X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/eb0f567ea3b6497311f5239ca54a4d059ba3485e..19e85e39155c45975c1508afaebf6c9cd17c3c63:/sdk/python/arvados.py diff --git a/sdk/python/arvados.py b/sdk/python/arvados.py index 336cec9907..5c547fa421 100644 --- a/sdk/python/arvados.py +++ b/sdk/python/arvados.py @@ -12,6 +12,8 @@ import re import hashlib import string import bz2 +import zlib +import fcntl from apiclient import errors from apiclient.discovery import build @@ -62,6 +64,7 @@ def current_task(): t = service.job_tasks().get(uuid=os.environ['TASK_UUID']).execute() t = UserDict.UserDict(t) t.set_output = types.MethodType(task_set_output, t) + t.tmpdir = os.environ['TASK_WORK'] _current_task = t return t @@ -71,6 +74,8 @@ def current_job(): if _current_job: return _current_job t = service.jobs().get(uuid=os.environ['JOB_UUID']).execute() + t = UserDict.UserDict(t) + t.tmpdir = os.environ['JOB_WORK'] _current_job = t return t @@ -93,7 +98,7 @@ class job_setup: task_input = f.as_manifest() new_task_attrs = { 'job_uuid': current_job()['uuid'], - 'created_by_job_task': current_task()['uuid'], + 'created_by_job_task_uuid': current_task()['uuid'], 'sequence': if_sequence + 1, 'parameters': { 'input':task_input @@ -106,6 +111,299 @@ class job_setup: ).execute() exit(0) + @staticmethod + def one_task_per_input_stream(if_sequence=0, and_end_task=True): + if if_sequence != current_task()['sequence']: + return + job_input = current_job()['script_parameters']['input'] + cr = CollectionReader(job_input) + for s in cr.all_streams(): + task_input = s.tokens() + new_task_attrs = { + 'job_uuid': current_job()['uuid'], + 'created_by_job_task_uuid': current_task()['uuid'], + 'sequence': if_sequence + 1, + 'parameters': { + 'input':task_input + } + } + service.job_tasks().create(job_task=json.dumps(new_task_attrs)).execute() + if and_end_task: + service.job_tasks().update(uuid=current_task()['uuid'], + job_task=json.dumps({'success':True}) + ).execute() + exit(0) + +class util: + @staticmethod + def run_command(execargs, **kwargs): + if 'stdin' not in kwargs: + kwargs['stdin'] = subprocess.PIPE + if 'stdout' not in kwargs: + kwargs['stdout'] = subprocess.PIPE + if 'stderr' not in kwargs: + kwargs['stderr'] = subprocess.PIPE + p = subprocess.Popen(execargs, close_fds=True, shell=False, + **kwargs) + stdoutdata, stderrdata = p.communicate(None) + if p.returncode != 0: + raise Exception("run_command %s exit %d:\n%s" % + (execargs, p.returncode, stderrdata)) + return stdoutdata, stderrdata + + @staticmethod + def git_checkout(url, version, path): + if not re.search('^/', path): + path = os.path.join(current_job().tmpdir, path) + if not os.path.exists(path): + util.run_command(["git", "clone", url, path], + cwd=os.path.dirname(path)) + util.run_command(["git", "checkout", version], + cwd=path) + return path + + @staticmethod + def tar_extractor(path, decompress_flag): + return subprocess.Popen(["tar", + "-C", path, + ("-x%sf" % decompress_flag), + "-"], + stdout=None, + stdin=subprocess.PIPE, stderr=sys.stderr, + shell=False, close_fds=True) + + @staticmethod + def tarball_extract(tarball, path): + """Retrieve a tarball from Keep and extract it to a local + directory. Return the absolute path where the tarball was + extracted. If the top level of the tarball contained just one + file or directory, return the absolute path of that single + item. + + tarball -- collection locator + path -- where to extract the tarball: absolute, or relative to job tmp + """ + if not re.search('^/', path): + path = os.path.join(current_job().tmpdir, path) + lockfile = open(path + '.lock', 'w') + fcntl.flock(lockfile, fcntl.LOCK_EX) + try: + os.stat(path) + except OSError: + os.mkdir(path) + already_have_it = False + try: + if os.readlink(os.path.join(path, '.locator')) == tarball: + already_have_it = True + except OSError: + pass + if not already_have_it: + + # emulate "rm -f" (i.e., if the file does not exist, we win) + try: + os.unlink(os.path.join(path, '.locator')) + except OSError: + if os.path.exists(os.path.join(path, '.locator')): + os.unlink(os.path.join(path, '.locator')) + + for f in CollectionReader(tarball).all_files(): + if re.search('\.(tbz|tar.bz2)$', f.name()): + p = tar_extractor(path, 'j') + elif re.search('\.