def testCheckWithModelAPI(self):
######################################################################
# Now look for kmeans_model_name using the one-model API and find_compatible_frames, and check it
model = self.a_node.models(key=self.kmeans_model_name,
find_compatible_frames=True)
found_kmeans = False
h2o_util.assertKeysExist(model['models'][0], '', ['compatible_frames'])
assert self.prostate_key in model['models'][0]['compatible_frames'], \
"Failed to find " + self.prostate_key + " in compatible_frames list."
######################################################################
# Now look for prostate_key using the one-frame API and find_compatible_models, and check it
result = self.a_node.frames(key='prostate.hex',
find_compatible_models=True)
frames = result['frames']
frames_dict = h2o_util.list_to_dict(frames, 'key/name')
assert_true('prostate.hex' in frames_dict,
"Failed to find prostate.hex in Frames list.")
compatible_models = result['compatible_models']
models_dict = h2o_util.list_to_dict(compatible_models, 'key')
assert_true ( self.dl_prostate_model_name in models_dict, "Failed to find " + \
self.dl_prostate_model_name + " in compatible models list.")
assert_true(
self.dl_prostate_model_name in frames[0]['compatible_models'])
assert_true(self.kmeans_model_name in frames[0]['compatible_models'])
def validate_predictions(result, model_name, frame_key, expected_rows):
'''
Validate a /Predictions result.
'''
assert p is not None, "FAIL: Got a null result for scoring: " + model_name + " on: " + frame_key
assert 'model_metrics' in p, "FAIL: Predictions for scoring: " + model_name + " on: " + frame_key + " does not contain a model_metrics object."
mm = p['model_metrics'][0]
h2o.H2O.verboseprint('mm: ', repr(mm))
assert 'auc' in mm, "FAIL: Predictions for scoring: " + model_name + " on: " + frame_key + " does not contain an AUC."
assert 'cm' in mm, "FAIL: Predictions for scoring: " + model_name + " on: " + frame_key + " does not contain a CM."
assert 'predictions' in mm, "FAIL: Predictions for scoring: " + model_name + " on: " + frame_key + " does not contain an predictions section."
assert 'key' in mm['predictions'], "FAIL: Predictions for scoring: " + model_name + " on: " + frame_key + " does not contain a key."
assert 'name' in mm['predictions']['key'], "FAIL: Predictions for scoring: " + model_name + " on: " + frame_key + " does not contain a key name."
predictions_key = mm['predictions']['key']['name']
result = a_node.frames(key=predictions_key, find_compatible_models=True, len=5)
frames = result['frames']
frames_dict = h2o_util.list_to_dict(frames, 'key/name')
assert predictions_key in frames_dict, "FAIL: Failed to find predictions key" + predictions_key + " in Frames list."
predictions = mm['predictions']
h2o.H2O.verboseprint('p: ', repr(p))
assert 'columns' in predictions, "FAIL: Predictions for scoring: " + model_name + " on: " + frame_key + " does not contain an columns section."
assert len(predictions['columns']) > 0, "FAIL: Predictions for scoring: " + model_name + " on: " + frame_key + " does not contain any columns."
assert 'label' in predictions['columns'][0], "FAIL: Predictions for scoring: " + model_name + " on: " + frame_key + " column 0 has no label element."
assert 'predict' == predictions['columns'][0]['label'], "FAIL: Predictions for scoring: " + model_name + " on: " + frame_key + " column 0 is not 'predict'."
assert expected_rows == predictions['rows'], "FAIL: Predictions for scoring: " + model_name + " on: " + frame_key + " has an unexpected number of rows."
def validate_predictions(result, model_name, frame_key, expected_rows, destination_key=None):
'''
Validate a /Predictions result.
'''
assert p is not None, "FAIL: Got a null result for scoring: " + model_name + " on: " + frame_key
assert 'model_metrics' in p, "FAIL: Predictions for scoring: " + model_name + " on: " + frame_key + " does not contain a model_metrics object."
mm = p['model_metrics'][0]
h2o.H2O.verboseprint('mm: ', repr(mm))
#assert 'auc' in mm, "FAIL: Predictions for scoring: " + model_name + " on: " + frame_key + " does not contain an AUC."
