def pack_kv(e):
if isinstance(e, tuple):
k, v = e
else:
k = b''
v = e
return struct.pack("I", len(k)) + str_to_bytes(k) \
+ struct.pack("I", len(v)) + str_to_bytes(v)
def pack_kv(e):
if isinstance(e, tuple):
k, v = e
else:
k = b''
v = e
return struct.pack("I", len(k)) + str_to_bytes(k) \
+ struct.pack("I", len(v)) + str_to_bytes(v)
def disk_sort(worker, input, filename, sort_buffer_size='10%'):
from os.path import getsize
from disco.comm import open_local
from disco.fileutils import AtomicFile
from disco.worker.task_io import re_reader
if worker:
worker.send('MSG', "Downloading {0}".format(filename))
out_fd = AtomicFile(filename)
for key, value in input:
if not isinstance(key, bytes):
raise ValueError("Keys must be bytes for external sort", key)
if b'\xff' in key or b'\x00' in key:
raise ValueError("Cannot sort key with 0xFF or 0x00 bytes", key)
else:
# value pickled using protocol 0 will always be printable ASCII
out_fd.write(key + b'\xff')
out_fd.write(encode(pickle_dumps(value, 0)) + b'\x00')
out_fd.close()
if worker:
worker.send('MSG', "Downloaded {0:s} OK".format(format_size(getsize(filename))))
worker.send('MSG', "Sorting {0}...".format(filename))
unix_sort(filename, sort_buffer_size=sort_buffer_size)
if worker:
worker.send('MSG', ("Finished sorting"))
fd = open_local(filename)
for k, v in sort_reader(fd, fd.url):
yield k, bytes_to_str(decode(str_to_bytes(pickle_loads(v))))
def disk_sort(worker, input, filename, sort_buffer_size='10%'):
from os.path import getsize
from disco.comm import open_local
from disco.fileutils import AtomicFile
from disco.worker.task_io import re_reader
if worker:
worker.send('MSG', "Downloading {0}".format(filename))
out_fd = AtomicFile(filename)
for key, value in input:
if not isinstance(key, bytes):
raise ValueError("Keys must be bytes for external sort", key)
if b'\xff' in key or b'\x00' in key:
raise ValueError("Cannot sort key with 0xFF or 0x00 bytes", key)
else:
# value pickled using protocol 0 will always be printable ASCII
out_fd.write(key + b'\xff')
out_fd.write(encode(pickle_dumps(value, 0)) + b'\x00')
out_fd.close()
if worker:
worker.send(
'MSG',
"Downloaded {0:s} OK".format(format_size(getsize(filename))))
worker.send('MSG', "Sorting {0}...".format(filename))
unix_sort(filename, sort_buffer_size=sort_buffer_size)
if worker:
worker.send('MSG', ("Finished sorting"))
fd = open_local(filename)
for k, v in sort_reader(fd, fd.url):
yield k, bytes_to_str(decode(str_to_bytes(pickle_loads(v))))
def test_save_map(self):
input = range(10)
self.job = SaveMapJob().run(input=self.test_server.urls(input))
results = sorted(self.results(self.job))
self.tag = self.disco.results(self.job.name)[1][0]
self.assertEquals(len(list(self.ddfs.blobs(self.tag))), len(input))
self.assertEquals(results, [(str_to_bytes(str(e)+'!'), '') for e in input])
def test_save_map(self):
input = range(10)
self.job = SaveMapJob().run(input=self.test_server.urls(input))
results = sorted(self.results(self.job))
self.tag = self.disco.results(self.job.name)[1][0]
self.assertEquals(len(list(self.ddfs.blobs(self.tag))), len(input))
self.assertEquals(results,
[(str_to_bytes(str(e) + '!'), '') for e in input])
def test_save(self):
ducks = ['dewey', 'huey', 'louie']
a, b = SaveJob1(), SaveJob2()
self.job = JobChain({a: self.test_server.urls(ducks),
b: a}).wait()
self.tag = self.disco.results(b)[1][0]
self.assertAllEqual(sorted(self.results(b)),