(tgz|tar.gz)$', f.name()): + p = tar_extractor(path, 'z') + elif re.search('\.tar$', f.name()): + p = tar_extractor(path, '') + else: + raise Exception("tarball_extract cannot handle filename %s" + % f.name()) + while True: + buf = f.read(2**20) + if len(buf) == 0: + break + p.stdin.write(buf) + p.stdin.close() + p.wait() + if p.returncode != 0: + lockfile.close() + raise Exception("tar exited %d" % p.returncode) + os.symlink(tarball, os.path.join(path, '.locator')) + tld_extracts = filter(lambda f: f != '.locator', os.listdir(path)) + lockfile.close() + if len(tld_extracts) == 1: + return os.path.join(path, tld_extracts[0]) + return path + + @staticmethod + def zipball_extract(zipball, path): + """Retrieve a zip archive from Keep and extract it to a local + directory. Return the absolute path where the archive was + extracted. If the top level of the archive contained just one + file or directory, return the absolute path of that single + item. + + zipball -- collection locator + path -- where to extract the archive: absolute, or relative to job tmp + """ + if not re.search('^/', path): + path = os.path.join(current_job().tmpdir, path) + lockfile = open(path + '.lock', 'w') + fcntl.flock(lockfile, fcntl.LOCK_EX) + try: + os.stat(path) + except OSError: + os.mkdir(path) + already_have_it = False + try: + if os.readlink(os.path.join(path, '.locator')) == zipball: + already_have_it = True + except OSError: + pass + if not already_have_it: + + # emulate "rm -f" (i.e., if the file does not exist, we win) + try: + os.unlink(os.path.join(path, '.locator')) + except OSError: + if os.path.exists(os.path.join(path, '.locator')): + os.unlink(os.path.join(path, '.locator')) + + for f in CollectionReader(zipball).all_files(): + if not re.search('\.zip$', f.name()): + raise Exception("zipball_extract cannot handle filename %s" + % f.name()) + zip_filename = os.path.join(path, os.path.basename(f.name())) + zip_file = open(zip_filename, 'wb') + while True: + buf = f.read(2**20) + if len(buf) == 0: + break + zip_file.write(buf) + zip_file.close() + + p = subprocess.Popen(["unzip", + "-q", "-o", + "-d", path, + zip_filename], + stdout=None, + stdin=None, stderr=sys.stderr, + shell=False, close_fds=True) + p.wait() + if p.returncode != 0: + lockfile.close() + raise Exception("unzip exited %d" % p.returncode) + os.unlink(zip_filename) + os.symlink(zipball, os.path.join(path, '.locator')) + tld_extracts = filter(lambda f: f != '.locator', os.listdir(path)) + lockfile.close() + if len(tld_extracts) == 1: + return os.path.join(path, tld_extracts[0]) + return path + + @staticmethod + def collection_extract(collection, path, files=[], decompress=True): + """Retrieve a collection from Keep and extract it to a local + directory. Return the absolute path where the collection was + extracted. + + collection -- collection locator + path -- where to extract: absolute, or relative to job tmp + """ + if not re.search('^/', path): + path = os.path.join(current_job().tmpdir, path) + lockfile = open(path + '.lock', 'w') + fcntl.flock(lockfile, fcntl.LOCK_EX) + try: + os.stat(path) + except OSError: + os.mkdir(path) + already_have_it = False + try: + if os.readlink(os.path.join(path, '.locator')) == collection: + already_have_it = True + except OSError: + pass + + # emulate "rm -f" (i.e., if the file does not exist, we win) + try: + os.unlink(os.path.join(path, '.locator')) + except OSError: + if os.path.exists(os.path.join(path, '.locator')): + os.unlink(os.path.join(path, '.locator')) + + files_got = [] + for f in CollectionReader(collection).all_files(): + if (files == [] or + ((f.name() not in files_got) and + (f.name() in files or + (decompress and f.decompressed_name() in files)))): + outname = f.decompressed_name() if decompress else f.name() + files_got += [outname] + if os.path.exists(os.path.join(path, outname)): + continue + util.mkdir_dash_p(os.path.dirname(os.path.join(path, outname))) + outfile = open(os.path.join(path, outname), 'wb') + for buf in (f.readall_decompressed() if decompress + else f.readall()): + outfile.write(buf) + outfile.close() + if len(files_got) < len(files): + raise Exception("Wanted files %s but only got %s from %s" % (files, files_got, map(lambda z: z.name(), list(CollectionReader(collection).all_files())))) + os.symlink(collection, os.path.join(path, '.locator')) + + lockfile.close() + return path + + @staticmethod + def mkdir_dash_p(path): + if not os.path.exists(path): + mkdir_dash_p(os.dirname(path)) + try: + os.mkdir(path) + except OSError: + if not os.path.exists(path): + os.mkdir(path) + + @staticmethod + def stream_extract(stream, path, files=[], decompress=True): + """Retrieve a stream from Keep and extract it to a local + directory. Return the absolute path where the stream was + extracted. + + stream -- StreamReader object + path -- where to extract: absolute, or relative to job tmp + """ + if not re.search('^/', path): + path = os.path.join(current_job().tmpdir, path) + lockfile = open(path + '.lock', 'w') + fcntl.flock(lockfile, fcntl.LOCK_EX) + try: + os.stat(path) + except OSError: + os.mkdir(path) + + files_got = [] + for