr
print rowExpr
rowExprList.append(rowExpr)
# CREATE DATASET*******************************************
SEEDPERFILE = random.randint(0, sys.maxint)
csvFilename = 'syn_enums_' + str(rowCount) + 'x' + str(
colCount) + '.csv'
csvPathname = SYNDATASETS_DIR + '/' + csvFilename
print "Creating random", csvPathname
write_syn_dataset(csvPathname,
rowCount,
iColCount,
oColCount,
SEEDPERFILE,
colEnumList=colEnumList)
# PARSE*******************************************************
src_key = csvFilename
parseResult = h2i.import_only(path=csvPathname,
schema='put',
src_key='A' + src_key,
timeoutSecs=200)
parseResult = h2i.import_only(path=csvPathname,
schema='put',
src_key='B' + src_key,
timeoutSecs=200)
parseResult = h2i.import_only(path=csvPathname,
schema='put',
src_key='C' + src_key,
timeoutSecs=200)
parseResult = h2i.import_only(path=csvPathname,
schema='put',
src_key='D' + src_key,
timeoutSecs=200)
parseResult = h2i.import_only(path=csvPathname,
schema='put',
src_key='E' + src_key,
timeoutSecs=200)
parseResult = h2i.import_only(path=csvPathname,
schema='put',
src_key='F' + src_key,
timeoutSecs=200)
parseResult = h2i.import_only(path=csvPathname,
schema='put',
src_key='G' + src_key,
timeoutSecs=200)
parseResult = h2i.import_only(path=csvPathname,
schema='put',
src_key='H' + src_key,
timeoutSecs=200)
parseResult = h2i.import_only(path=csvPathname,
schema='put',
src_key='I' + src_key,
timeoutSecs=200)
parseResult = h2i.import_only(path=csvPathname,
schema='put',
src_key='J' + src_key,
timeoutSecs=200)
parseResult = h2i.parse_only(pattern='*' + src_key,
hex_key=hex_key,
timeoutSecs=800)
print "Parse result['destination_key']:", parseResult[
'destination_key']
inspect = h2o_cmd.runInspect(key=parseResult['destination_key'])
h2o_cmd.infoFromInspect(inspect, csvPathname)
pNumRows = inspect['numRows']
pNumCols = inspect['numCols']
# print h2o.dump_json(inspect)
levels = h2o.nodes[0].levels(source=hex_key)
print "levels result:", h2o.dump_json(levels)
(missingValuesDict, constantValuesDict, enumSizeDict, colTypeDict, colNameDict) = \
h2o_cmd.columnInfoFromInspect(parseResult['destination_key'], exceptionOnMissingValues=False)
# error if any col has constant values
if len(constantValuesDict) != 0:
raise Exception(
"Probably got a col NA'ed and constant values as a result %s"
% constantValuesDict)
# INIT all possible key names used***************************
# remember. 1 indexing!
# is this needed?
if 1 == 1:
a = 'a=c(1,2,3);' + ';'.join(
['a[,%s]=a[,%s-1]' % (i, i) for i in range(2, colCount)])
print a
for eKey in eKeys:
# build up the columns
e = h2o.nodes[0].exec_query(str='%s;%s=a' % (a, eKey),
print_params=False)
## print h2o.dump_json(e)
xList = []
eList = []
fList = []
for repeat in range(CUT_LOOP_CNT):
# EXEC*******************************************************
# don't use exec_expr to avoid issues with Inspect following etc.
randICol = random.randint(0, iColCount - 1)
randOCol = random.randint(iColCount, iColCount + oColCount - 1)
# should be two different keys in the sample
e = random.sample(eKeys, 2)
fKey = e[0]
eKey = e[1]
start = time.time()
h2o.nodes[0].exec_query(str="%s=%s" %
(fKey, random.choice(rowExprList)))
elapsed = time.time() - start
execTime = elapsed
print "exec 2 took", elapsed, "seconds."