[(str_to_bytes('{0}!?!?'.format(d)), '') for d in ducks])
def test_save(self):
ducks = ['dewey', 'huey', 'louie']
a, b = SaveJob1(), SaveJob2()
self.job = JobChain({a: self.test_server.urls(ducks), b: a}).wait()
self.tag = self.disco.results(b)[1][0]
self.assertAllEqual(sorted(self.results(b)),
[(str_to_bytes('{0}!?!?'.format(d)), '')
for d in ducks])
def test_extreduce(self):
self.job = ExternalJob().run(input=self.test_server.urls(self.inputs),
map=lambda e, params: [('', e)],
reduce=external([self.binary]))
ans = str_to_bytes(
str(
sum(
map(ord, ''.join('test_{0}\n'.format(i)
for i in self.inputs)))))
self.assertEquals([(ans, ans)] * 10, list(self.results(self.job)))
def prepare(params, mode):
global proc
# op -> worker
# find required files
path = os.path.join('ext.{0}'.format(mode), 'op')
os.chmod(path, stat.S_IEXEC)
proc = Popen([path, mode], stdin=PIPE, stdout=PIPE, stderr=PIPE)
register_poll()
if params and isinstance(params, dict):
proc.stdin.write(str_to_bytes(encode_netstring_fd(params)))
else:
proc.stdin.write('0\n')
return globals()[mode]
def prepare(params, mode):
global proc
# op -> worker
# find required files
path = os.path.join('ext.{0}'.format(mode), 'op')
os.chmod(path, stat.S_IEXEC)
proc = Popen([path, mode], stdin=PIPE, stdout=PIPE, stderr=PIPE)
register_poll()
if params and isinstance(params, dict):
proc.stdin.write(str_to_bytes(encode_netstring_fd(params)))
else:
proc.stdin.write('0\n')
return globals()[mode]
def test_save_map(self):
input = range(10)
self.job = SaveMapJob().run(input=self.test_server.urls(input))
results = sorted(self.results(self.job))
self.tag = self.disco.results(self.job.name)[1][0]
# Previously, each map would save one blob into DDFS. Now,
# the pipeline termination does it, using the output of the
# shuffle stage. So now, the number of blobs in the tag
# depends on the grouping used for shuffle, and also the
# number of nodes used. Hence, we cannot anymore assert on
# the number of blobs in the tag.
# self.assertEquals(len(list(self.ddfs.blobs(self.tag))), len(input))
self.assertEquals(results, [(str_to_bytes(str(e)+'!'), '') for e in input])
def test_save_map(self):
input = range(10)
self.job = SaveMapJob().run(input=self.test_server.urls(input))
results = sorted(self.results(self.job))
self.tag = self.disco.results(self.job.name)[1][0]
# Previously, each map would save one blob into DDFS. Now,
# the pipeline termination does it, using the output of the
# shuffle stage. So now, the number of blobs in the tag
# depends on the grouping used for shuffle, and also the
# number of nodes used. Hence, we cannot anymore assert on
# the number of blobs in the tag.
# self.assertEquals(len(list(self.ddfs.blobs(self.tag))), len(input))
self.assertEquals(results,
[(str_to_bytes(str(e) + '!'), '') for e in input])
def read(interface, state, label, inp):
from disco import util
for e in inp:
scheme, netloc, _ = util.urlsplit(e)
fileName, joinColumn = str(netloc).split('?')
File = open(PREFIX + fileName, 'r')
col = int(joinColumn)
reader = csv.reader(File)
firstRow = True
for row in reader:
if firstRow:
tableName = row[0]
firstRow = False
else:
fullName = tableName + '?' + str(col)
Hash = int(
hashlib.md5(str_to_bytes(row[col])).hexdigest(), 16) % 160
interface.output(Hash).add(fullName, row)
def read(interface, state, label, inp):
from disco import util
for e in inp:
scheme, netloc, _ = util.urlsplit(e)
fileName, joinColumn = str(netloc).split("?")