f in stream.all_files(): + if (files == [] or + ((f.name() not in files_got) and + (f.name() in files or + (decompress and f.decompressed_name() in files)))): + outname = f.decompressed_name() if decompress else f.name() + files_got += [outname] + if os.path.exists(os.path.join(path, outname)): + continue + mkdir_dash_p(os.path.dirname(os.path.join(path, outname))) + outfile = open(os.path.join(path, outname), 'wb') + for buf in (f.readall_decompressed() if decompress + else f.readall()): + outfile.write(buf) + outfile.close() + if len(files_got) < len(files): + raise Exception("Wanted files %s but only got %s from %s" % + (files, files_got, map(lambda z: z.name(), + list(stream.all_files())))) + lockfile.close() + return path + class DataReader: def __init__(self, data_locator): self.data_locator = data_locator @@ -138,6 +436,8 @@ class StreamFileReader: self._filepos = 0 def name(self): return self._name + def decompressed_name(self): + return re.sub('\.(bz2|gz)$', '', self._name) def size(self): return self._size def stream_name(self): @@ -147,7 +447,7 @@ class StreamFileReader: data = self._stream.read(min(size, self._size - self._filepos)) self._filepos += len(data) return data - def readall(self, size, **kwargs): + def readall(self, size=2**20, **kwargs): while True: data = self.read(size, **kwargs) if data == '': @@ -159,12 +459,26 @@ class StreamFileReader: data = decompressor.decompress(chunk) if data and data != '': yield data - def readlines(self, decompress=True): + def gunzip(self, size): + decompressor = zlib.decompressobj(16+zlib.MAX_WBITS) + for chunk in self.readall(size): + data = decompressor.decompress(decompressor.unconsumed_tail + chunk) + if data and data != '': + yield data + def readall_decompressed(self, size=2**20): self._stream.seek(self._pos + self._filepos) - if decompress and re.search('\.bz2$', self._name): - datasource = self.bunzip2(2**10) + if re.search('\.bz2$', self._name): + return self.bunzip2(size) + elif re.search('\.gz$', self._name): + return self.gunzip(size) + else: + return self.readall(size) + def readlines(self, decompress=True): + if decompress: + datasource = self.readall_decompressed() else: - datasource = self.readall(2**10) + self._stream.seek(self._pos + self._filepos) + datasource = self.readall() data = '' for newdata in datasource: data += newdata @@ -207,6 +521,9 @@ class StreamReader: self.files += [[int(pos), int(size), name]] else: raise Exception("Invalid manifest format") + + def tokens(self): + return self._tokens def tokens_for_range(self, range_start, range_size): resp = [self._stream_name] return_all_tokens = False @@ -326,11 +643,15 @@ class CollectionReader: for s in self.all_streams(): for f in s.all_files(): yield f + def manifest_text(self): + self._populate() + return self._manifest_text class CollectionWriter: KEEP_BLOCK_SIZE = 2**26 def __init__(self): - self._data_buffer = '' + self._data_buffer = [] + self._data_buffer_len = 0 self._current_stream_files = [] self._current_stream_length = 0 self._current_stream_locators = [] @@ -343,14 +664,17 @@ class CollectionWriter: def __exit__(self): self.finish() def write(self, newdata): - self._data_buffer += newdata + self._data_buffer += [newdata] + self._data_buffer_len += len(newdata) self._current_stream_length += len(newdata) - while len(self._data_buffer) >= self.KEEP_BLOCK_SIZE: + while self._data_buffer_len >= self.KEEP_BLOCK_SIZE: self.flush_data() def flush_data(self): - if self._data_buffer != '': - self._current_stream_locators += [Keep.put(self._data_buffer[0:self.KEEP_BLOCK_SIZE])] - self._data_buffer = self._data_buffer[self.KEEP_BLOCK_SIZE:] + data_buffer = ''.join(self._data_buffer) + if data_buffer != '': + self._current_stream_locators += [Keep.put(data_buffer[0:self.KEEP_BLOCK_SIZE])] + self._data_buffer = [data_buffer[self.KEEP_BLOCK_SIZE:]] + self._data_buffer_len = len(self._data_buffer[0]) def start_new_file(self, newfilename=None): self.finish_current_file() self.set_current_file_name(newfilename) @@ -369,7 +693,7 @@ class CollectionWriter: self._current_stream_length - self._current_file_pos, self._current_file_name]] self._current_file_pos = self._current_stream_length - def start_new_stream(self, newstreamname=None): + def start_new_stream(self, newstreamname='.'): self.finish_current_stream() self.set_current_stream_name(newstreamname) def set_current_stream_name(self, newstreamname): @@ -401,6 +725,8 @@ class CollectionWriter: self.finish_current_stream() manifest = '' for stream in self._finished_streams: + if not re.search(r'^\.(/.*)?$', stream[0]): + manifest += './' manifest += stream[0] if len(stream[1]) == 0: manifest += " d41d8cd98f00b204e9800998ecf8427e+0"