def _recover(self, mr_spec, shard_number, shard_attempt):
next_seg_index = self._seg_index
# Save the current seg if it actually has something.
# Remember self._streaming_buffer is the pickled instance
# from the previous slice.
if self._seg_valid_length != 0:
try:
gcs_next_offset = self._streaming_buffer._get_offset_from_gcs(
) + 1
# If GCS is ahead of us, just force close.
if gcs_next_offset > self._streaming_buffer.tell():
self._streaming_buffer._force_close(gcs_next_offset)
# Otherwise flush in memory contents too.
else:
self._streaming_buffer.close()
except cloudstorage.FileClosedError:
pass
cloudstorage_api.copy2(
self._streaming_buffer.name,
self._streaming_buffer.name,
metadata={self._VALID_LENGTH: self._seg_valid_length})
next_seg_index = self._seg_index + 1
writer_spec = self.get_params(mr_spec.mapper, allow_old=False)
# Create name for the new seg.
key = self._generate_filename(writer_spec, mr_spec.name,
mr_spec.mapreduce_id, shard_number,
shard_attempt, next_seg_index)
new_writer = self._create(writer_spec, key)
new_writer._seg_index = next_seg_index
return new_writer
def _recover(self, mr_spec, shard_number, shard_attempt):
next_seg_index = self._seg_index
# Save the current seg if it actually has something.
# Remember self._streaming_buffer is the pickled instance
# from the previous slice.
if self._seg_valid_length != 0:
try:
gcs_next_offset = self._streaming_buffer._get_offset_from_gcs() + 1
# If GCS is ahead of us, just force close.
if gcs_next_offset > self._streaming_buffer.tell():
self._streaming_buffer._force_close(gcs_next_offset)
# Otherwise flush in memory contents too.
else:
self._streaming_buffer.close()
except cloudstorage.FileClosedError:
pass
cloudstorage_api.copy2(
self._streaming_buffer.name,
self._streaming_buffer.name,
metadata={self._VALID_LENGTH:
self._seg_valid_length})
next_seg_index = self._seg_index + 1
writer_spec = self.get_params(mr_spec.mapper, allow_old=False)
# Create name for the new seg.
key = self._generate_filename(
writer_spec, mr_spec.name,
mr_spec.mapreduce_id,
shard_number,
shard_attempt,
next_seg_index)
new_writer = self._create(writer_spec, key)
new_writer._seg_index = next_seg_index
return new_writer
def finalize(self, ctx, shard_state):
self._streaming_buffer.close()
if self._no_dup:
cloudstorage_api.copy2(
self._streaming_buffer.name,
self._streaming_buffer.name,
metadata={self._VALID_LENGTH: self._streaming_buffer.tell()})
# The filename user requested.
mr_spec = ctx.mapreduce_spec
writer_spec = self.get_params(mr_spec.mapper, allow_old=False)
filename = self._generate_filename(writer_spec, mr_spec.name,
mr_spec.mapreduce_id,
shard_state.shard_number)
seg_filename = self._streaming_buffer.name
prefix, last_index = seg_filename.rsplit("-", 1)
# These info is enough for any external process to combine
# all segs into the final file.
# TODO(user): Create a special input reader to combine segs.
shard_state.writer_state = {
self._SEG_PREFIX: prefix + "-",
self._LAST_SEG_INDEX: int(last_index),
"filename": filename
}
else:
shard_state.writer_state = {
"filename": self._streaming_buffer.name
}
def finalize(self, ctx, shard_state):
self._streaming_buffer.close()
if self._no_dup:
cloudstorage_api.copy2(
self._streaming_buffer.name,
self._streaming_buffer.name,
metadata={self._VALID_LENGTH: self._streaming_buffer.tell()})
# The filename user requested.
mr_spec = ctx.mapreduce_spec
writer_spec = self.get_params(mr_spec.mapper, allow_old=False)
filename = self._generate_filename(writer_spec,
mr_spec.name,
mr_spec.mapreduce_id,
shard_state.shard_number)
seg_filename = self._streaming_buffer.name
prefix, last_index = seg_filename.rsplit("-", 1)