#assert 'cm' in mm, "FAIL: Predictions for scoring: " + model_name + " on: " + frame_key + " does not contain a CM."
assert 'predictions' in mm, "FAIL: Predictions for scoring: " + model_name + " on: " + frame_key + " does not contain an predictions section."
assert 'key' in mm['predictions'], "FAIL: Predictions for scoring: " + model_name + " on: " + frame_key + " does not contain a key."
assert 'name' in mm['predictions']['key'], "FAIL: Predictions for scoring: " + model_name + " on: " + frame_key + " does not contain a key name."
predictions_key = mm['predictions']['key']['name']
f = a_node.frames(key=predictions_key, find_compatible_models=True, row_count=5)
frames = f['frames']
frames_dict = h2o_util.list_to_dict(frames, 'key/name')
assert predictions_key in frames_dict, "FAIL: Failed to find predictions key" + predictions_key + " in Frames list."
predictions = mm['predictions']
h2o.H2O.verboseprint('p: ', repr(p))
assert 'columns' in predictions, "FAIL: Predictions for scoring: " + model_name + " on: " + frame_key + " does not contain an columns section."
assert len(predictions['columns']) > 0, "FAIL: Predictions for scoring: " + model_name + " on: " + frame_key + " does not contain any columns."
assert 'label' in predictions['columns'][0], "FAIL: Predictions for scoring: " + model_name + " on: " + frame_key + " column 0 has no label element."
assert 'predict' == predictions['columns'][0]['label'], "FAIL: Predictions for scoring: " + model_name + " on: " + frame_key + " column 0 is not 'predict'."
assert expected_rows == predictions['rows'], "FAIL: Predictions for scoring: " + model_name + " on: " + frame_key + " has an unexpected number of rows."
assert 'destination_key' in result, "FAIL: failed to find 'destination_key' in predict result:" + h2o_util.dump_json(result)
assert 'name' in result['destination_key'], "FAIL: failed to find name in 'destination_key' in predict result:" + h2o_util.dump_json(result)
if destination_key is not None:
assert destination_key == result['destination_key']['name'], "FAIL: bad value for 'destination_key' in predict result; expected: " + destination_key + ", got: " + result['destination_key']['name']
def test_simple2(self):
# h2o-dev doesn't take ../.. type paths? make find_file return absolute path
# csvPathname = find_file("bigdata/laptop/poker-hand-testing.data")
csvPathname = find_file("smalldata/logreg/prostate.csv")
import_result = h2o.n0.import_files(path=csvPathname)
# print dump_json(import_result)
k = import_result['keys'][0]
frames_result = h2o.n0.frames(key=k)
frame = frames_result['frames'][0]
rows = frame['rows']
columns = frame['columns']
for c in columns:
label = c['label']
missing = c['missing_count']
stype = c['type']
domain = c['domain']
# print dump_json(frame)
# let's see what ray's util does
frames = h2o.n0.frames()['frames']
frames_dict = h2o_util.list_to_dict(frames, 'key/name')
# print "frames:", dump_json(frames)
# print "frames_dict:", dump_json(frames_dict)
for k,v in frames_dict.items():
print "frames_dict key:", k
# interesting. we can do dictionary comprehensions
# { k:v for k,v in my_dict.items() if 'Peter' in k }
# how do you parse multiple files
parse_result = h2o.n0.parse(key=k, intermediateResults=DO_INTERMEDIATE_RESULTS)
frame = parse_result['frames'][0]
hex_key = frame['key']['name']
colCount = 9
rowCount = 380
# colCount = 11
# rowCount = 1000000
start = time.time()
inspect = h2o_cmd.runInspect(None, hex_key)
print "Inspect:", hex_key, "took", time.time() - start, "seconds"
numCols = len(inspect['frames'][0]['columns'])
numRows = inspect['frames'][0]['rows']
print "\n" + csvPathname, \
" rows:", "{:,}".format(numRows), \
" len(columns):", "{:,}".format(numCols)