inspect = h2o_cmd.runInspect(key=fKey)
h2o_cmd.infoFromInspect(inspect, fKey)
numRows = inspect['numRows']
numCols = inspect['numCols']
if numRows == 0 or numCols != colCount:
h2p.red_print("Warning: Cut resulted in", numRows,
"rows and", numCols,
"cols. Quantile will abort")
# QUANTILE*******************************************************
quantile = 0.5 if DO_MEDIAN else .999
# first output col. always fed by an exec cut, so 0?
column = iColCount
column = 0
start = time.time()
q = h2o.nodes[0].quantiles(source_key=fKey,
column=column,
quantile=quantile,
max_qbins=MAX_QBINS,
multiple_pass=MULTI_PASS)
h2p.red_print("quantile", quantile, q['result'])
elapsed = time.time() - start
print "quantile end on ", fKey, 'took', elapsed, 'seconds.'
quantileTime = elapsed
# remove all keys*******************************************************
# what about hex_key?
if 1 == 0:
start = time.time()
h2o.nodes[0].remove_all_keys()
elapsed = time.time() - start
print "remove all keys end on ", csvFilename, 'took', elapsed, 'seconds.'
trial += 1
xList.append(trial)
eList.append(execTime)
fList.append(quantileTime)
#****************************************************************
# QUANTILE APPROX. BASELINE FOR SINGLE COL WALK FULL DATASET
print "QUANTILE APPROX. BASELINE FOR SINGLE COL WALK FULL DATASET. Although it's a real col, not an enum col"
quantile = 0.5 if DO_MEDIAN else .999
# first output col. always fed by an exec cut, so 0?
column = iColCount
start = time.time()
q = h2o.nodes[0].quantiles(source_key=hex_key,
column='C' + str(iColCount + 1),
quantile=quantile,
max_qbins=MAX_QBINS,
multiple_pass=0)
elapsed = time.time() - start
h2p.red_print(
hex_key, pNumRows,
"rows Baseline: quantile single col (C" + str(iColCount + 1) + ")",
"one iteration", elapsed, "secs. threshold:", quantile,
q['result'])
print "quantile single col 1 iteration end on", hex_key, "took", elapsed, 'seconds.'
quantileTime = elapsed
#****************************************************************
# PLOTS. look for eplot.jpg and fplot.jpg in local dir?
if DO_PLOT:
xLabel = 'trial'
eLabel = 'exec cut time'
fLabel = 'quantile time'
eListTitle = ""
fListTitle = ""
h2o_gbm.plotLists(xList,
xLabel,
eListTitle,
eList,
eLabel,
fListTitle,
fList,
fLabel,
server=True)
def test_exec_enums_rand_cut(self):
h2o.beta_features = True
SYNDATASETS_DIR = h2o.make_syn_dir()
tryList = [
(ROWS, 3, 2, 'cE', 300),
]
# create key names to use for exec
eKeys = ['e%s' % i for i in range(10)]
# h2b.browseTheCloud()
trial = 0
for (rowCount, iColCount, oColCount, hex_key, timeoutSecs) in tryList:
colCount = iColCount + oColCount
hex_key = 'p'
colEnumList = create_col_enum_list(iColCount)
# create 100 possible cut expressions here, so we don't waste time below
rowExprList = []
for j in range(CUT_EXPR_CNT):
print "Creating", CUT_EXPR_CNT, 'cut expressions'
# init cutValue. None means no compare
cutValue = [None for i in range(iColCount)]
# build up a random cut expression
cols = random.sample(range(iColCount), random.randint(1,iColCount))
for c in cols:
# possible choices within the column
# cel = colEnumList[c]
cel = colEnumList
# for now the cutValues are numbers for the enum mappings
if 1==1:
# FIX! hack. don't use encoding 0, maps to NA here? h2o doesn't like
celChoice = str(random.choice(range(len(cel))))
else:
celChoice = random.choice(cel)
cutValue[c] = celChoice
cutExprList = []
for i,c in enumerate(cutValue):
if c is None:
continue
else:
# new ...ability to reference cols
# src[ src$age<17 && src$zip=95120 && ... , ]
cutExprList.append('p$C'+str(i+1)+'=='+c)