# These info is enough for any external process to combine
# all segs into the final file.
# TODO(user): Create a special input reader to combine segs.
shard_state.writer_state = {self._SEG_PREFIX: prefix + "-",
self._LAST_SEG_INDEX: int(last_index),
"filename": filename}
else:
shard_state.writer_state = {"filename": self._streaming_buffer.name}
def testCopy2(self):
with cloudstorage.open(TESTFILE, 'w', 'text/foo',
{'x-goog-meta-foo': 'foo'}) as f:
f.write('abcde')
dst = TESTFILE + 'copy'
self.assertRaises(cloudstorage.NotFoundError, cloudstorage.stat, dst)
cloudstorage_api.copy2(TESTFILE, dst)
src_stat = cloudstorage.stat(TESTFILE)
dst_stat = cloudstorage.stat(dst)
self.assertEqual(src_stat.st_ctime, dst_stat.st_ctime)
self.assertEqual(src_stat.st_size, dst_stat.st_size)
self.assertEqual(src_stat.etag, dst_stat.etag)
self.assertEqual(src_stat.content_type, dst_stat.content_type)
self.assertEqual(src_stat.metadata, dst_stat.metadata)
with cloudstorage.open(dst) as f:
self.assertEqual('abcde', f.read())
def testCopy2(self):
with cloudstorage.open(TESTFILE, 'w',
'text/foo', {'x-goog-meta-foo': 'foo'}) as f:
f.write('abcde')
dst = TESTFILE + 'copy'
self.assertRaises(cloudstorage.NotFoundError, cloudstorage.stat, dst)
cloudstorage_api.copy2(TESTFILE, dst)
src_stat = cloudstorage.stat(TESTFILE)
dst_stat = cloudstorage.stat(dst)
self.assertEqual(src_stat.st_ctime, dst_stat.st_ctime)
self.assertEqual(src_stat.st_size, dst_stat.st_size)
self.assertEqual(src_stat.etag, dst_stat.etag)
self.assertEqual(src_stat.content_type, dst_stat.content_type)
self.assertEqual(src_stat.metadata, dst_stat.metadata)
with cloudstorage.open(dst) as f:
self.assertEqual('abcde', f.read())
def testCopy2ReplacesMetadata(self):
with cloudstorage.open(TESTFILE, 'w',
'text/foo', {'x-goog-meta-foo': 'foo'}) as f:
f.write('abcde')
src_stat = cloudstorage.stat(TESTFILE)
cloudstorage_api.copy2(TESTFILE, TESTFILE,
metadata={'x-goog-meta-foo': 'bar',
'content-type': 'text/bar'})
dst_stat = cloudstorage.stat(TESTFILE)
self.assertEqual(src_stat.st_size, dst_stat.st_size)
self.assertEqual(src_stat.etag, dst_stat.etag)
self.assertEqual(src_stat.st_ctime, dst_stat.st_ctime)
self.assertEqual('text/foo', src_stat.content_type)
self.assertEqual('text/bar', dst_stat.content_type)
self.assertEqual('foo', src_stat.metadata['x-goog-meta-foo'])
self.assertEqual('bar', dst_stat.metadata['x-goog-meta-foo'])
with cloudstorage.open(TESTFILE) as f:
self.assertEqual('abcde', f.read())
def testCopy2ReplacesMetadata(self):
with cloudstorage.open(TESTFILE, 'w', 'text/foo',
{'x-goog-meta-foo': 'foo'}) as f:
f.write('abcde')
src_stat = cloudstorage.stat(TESTFILE)
cloudstorage_api.copy2(TESTFILE,
TESTFILE,
metadata={
'x-goog-meta-foo': 'bar',
'content-type': 'text/bar'
})
dst_stat = cloudstorage.stat(TESTFILE)
self.assertEqual(src_stat.st_size, dst_stat.st_size)
self.assertEqual(src_stat.etag, dst_stat.etag)
self.assertEqual(src_stat.st_ctime, dst_stat.st_ctime)
self.assertEqual('text/foo', src_stat.content_type)
self.assertEqual('text/bar', dst_stat.content_type)
self.assertEqual('foo', src_stat.metadata['x-goog-meta-foo'])
self.assertEqual('bar', dst_stat.metadata['x-goog-meta-foo'])
with cloudstorage.open(TESTFILE) as f:
self.assertEqual('abcde', f.read())
def copy_to(path, target_path):
return cloudstorage_api.copy2(path, target_path)
def copy_to(path, target_path):
return cloudstorage_api.copy2(path, target_path)