# should match # of cols in header or ??
self.assertEqual(numCols, colCount,
"parse created result with the wrong number of cols %s %s" % (numCols, colCount))
self.assertEqual(numRows, rowCount,
"parse created result with the wrong number of rows (header shouldn't count) %s %s" % \
(numRows, rowCount))
verboseprint(hex_key, ":", dump_json(parse_result))
def testCheckWithModelAPI(self):
######################################################################
# Now look for kmeans_model_name using the one-model API and find_compatible_frames, and check it
model = self.a_node.models(key=self.kmeans_model_name, find_compatible_frames=True)
found_kmeans = False;
h2o_util.assertKeysExist(model['models'][0], '', ['compatible_frames'])
assert self.prostate_key in model['models'][0]['compatible_frames'], \
"Failed to find " + self.prostate_key + " in compatible_frames list."
######################################################################
# Now look for prostate_key using the one-frame API and find_compatible_models, and check it
result = self.a_node.frames(key='prostate.hex', find_compatible_models=True)
frames = result['frames']
frames_dict = h2o_util.list_to_dict(frames, 'key/name')
assert 'prostate.hex' in frames_dict, "Failed to find prostate.hex in Frames list."
compatible_models = result['compatible_models']
models_dict = h2o_util.list_to_dict(compatible_models, 'key')
assert self.dl_prostate_model_name in models_dict, "Failed to find " + \
self.dl_prostate_model_name + " in compatible models list."
assert self.dl_prostate_model_name in frames[0]['compatible_models']
assert self.kmeans_model_name in frames[0]['compatible_models']
def testImportProstate(self):
cleanup(self.a_node)
import_result = self.a_node.import_files(path="/Users/radu/h2o-dev/smalldata/logreg/prostate.csv")
parse_result = self.a_node.parse(key=import_result['keys'][0]) # TODO: handle multiple files
self.prostate_key = parse_result['frames'][0]['key']['name']
# Test /Frames for prostate.csv
frames = self.a_node.frames()['frames']
frames_dict = h2o_util.list_to_dict(frames, 'key/name')
assert 'prostate.hex' in frames_dict, "Failed to find prostate.hex in Frames list."
# Test /Frames/{key} for prostate.csv
frames = self.a_node.frames(key='prostate.hex')['frames']
frames_dict = h2o_util.list_to_dict(frames, 'key/name')
assert 'prostate.hex' in frames_dict, "Failed to find prostate.hex in Frames list."
columns_dict = h2o_util.list_to_dict(frames[0]['columns'], 'label')
assert 'CAPSULE' in columns_dict, "Failed to find CAPSULE in Frames/prostate.hex."
assert 'AGE' in columns_dict, "Failed to find AGE in Frames/prostate.hex/columns."
assert 'bins' in columns_dict['AGE'], "Failed to find bins in Frames/prostate.hex/columns/AGE."
assert None is columns_dict['AGE']['bins'], "Failed to clear bins field." # should be cleared except for /summary
frames = self.a_node.columns(key='prostate.hex')['frames']
columns_dict = h2o_util.list_to_dict(frames[0]['columns'], 'label')
assert 'ID' in columns_dict, "Failed to find ID in Frames/prostate.hex/columns."
assert 'AGE' in columns_dict, "Failed to find AGE in Frames/prostate.hex/columns."
assert 'bins' in columns_dict['AGE'], "Failed to find bins in Frames/prostate.hex/columns/AGE."
assert None is columns_dict['AGE']['bins'], "Failed to clear bins field." # should be cleared except for /summary
frames = self.a_node.column(key='prostate.hex', column='AGE')['frames']
columns_dict = h2o_util.list_to_dict(frames[0]['columns'], 'label')
assert 'AGE' in columns_dict, "Failed to find AGE in Frames/prostate.hex/columns."
assert 'bins' in columns_dict['AGE'], "Failed to find bins in Frames/prostate.hex/columns/AGE."