cutExpr = ' && '.join(cutExprList)
print "cutExpr:", cutExpr
# should be two different keys in the sample
e = random.sample(eKeys,2)
fKey = e[0]
eKey = e[1]
rowExpr = '%s[%s,];' % (hex_key, cutExpr)
print "rowExpr:", rowExpr
rowExprList.append(rowExpr)
print "j:", j
# CREATE DATASET*******************************************
SEEDPERFILE = random.randint(0, sys.maxint)
csvFilename = 'syn_enums_' + str(rowCount) + 'x' + str(colCount) + '.csv'
csvPathname = SYNDATASETS_DIR + '/' + csvFilename
print "Creating random", csvPathname
write_syn_dataset(csvPathname, rowCount, iColCount, oColCount, SEEDPERFILE, colEnumList=colEnumList)
# PARSE*******************************************************
parseResult = h2i.import_parse(path=csvPathname, schema='put', hex_key=hex_key, timeoutSecs=30, doSummary=False, header=0)
print "Parse result['destination_key']:", parseResult['destination_key']
inspect = h2o_cmd.runInspect(key=parseResult['destination_key'])
h2o_cmd.infoFromInspect(inspect, csvPathname)
# print h2o.dump_json(inspect)
rSummary = h2o_cmd.runSummary(key=parseResult['destination_key'])
h2o_cmd.infoFromSummary(rSummary)
(missingValuesDict, constantValuesDict, enumSizeDict, colTypeDict, colNameDict) = \
h2o_cmd.columnInfoFromInspect(parseResult['destination_key'], exceptionOnMissingValues=False)
# error if any col has constant values
if len(constantValuesDict) != 0:
raise Exception("Probably got a col NA'ed and constant values as a result %s" % constantValuesDict)
# INIT all possible key names used***************************
# remember. 1 indexing!
# is this needed?
if 1==1:
a = 'a=c(1,2,3);' + ';'.join(['a[,%s]=a[,%s-1]'% (i,i) for i in range(2,colCount)])
print a
for eKey in eKeys:
# build up the columns
e = h2o.nodes[0].exec_query(str='%s;%s=a' % (a, eKey), print_params=False)
## print h2o.dump_json(e)
xList = []
eList = []
fList = []
for repeat in range(200):
# EXEC*******************************************************
# don't use exec_expr to avoid issues with Inspect following etc.
randICol = random.randint(0,iColCount-1)
randOCol = random.randint(iColCount, iColCount+oColCount-1)
# should be two different keys in the sample
e = random.sample(eKeys,2)
fKey = e[0]
eKey = e[1]
if 1==0:
start = time.time()
e = h2o.nodes[0].exec_query(str='%s=%s[,%s]' % (fKey, hex_key, randOCol+1))
elapsed = time.time() - start
print "exec 1 took", elapsed, "seconds."
execTime = elapsed
if 1==1:
start = time.time()
h2o.nodes[0].exec_query(str="%s=%s" % (fKey, random.choice(rowExprList)))
elapsed = time.time() - start
execTime = elapsed
print "exec 2 took", elapsed, "seconds."
if 1==0:
gKey = random.choice(eKeys)
# do a 2nd random to see if things blow up
start = time.time()
h2o.nodes[0].exec_query(str="%s=%s" % (gKey, fKey))
elapsed = time.time() - start
print "exec 3 took", elapsed, "seconds."
if 1==1:
inspect = h2o_cmd.runInspect(key=fKey)
h2o_cmd.infoFromInspect(inspect, fKey)
numRows = inspect['numRows']
numCols = inspect['numCols']
if numRows==0 or numCols!=colCount:
h2p.red_print("Warning: Cut resulted in", numRows, "rows and", numCols, "cols. Quantile will abort")
# QUANTILE*******************************************************
quantile = 0.5 if DO_MEDIAN else .999
# first output col. always fed by an exec cut, so 0?
column = iColCount
start = time.time()
q = h2o.nodes[0].quantiles(source_key=fKey, column=column,
quantile=quantile, max_qbins=MAX_QBINS, multiple_pass=MULTI_PASS)
h2p.red_print("quantile", quantile, q['result'])
elapsed = time.time() - start
print "quantile end on ", fKey, 'took', elapsed, 'seconds.'