File = open(PREFIX + fileName, "r")
col = int(joinColumn)
reader = csv.reader(File)
firstRow = True
for row in reader:
if firstRow:
tableName = row[0]
firstRow = False
else:
fullName = tableName + "?" + str(col)
Hash = int(hashlib.md5(str_to_bytes(row[col])).hexdigest(), 16) % 160
interface.output(Hash).add(fullName, row)
def runTest(self):
self.job = RawJob().run(input=self.test_server.urls(self.input))
self.assertEqual(sorted(self.results(self.job)),
sorted((str_to_bytes(i), '') for i in self.input))
def send_data(self, data):
self.send_response(OK)
self.send_header('Content-length', len(data or []))
self.end_headers()
self.wfile.write(str_to_bytes(data))
def ansi_text(self, text, bgcolor=WHITE, fgcolor=BLACK):
return self.background(bgcolor) + self.foreground(
fgcolor) + str_to_bytes(text)
def runTest(self):
self.job = RawJob().run(input=self.test_server.urls(self.input))
self.assertEqual(sorted(self.results(self.job)),
sorted((str_to_bytes(i), '') for i in self.input))
def map(e, params):
for i in range(10):
put('{0}-{1}'.format(e, i),
str_to_bytes('val:{0}-{1}'.format(e, i)))
return []
def map(e, params):
k = bytes_to_str(e)
v = str_to_bytes('value:{0}'.format(k))
put(k, v)
yield k, v
def ansi_text(self, text, bgcolor=WHITE, fgcolor=BLACK):
return self.background(bgcolor) + self.foreground(fgcolor) + str_to_bytes(text)
def Map(interface, state, label, inp):
out = interface.output(0)
for i in inp:
for k, v in shuffled((base64.encodestring(str_to_bytes(c)), b'')
for c in bytes_to_str(str_to_bytes(i) * 10)):
out.add(k, v)
def checkAnswers(self, job, input):
self.assertEquals(sorted(self.results(job)),
sorted((str_to_bytes(str(i)), '') for i in input))
def Map(interface, state, label, inp):
out = interface.output(0)
for i in inp:
for k, v in shuffled((base64.encodestring(str_to_bytes(c)), b'') for c in bytes_to_str(str_to_bytes(i) * 10)):
out.add(k, v)
def runTest(self):
self.job = SortJob().run(input=self.test_server.urls([''] * 100))
result = [i for i in self.results(self.job)]
self.assertResults(self.job, sorted((str_to_bytes(c), 1000) for c in alphanum))
def map(string, params):
return shuffled((base64.encodestring(str_to_bytes(c)), b'')
for c in bytes_to_str(string * 10))
def Map(interface, state, label, inp):
out = interface.output(0)
for i in inp:
out.add(str_to_bytes(i), u'\x00\x00')
def Reduce(interface, state, label, inp):
out = interface.output(0)
for k, vs in kvgroup(inp):
out.add(str_to_bytes(k), 0)
def add(self, k, v):
k, v = str(k), str(v)
self.stream.write(
str_to_bytes("%d %s %d %s\n" % (len(k), k, len(v), v)))
def read(self):
if self.isopen:
return BytesIO(str_to_bytes(self.source)).read
return open(self.source, 'rb').read
def map(e, params):
for i in range(10):
put('{0}-{1}'.format(e, i), str_to_bytes('val:{0}-{1}'.format(e, i)))
return []
def send_data(self, data):
self.send_response(OK)
self.send_header('Content-length', len(data or []))
self.end_headers()
self.wfile.write(str_to_bytes(data))
def reduce(iter, params):
for k, v in iter:
assert v == get(k)
x = 'reduce:{0}'.format(this_partition())
put(x, str_to_bytes('value:{0}'.format(x)))
yield 'all', 'ok'
def reduce(iter, params):
for k, v in iter:
assert v == get(k)
x = 'reduce:{0}'.format(this_partition())
put(x, str_to_bytes('value:{0}'.format(x)))
yield 'all', 'ok'
def Map(interface, state, label, inp):
out = interface.output(0)
for i in inp:
out.add(str_to_bytes(i), u'\x00\x00')
def map(e, params):
k = bytes_to_str(e)
v = str_to_bytes('value:{0}'.format(k))
put(k, v)
yield k, v
def Reduce(interface, state, label, inp):
out = interface.output(0)
for k, vs in kvgroup(inp):
out.add(str_to_bytes(k), 0)
def test_extreduce(self):
self.job = ExternalJob().run(input=self.test_server.urls(self.inputs),
map=lambda e, params: [('', e)],
reduce=external([self.binary]))
ans = str_to_bytes(str(sum(map(ord, ''.join('test_{0}\n'.format(i) for i in self.inputs)))))
self.assertEquals([(ans, ans)] * 10, list(self.results(self.job)))
def runTest(self):
self.job = SortJob().run(input=self.test_server.urls([''] * 100))
result = [i for i in self.results(self.job)]
self.assertResults(self.job,
sorted((str_to_bytes(c), 1000) for c in alphanum))
def add(self, k, v):
k, v = str(k), str(v)
self.stream.write(str_to_bytes("%d %s %d %s\n" % (len(k), k, len(v), v)))
def map(string, params):
return shuffled((base64.encodestring(str_to_bytes(c)), b'') for c in bytes_to_str(string * 10))
def read(self):
if self.isopen:
return BytesIO(str_to_bytes(self.source)).read
return open(self.source, 'rb').read
def checkAnswers(self, job, input):
self.assertEquals(sorted(self.results(job)),
sorted((str_to_bytes(str(i)), '') for i in input))
def getHash(line):
return int(hashlib.md5(str_to_bytes(line)).hexdigest(), 16) % 128
def getHash(line):
return int(hashlib.md5(str_to_bytes(line)).hexdigest(), 16) % 128