assert None is columns_dict['AGE']['bins'], "Failed to clear bins field." # should be cleared except for /summary
frames = self.a_node.summary(key='prostate.hex', column='AGE')['frames']
columns_dict = h2o_util.list_to_dict(frames[0]['columns'], 'label')
assert 'AGE' in columns_dict, "Failed to find AGE in Frames/prostate.hex/columns/AGE/summary."
col = columns_dict['AGE']
h2o_util.assertKeysExistAndNonNull(col, '', ['label', 'missing', 'zeros', 'pinfs', 'ninfs', 'mins',
'maxs', 'mean', 'sigma', 'type', 'data', 'precision', 'bins', 'base', 'stride', 'pctiles'])
h2o_util.assertKeysExist(col, '', ['domain', 'str_data'])
assert col['mins'][0] == 43, 'Failed to find 43 as the first min for AGE.'
assert col['maxs'][0] == 79, 'Failed to find 79 as the first max for AGE.'
assert col['mean'] == 66.03947368421052, 'Failed to find 66.03947368421052 as the mean for AGE.'
assert col['sigma'] == 6.527071269173308, 'Failed to find 6.527071269173308 as the sigma for AGE.'
assert col['type'] == 'int', 'Failed to find int as the type for AGE.'
assert col['data'][0] == 65, 'Failed to find 65 as the first data for AGE.'
assert col['precision'] == -1, 'Failed to find -1 as the precision for AGE.'
assert col['bins'][0] == 1, 'Failed to find 1 as the first bin for AGE.'
assert col['base'] == 43, 'Failed to find 43 as the base for AGE.'
assert col['stride'] == 1, 'Failed to find 1 as the stride for AGE.'
assert col['pctiles'][0] == 50.5, 'Failed to find 50.5 as the first pctile for AGE.'
def find_key(pattern=None):
try:
patternObj = re.compile(pattern)
except:
raise Exception("Need legal pattern in find_key, not %s", pattern)
frames = h2o_nodes.nodes[0].frames()['frames']
frames_dict = h2o_util.list_to_dict(frames, 'key/name')
result = []
for key in frames_dict:
if patternObj.search(key):
result.append(key)
if len(result) == 0:
verboseprint("Warning: No match for %s" % pattern)
return None
if len(result) > 1:
verboseprint("Warning: multiple imported keys match the key pattern %s, Using: %s" % (pattern, result[0]))
return result[0]
DatasetSpec('airlines_binomial', '../../smalldata/airlines/allyears2k_headers.zip', 43978, 'Binomial', 'IsDepDelayed', ['IsArrDelayed', 'ArrDelay', 'DepDelay']), # TODO: more ignored?
DatasetSpec('iris_multinomial', '../../smalldata/iris/iris_wheader.csv', 150, 'Multinomial', 'class', []),
]
datasets = {} # the dataset spec
for dataset_spec in datasets_to_import:
dataset = dataset_spec.import_and_validate_dataset(a_node) # it's also stored in dataset_spec['dataset']
datasets[dataset_spec['dest_key']] = dataset_spec
################################################
# Test /Frames for prostate.csv
frames = a_node.frames(len=5)['frames']
frames_dict = h2o_util.list_to_dict(frames, 'key/name')
# TODO: remove:
if h2o.H2O.verbose:
print "frames: "
pp.pprint(frames)
if h2o.H2O.verbose:
print "frames_dict: "
pp.pprint(frames_dict)
# TODO: test len and offset (they aren't working yet)
assert 'prostate_binomial' in frames_dict, "FAIL: Failed to find " + 'prostate_binomial' + " in Frames list."
assert not frames_dict['prostate_binomial']['isText'], "FAIL: Parsed Frame is isText"
def testImportProstate(self):
cleanup(self.a_node)
prostate_tuple = self.cfg.data['prostate']
if (prostate_tuple[0] == "file"):
import_result = self.a_node.import_files(
path=os.path.abspath(prostate_tuple[1]))
else:
raise RuntimeError("Unsupported file type specified")
parse_result = self.a_node.parse(
key=import_result['keys'][0]) # TODO: handle multiple files
self.prostate_key = parse_result['frames'][0]['key']['name']
# Test /Frames for prostate.csv
frames = self.a_node.frames()['frames']
frames_dict = h2o_util.list_to_dict(frames, 'key/name')
assert 'prostate.hex' in frames_dict, "Failed to find prostate.hex in Frames list."