quantileTime = elapsed
# remove all keys*******************************************************
# what about hex_key?
if 1==0:
start = time.time()
h2o.nodes[0].remove_all_keys()
elapsed = time.time() - start
print "remove all keys end on ", csvFilename, 'took', elapsed, 'seconds.'
trial += 1
xList.append(trial)
eList.append(execTime)
fList.append(quantileTime)
# just get a plot of the last one (biggest)
if DO_PLOT:
xLabel = 'trial'
eLabel = 'exec cut time'
fLabel = 'quantile time'
eListTitle = ""
fListTitle = ""
h2o_gbm.plotLists(xList, xLabel, eListTitle, eList, eLabel, fListTitle, fList, fLabel)
def __do_json_request(self, jsonRequest=None, fullUrl=None, timeout=10, params=None, postData=None, returnFast=False,
cmd='get', extraComment=None, ignoreH2oError=False, noExtraErrorCheck=False, raiseIfNon200=True, **kwargs):
H2O.verboseprint("__do_json_request, timeout: " + str(timeout))
# if url param is used, use it as full url. otherwise crate from the jsonRequest
if fullUrl:
url = fullUrl
else:
url = self.__url(jsonRequest)
# remove any params that are 'None'
# need to copy dictionary, since can't delete while iterating
if params is not None:
params_serialized = params.copy()
for k in params_serialized:
if params_serialized[k] is None:
del params[k]
paramsStr = '?' + '&'.join(['%s=%s' % (k, v) for (k, v) in params.items()])
else:
paramsStr = ''
# The requests package takes array parameters and explodes them: ['f00', 'b4r'] becomes "f00,b4r".
# NOTE: this handles 1D arrays only; if we need ND this needs to be recursive.
# NOTE: we currently don't need to do this for GET, so that's not implemented.
if postData is not None:
munged_postData = {}
for k, v in postData.iteritems():
if type(v) is list:
if len(v) == 0:
munged_postData[k] = '[]'
else:
first = True
array_str = '['
for val in v:
if not first: array_str += ', '
if val is None:
array_str += 'null'
elif isinstance(val, basestring):
array_str += "\"" + str(val) + "\""
else:
array_str += str(val)
first = False
array_str += ']'
munged_postData[k] = array_str
else:
# not list:
munged_postData[k] = v
else:
# None
munged_postData = postData
if extraComment:
log('Start ' + url + paramsStr, comment=extraComment)
else:
log('Start ' + url + paramsStr)
log_rest("")
log_rest("----------------------------------------------------------------------\n")
if extraComment:
log_rest("# Extra comment info about this request: " + extraComment)
if cmd == 'get':
log_rest("GET")
else:
log_rest("POST")
log_rest(url + paramsStr)
# file get passed thru kwargs here
try:
if 'post' == cmd:
# NOTE == cmd: for now, since we don't have deserialization from JSON in h2o-dev, we use form-encoded POST.
# This is temporary.
#
# This following does application/json (aka, posting JSON in the body):
# r = requests.post(url, timeout=timeout, params=params, data=json.dumps(munged_postData), **kwargs)
#
# This does form-encoded, which doesn't allow POST of nested structures
r = requests.post(url, timeout=timeout, params=params, data=munged_postData, **kwargs)
elif 'delete' == cmd:
r = requests.delete(url, timeout=timeout, params=params, **kwargs)
elif 'get' == cmd:
r = requests.get(url, timeout=timeout, params=params, **kwargs)
else:
raise ValueError("Unknown HTTP command (expected 'get', 'post' or 'delete'): " + cmd)
except Exception, e:
# rethrow the exception after we've checked for stack trace from h2o
# out of memory errors maybe don't show up right away? so we should wait for h2o
# to get it out to h2o stdout. We don't want to rely on cloud teardown to check
# because there's no delay, and we don't want to delay all cloud teardowns by waiting.
# (this is new/experimental)
exc_info = sys.exc_info()
# use this to ignore the initial connection errors during build cloud when h2o is coming up
if not noExtraErrorCheck:
h2p.red_print(
"ERROR: got exception on %s to h2o. \nGoing to check sandbox, then rethrow.." % (url + paramsStr))
time.sleep(2)
H2O.check_sandbox_for_errors(python_test_name=H2O.python_test_name);
log_rest("")
log_rest("EXCEPTION CAUGHT DOING REQUEST: " + str(e.message))
raise exc_info[1], None, exc_info[2]
H2O.verboseprint("r: " + repr(r))
def __do_json_request(self, jsonRequest=None, fullUrl=None, timeout=10, params=None, postData=None, returnFast=False,
cmd='get', extraComment=None, ignoreH2oError=False, noExtraErrorCheck=False, raiseIfNon200=True, suppressErrorMsg=False, **kwargs):
H2O.verboseprint("__do_json_request, timeout: " + str(timeout))
# if url param is used, use it as full url. otherwise crate from the jsonRequest
if fullUrl:
url = fullUrl
else:
url = self.__url(jsonRequest)
# remove any params that are 'None'
# need to copy dictionary, since can't delete while iterating
if params is not None:
params_serialized = params.copy()
for k in params_serialized:
if params_serialized[k] is None:
del params[k]
paramsStr = '?' + '&'.join(['%s=%s' % (k, v) for (k, v) in params.items()])
else:
paramsStr = ''
# The requests package takes array parameters and explodes them: ['f00', 'b4r'] becomes "f00,b4r".
# NOTE: this handles 1D arrays only; if we need ND this needs to be recursive.
# NOTE: we currently don't need to do this for GET, so that's not implemented.
if postData is not None:
munged_postData = {}
for k, v in postData.iteritems():
if type(v) is list:
if len(v) == 0:
munged_postData[k] = '[]'
else:
first = True
array_str = '['
for val in v:
if not first: array_str += ', '
if val is None:
array_str += 'null'
elif isinstance(val, basestring):
array_str += "\"" + str(val) + "\""
else:
array_str += str(val)
first = False
array_str += ']'
munged_postData[k] = array_str
elif type(v) is dict:
if len(v) == 0:
munged_postData[k] = '{}'
else:
first = True
map_str = '{'
for key, val in v.iteritems():
if not first: map_str += ', '
if val is None:
map_str += "\"" + key + "\"" + ': null'
elif isinstance(val, basestring):
map_str += "\"" + str(key) + "\"" + ":" + "\"" + str(val) + "\""
else:
map_str += "\"" + key + "\"" + ':' + str(val)
first = False
map_str += '}'
munged_postData[k] = map_str
else:
# not list:
munged_postData[k] = v
else:
# None
munged_postData = postData
# print("munged_postData: " + repr(munged_postData))
if extraComment:
log('Start ' + url + paramsStr, comment=extraComment)
else:
log('Start ' + url + paramsStr)
log_rest("")
log_rest("----------------------------------------------------------------------\n")
if extraComment:
log_rest("# Extra comment info about this request: " + extraComment)
if cmd == 'get':
log_rest("GET")
else:
log_rest("POST")
log_rest(url + paramsStr)
# file get passed thru kwargs here
try:
if 'post' == cmd:
# NOTE == cmd: for now, since we don't have deserialization from JSON in h2o-dev, we use form-encoded POST.
# This is temporary.
#
# This following does application/json (aka, posting JSON in the body):
# r = requests.post(url, timeout=timeout, params=params, data=json.dumps(munged_postData), **kwargs)
#
# This does form-encoded, which doesn't allow POST of nested structures
r = requests.post(url, timeout=timeout, params=params, data=munged_postData, **kwargs)
elif 'delete' == cmd:
r = requests.delete(url, timeout=timeout, params=params, **kwargs)
elif 'get' == cmd:
r = requests.get(url, timeout=timeout, params=params, **kwargs)
else:
raise ValueError("Unknown HTTP command (expected 'get', 'post' or 'delete'): " + cmd)
except Exception as e:
# rethrow the exception after we've checked for stack trace from h2o
# out of memory errors maybe don't show up right away? so we should wait for h2o
# to get it out to h2o stdout. We don't want to rely on cloud teardown to check
# because there's no delay, and we don't want to delay all cloud teardowns by waiting.
# (this is new/experimental)
exc_info = sys.exc_info()
# use this to ignore the initial connection errors during build cloud when h2o is coming up
if not noExtraErrorCheck:
h2p.red_print(
"ERROR: got exception on %s to h2o. \nGoing to check sandbox, then rethrow.." % (url + paramsStr))
time.sleep(2)
H2O.check_sandbox_for_errors(python_test_name=H2O.python_test_name);
log_rest("")
log_rest("EXCEPTION CAUGHT DOING REQUEST: " + str(e.message))
raise (exc_info[1], None, exc_info[2])
H2O.verboseprint("r: " + repr(r))
if 200 != r.status_code:
pp = pprint.PrettyPrinter(indent=4)
msg = "JSON call returned non-200 status: " + url
json = r.json()
if None != json and 'dev_msg' in json:
msg += "\ndev_msg: "
msg += str(json['dev_msg'])
msg += "\nr.status_code: " + str(r.status_code)
msg += "\nr.headers: " + repr(r.headers)
if None == json:
msg += '\nERROR: the error output from H2O is not JSON!'
msg += "\nr.text: " + r.text
else:
msg += "\nr.json: "
msg += pp.pformat(json)
if raiseIfNon200:
pass # we'll pass msg up with the exception
elif not suppressErrorMsg:
print(msg)
log_rest(msg)
log_rest("")
try:
if r is None:
log_rest("r is None")
else:
log_rest("HTTP status code: " + str(r.status_code))
# The following accesses to r.text were taking most of the runtime:
log_text = False
if log_text:
if hasattr(r, 'text'):
if r.text is None:
log_rest("r.text is None")
else:
log_rest(r.text)
else:
log_rest("r does not have attr text")
except Exception as e:
# Paranoid exception catch.
# Ignore logging exceptions in the case that the above error checking isn't sufficient.
print("Caught exception from result logging: ", e, "; result: ", repr(r))
# fatal if no response
if raiseIfNon200 and not r:
raise Exception("Maybe bad url? no r in __do_json_request in %s:" % inspect.stack()[1][3] + "\n\n" + msg)
# this is used to open a browser on results, or to redo the operation in the browser
# we don't' have that may urls flying around, so let's keep them all
H2O.json_url_history.append(r.url)
# if r.json():
# raise Exception("Maybe bad url? no r.json in __do_json_request in %s:" % inspect.stack()[1][3])
rjson = None
if returnFast:
return
try:
rjson = r.json()
except:
print(h2o_test_utils.dump_json(r.text))
if not isinstance(r, (list, dict)):
raise Exception("h2o json responses should always be lists or dicts, see previous for text")
raise Exception("Could not decode any json from the request.")
# TODO
# TODO
# TODO
# TODO: we should really only look in the response object. This check
# prevents us from having a field called "error" (e.g., for a scoring result).
for e in ['error', 'Error', 'errors', 'Errors']:
# error can be null (python None). This happens in exec2
if e in rjson and rjson[e]:
H2O.verboseprint("rjson:" + h2o_test_utils.dump_json(rjson))
emsg = 'rjson %s in %s: %s' % (e, inspect.stack()[1][3], rjson[e])
if ignoreH2oError:
# well, we print it..so not totally ignore. test can look at rjson returned
print(emsg)
else:
print(emsg)
raise Exception(emsg)
for w in ['warning', 'Warning', 'warnings', 'Warnings']:
# warning can be null (python None).
if w in rjson and rjson[w]:
H2O.verboseprint(dump_json(rjson))
print('rjson %s in %s: %s' % (w, inspect.stack()[1][3], rjson[w]))
# Allow the caller to check things like __http_request.status_code.
# The response object is not JSON-serializable, so we capture the fields we want here:
response = {}
# response['headers'] = r.headers
response['url'] = r.url
response['status_code'] = r.status_code
response['text'] = r.text
rjson['__http_response'] = response
return rjson