# Test /Frames/{key} for prostate.csv
frames = self.a_node.frames(key='prostate.hex')['frames']
frames_dict = h2o_util.list_to_dict(frames, 'key/name')
assert 'prostate.hex' in frames_dict, "Failed to find prostate.hex in Frames list."
columns_dict = h2o_util.list_to_dict(frames[0]['columns'], 'label')
assert 'CAPSULE' in columns_dict, "Failed to find CAPSULE in Frames/prostate.hex."
assert 'AGE' in columns_dict, "Failed to find AGE in Frames/prostate.hex/columns."
assert 'bins' in columns_dict[
'AGE'], "Failed to find bins in Frames/prostate.hex/columns/AGE."
assert None is columns_dict['AGE'][
'bins'], "Failed to clear bins field." # should be cleared except for /summary
frames = self.a_node.columns(key='prostate.hex')['frames']
columns_dict = h2o_util.list_to_dict(frames[0]['columns'], 'label')
assert 'ID' in columns_dict, "Failed to find ID in Frames/prostate.hex/columns."
assert 'AGE' in columns_dict, "Failed to find AGE in Frames/prostate.hex/columns."
assert 'bins' in columns_dict[
'AGE'], "Failed to find bins in Frames/prostate.hex/columns/AGE."
assert None is columns_dict['AGE'][
'bins'], "Failed to clear bins field." # should be cleared except for /summary
frames = self.a_node.column(key='prostate.hex', column='AGE')['frames']
columns_dict = h2o_util.list_to_dict(frames[0]['columns'], 'label')
assert 'AGE' in columns_dict, "Failed to find AGE in Frames/prostate.hex/columns."
assert 'bins' in columns_dict[
'AGE'], "Failed to find bins in Frames/prostate.hex/columns/AGE."
assert None is columns_dict['AGE'][
'bins'], "Failed to clear bins field." # should be cleared except for /summary
frames = self.a_node.summary(key='prostate.hex',
column='AGE')['frames']
columns_dict = h2o_util.list_to_dict(frames[0]['columns'], 'label')
assert 'AGE' in columns_dict, "Failed to find AGE in Frames/prostate.hex/columns/AGE/summary."
col = columns_dict['AGE']
h2o_util.assertKeysExistAndNonNull(col, '', [
'label', 'missing', 'zeros', 'pinfs', 'ninfs', 'mins', 'maxs',
'mean', 'sigma', 'type', 'data', 'precision', 'bins', 'base',
'stride', 'pctiles'
])
h2o_util.assertKeysExist(col, '', ['domain', 'str_data'])
assert col['mins'][
0] == 43, 'Failed to find 43 as the first min for AGE.'
assert col['maxs'][
0] == 79, 'Failed to find 79 as the first max for AGE.'
assert col[
'mean'] == 66.03947368421052, 'Failed to find 66.03947368421052 as the mean for AGE.'
assert col[
'sigma'] == 6.527071269173308, 'Failed to find 6.527071269173308 as the sigma for AGE.'
assert col['type'] == 'int', 'Failed to find int as the type for AGE.'
assert col['data'][
0] == 65, 'Failed to find 65 as the first data for AGE.'
assert col[
'precision'] == -1, 'Failed to find -1 as the precision for AGE.'
assert col['bins'][
0] == 1, 'Failed to find 1 as the first bin for AGE.'
assert col['base'] == 43, 'Failed to find 43 as the base for AGE.'
assert col['stride'] == 1, 'Failed to find 1 as the stride for AGE.'
assert col['pctiles'][
0] == 50.5, 'Failed to find 50.5 as the first pctile for AGE.'
def test_parse_multi_header_single(self):
SYNDATASETS_DIR = h2o.make_syn_dir()
csvFilename = "syn_ints.csv"
csvPathname = SYNDATASETS_DIR + '/' + csvFilename
headerData = "ID,CAPSULE,AGE,RACE,DPROS,DCAPS,PSA,VOL,GLEASON,output"
# cols must be 9 to match the header above, otherwise a different bug is hit
# extra output is added, so it's 10 total
tryList = [
(57, 300, 9, 'cA', 60, 0),
# try with 1-3 data lines in the header file too
(57, 300, 9, 'cB', 60, 1),
(57, 300, 9, 'cC', 60, 2),
(57, 300, 9, 'cD', 60, 3),
]
trial = 0
for (fileNum, rowCount, colCount, hex_key, timeoutSecs,
dataRowsWithHeader) in tryList:
trial += 1
# FIX! should we add a header to them randomly???
print "Wait while", fileNum, "synthetic files are created in", SYNDATASETS_DIR
rowxcol = str(rowCount) + 'x' + str(colCount)
totalCols = colCount + 1 # 1 extra for output
totalDataRows = 0
for fileN in range(fileNum):
csvFilename = 'syn_' + str(fileN) + "_" + str(
SEED) + "_" + rowxcol + '.csv'
csvPathname = SYNDATASETS_DIR + '/' + csvFilename
rList = rand_rowData(colCount)
dataRowsDone = write_syn_dataset(csvPathname,
rowCount,
headerData=None,
rList=rList)
totalDataRows += dataRowsDone
# create the header file
# can make it pass by not doing this
if HEADER:
csvFilename = 'syn_header_' + str(
SEED) + "_" + rowxcol + '.csv'
csvPathname = SYNDATASETS_DIR + '/' + csvFilename
dataRowsDone = write_syn_dataset(csvPathname,
dataRowsWithHeader,
headerData, rList)
totalDataRows += dataRowsDone
# make sure all key names are unique, when we re-put and re-parse (h2o caching issues)
src_key = "syn_" + str(trial)
hex_key = "syn_" + str(trial) + ".hex"
# DON"T get redirected to S3! (EC2 hack in config, remember!)
# use it at the node level directly (because we gen'ed the files.
# I suppose we could force the redirect state bits in h2o.nodes[0] to False, instead?
# put them, rather than using import files, so this works if remote h2o is used
# and python creates the files locally
fileList = os.listdir(SYNDATASETS_DIR)
for f in fileList:
h2i.import_only(path=SYNDATASETS_DIR + "/" + f,
schema='put',
noPrint=True)
print f
# fix. should we have a h2o.n0 for brevity? or h2o.n. ? so we can change it around if multi-node?
# frames = h2o.nodes[0].frames()['frames']
frames = h2o.n0.frames()['frames']
frames_dict = h2o_util.list_to_dict(frames, 'key/name')
# print "frames:", dump_json(frames)
# print "frames_dict:", dump_json(frames_dict)
if HEADER:
header = h2i.find_key('syn_header')
if not header:
raise Exception(
"Didn't find syn_header* key in the import")
# use regex. the only files in the dir will be the ones we just created with *fileN* match
print "Header Key = " + header
start = time.time()
# does h2o-dev take a regex? or do we need to glob
parseResult = h2i.parse_only(
pattern='*' + rowxcol + '*',
hex_key=hex_key,
timeoutSecs=timeoutSecs,
checkHeader="1") # header_from_file=header
pA = h2o_cmd.ParseObj(parseResult,
expectedNumRows=totalDataRows,
expectedNumCols=totalCols)
print pA.numRows
print pA.numCols
print pA.parse_key
expectedLabelList = headerData.split(",")
iA = h2o_cmd.InspectObj(pA.parse_key,
expectedNumRows=totalDataRows,
expectedNumCols=totalCols,
expectedMissinglist=[],
expectedLabelList=expectedLabelList)
if DO_RF:
# put in an ignore param, that will fail unless headers were parsed correctly
if HEADER:
kwargs = {
'sample_rate': 0.75,
'max_depth': 25,
'ntrees': 1,
'ignored_cols_by_name': 'ID,CAPSULE'
}
else:
kwargs = {
'sample_rate': 0.75,
'max_depth': 25,
'ntrees': 1
}
rfv = h2o_cmd.runRF(parseResult=parseResult,
timeoutSecs=timeoutSecs,
**kwargs)
h2o.check_sandbox_for_errors()
DatasetSpec('airlines_binomial', '../../smalldata/airlines/allyears2k_headers.zip', 43978, 'Binomial', 'IsDepDelayed', ['IsArrDelayed', 'ArrDelay', 'DepDelay']), # TODO: more ignored?
DatasetSpec('iris_multinomial', '../../smalldata/iris/iris_wheader.csv', 150, 'Multinomial', 'class', []),
]
datasets = {} # the dataset spec
for dataset_spec in datasets_to_import:
dataset = dataset_spec.import_and_validate_dataset(a_node) # it's also stored in dataset_spec['dataset']
datasets[dataset_spec['dest_key']] = dataset_spec
################################################
# Test /Frames for prostate.csv
frames = a_node.frames(row_count=5)['frames']
frames_dict = h2o_util.list_to_dict(frames, 'key/name')
# TODO: remove:
if h2o.H2O.verbose:
print "frames: "
pp.pprint(frames)
if h2o.H2O.verbose:
print "frames_dict: "
pp.pprint(frames_dict)
assert 'prostate_binomial' in frames_dict, "FAIL: Failed to find " + 'prostate_binomial' + " in Frames list."
assert not frames_dict['prostate_binomial']['is_text'], "FAIL: Parsed Frame is is_text"
# Test /Frames/{key} for prostate.csv
def test_parse_multi_header_single(self):
SYNDATASETS_DIR = h2o.make_syn_dir()
csvFilename = "syn_ints.csv"
csvPathname = SYNDATASETS_DIR + '/' + csvFilename
headerData = "ID,CAPSULE,AGE,RACE,DPROS,DCAPS,PSA,VOL,GLEASON,output"
# cols must be 9 to match the header above, otherwise a different bug is hit
# extra output is added, so it's 10 total
tryList = [
(57, 300, 9, 'cA', 60, 0),
# try with 1-3 data lines in the header file too
(57, 300, 9, 'cB', 60, 1),
(57, 300, 9, 'cC', 60, 2),
(57, 300, 9, 'cD', 60, 3),
]
trial = 0
for (fileNum, rowCount, colCount, hex_key, timeoutSecs, dataRowsWithHeader) in tryList:
trial += 1
# FIX! should we add a header to them randomly???
print "Wait while", fileNum, "synthetic files are created in", SYNDATASETS_DIR
rowxcol = str(rowCount) + 'x' + str(colCount)
totalCols = colCount + 1 # 1 extra for output
totalDataRows = 0
for fileN in range(fileNum):
csvFilename = 'syn_' + str(fileN) + "_" + str(SEED) + "_" + rowxcol + '.csv'
csvPathname = SYNDATASETS_DIR + '/' + csvFilename
rList = rand_rowData(colCount)
dataRowsDone = write_syn_dataset(csvPathname, rowCount, headerData=None, rList=rList)
totalDataRows += dataRowsDone
# create the header file
# can make it pass by not doing this
if HEADER:
csvFilename = 'syn_header_' + str(SEED) + "_" + rowxcol + '.csv'
csvPathname = SYNDATASETS_DIR + '/' + csvFilename
dataRowsDone = write_syn_dataset(csvPathname, dataRowsWithHeader, headerData, rList)
totalDataRows += dataRowsDone
# make sure all key names are unique, when we re-put and re-parse (h2o caching issues)
src_key = "syn_" + str(trial)
hex_key = "syn_" + str(trial) + ".hex"
# DON"T get redirected to S3! (EC2 hack in config, remember!)
# use it at the node level directly (because we gen'ed the files.
# I suppose we could force the redirect state bits in h2o.nodes[0] to False, instead?
# put them, rather than using import files, so this works if remote h2o is used
# and python creates the files locally
fileList = os.listdir(SYNDATASETS_DIR)
for f in fileList:
h2i.import_only(path=SYNDATASETS_DIR + "/" + f, schema='put', noPrint=True)
print f
# fix. should we have a h2o.n0 for brevity? or h2o.n. ? so we can change it around if multi-node?
# frames = h2o.nodes[0].frames()['frames']
frames = h2o.n0.frames()['frames']
frames_dict = h2o_util.list_to_dict(frames, 'key/name')
# print "frames:", dump_json(frames)
# print "frames_dict:", dump_json(frames_dict)
if HEADER:
header = h2i.find_key('syn_header')
if not header:
raise Exception("Didn't find syn_header* key in the import")
# use regex. the only files in the dir will be the ones we just created with *fileN* match
print "Header Key = " + header
start = time.time()
# does h2o-dev take a regex? or do we need to glob
parseResult = h2i.parse_only(pattern='*'+rowxcol+'*',
hex_key=hex_key, timeoutSecs=timeoutSecs, checkHeader="1") # header_from_file=header
pA = h2o_cmd.ParseObj(parseResult, expectedNumRows=totalDataRows, expectedNumCols=totalCols)
print pA.numRows
print pA.numCols
print pA.parse_key
expectedLabelList = headerData.split(",")
iA = h2o_cmd.InspectObj(pA.parse_key, expectedNumRows=totalDataRows, expectedNumCols=totalCols,
expectedMissinglist=[], expectedLabelList=expectedLabelList)
if DO_RF:
# put in an ignore param, that will fail unless headers were parsed correctly
if HEADER:
kwargs = {'sample_rate': 0.75, 'max_depth': 25, 'ntrees': 1, 'ignored_cols_by_name': 'ID,CAPSULE'}
else:
kwargs = {'sample_rate': 0.75, 'max_depth': 25, 'ntrees': 1}
rfv = h2o_cmd.runRF(parseResult=parseResult, timeoutSecs=timeoutSecs, **kwargs)
h2o.check_sandbox_for_errors()
def test_simple2(self):
# h2o-dev doesn't take ../.. type paths? make find_file return absolute path
# csvPathname = find_file("bigdata/laptop/poker-hand-testing.data")
csvPathname = find_file("smalldata/logreg/prostate.csv")
import_result = h2o.n0.import_files(path=csvPathname)
# print dump_json(import_result)
k = import_result['keys'][0]
frames_result = h2o.n0.frames(key=k)
frame = frames_result['frames'][0]
rows = frame['rows']
columns = frame['columns']
for c in columns:
label = c['label']
missing = c['missing_count']
stype = c['type']
domain = c['domain']
# print dump_json(frame)
# let's see what ray's util does
frames = h2o.n0.frames()['frames']
frames_dict = h2o_util.list_to_dict(frames, 'key/name')
# print "frames:", dump_json(frames)
# print "frames_dict:", dump_json(frames_dict)
for k, v in frames_dict.items():
print "frames_dict key:", k
# interesting. we can do dictionary comprehensions
# { k:v for k,v in my_dict.items() if 'Peter' in k }
# how do you parse multiple files
parse_result = h2o.n0.parse(
key=k, intermediateResults=DO_INTERMEDIATE_RESULTS)
frame = parse_result['frames'][0]
hex_key = frame['key']['name']
colCount = 9
rowCount = 380
# colCount = 11
# rowCount = 1000000
start = time.time()
inspect = h2o_cmd.runInspect(None, hex_key)
print "Inspect:", hex_key, "took", time.time() - start, "seconds"
numCols = len(inspect['frames'][0]['columns'])
numRows = inspect['frames'][0]['rows']
print "\n" + csvPathname, \
" rows:", "{:,}".format(numRows), \
" len(columns):", "{:,}".format(numCols)
# should match # of cols in header or ??
self.assertEqual(
numCols, colCount,
"parse created result with the wrong number of cols %s %s" %
(numCols, colCount))
self.assertEqual(numRows, rowCount,
"parse created result with the wrong number of rows (header shouldn't count) %s %s" % \
(numRows, rowCount))
verboseprint(hex_key, ":", dump_json(parse_result))