4 import arvados.commands.ws as ws
16 import arvados.commands._util as arv_cmd
17 import arvados.collection
19 from arvados._version import __version__
21 logger = logging.getLogger('arvados.arv-run')
22 logger.setLevel(logging.INFO)
24 arvrun_parser = argparse.ArgumentParser(parents=[arv_cmd.retry_opt])
25 arvrun_parser.add_argument('--dry-run', action="store_true",
26 help="Print out the pipeline that would be submitted and exit")
27 arvrun_parser.add_argument('--local', action="store_true",
28 help="Run locally using arv-run-pipeline-instance")
29 arvrun_parser.add_argument('--docker-image', type=str,
30 help="Docker image to use, otherwise use instance default.")
31 arvrun_parser.add_argument('--ignore-rcode', action="store_true",
32 help="Commands that return non-zero return codes should not be considered failed.")
33 arvrun_parser.add_argument('--no-reuse', action="store_true",
34 help="Do not reuse past jobs.")
35 arvrun_parser.add_argument('--no-wait', action="store_true",
36 help="Do not wait and display logs after submitting command, just exit.")
37 arvrun_parser.add_argument('--project-uuid', type=str,
38 help="Parent project of the pipeline")
39 arvrun_parser.add_argument('--git-dir', type=str, default="",
40 help="Git repository passed to arv-crunch-job when using --local")
41 arvrun_parser.add_argument('--repository', type=str, default="arvados",
42 help="repository field of component, default 'arvados'")
43 arvrun_parser.add_argument('--script-version', type=str, default="master",
44 help="script_version field of component, default 'master'")
45 arvrun_parser.add_argument('--version', action='version',
46 version="%s %s" % (sys.argv[0], __version__),
47 help='Print version and exit.')
48 arvrun_parser.add_argument('args', nargs=argparse.REMAINDER)
50 class ArvFile(object):
51 def __init__(self, prefix, fn):
56 return (self.prefix+self.fn).__hash__()
58 def __eq__(self, other):
59 return (self.prefix == other.prefix) and (self.fn == other.fn)
61 class UploadFile(ArvFile):
64 # Determine if a file is in a collection, and return a tuple consisting of the
65 # portable data hash and the path relative to the root of the collection.
66 # Return None if the path isn't with an arv-mount collection or there was is error.
67 def is_in_collection(root, branch):
71 fn = os.path.join(root, ".arvados#collection")
72 if os.path.exists(fn):
73 with file(fn, 'r') as f:
75 return (c["portable_data_hash"], branch)
77 sp = os.path.split(root)
78 return is_in_collection(sp[0], os.path.join(sp[1], branch))
79 except (IOError, OSError):
82 # Determine the project to place the output of this command by searching upward
83 # for arv-mount psuedofile indicating the project. If the cwd isn't within
84 # an arv-mount project or there is an error, return current_user.
85 def determine_project(root, current_user):
89 fn = os.path.join(root, ".arvados#project")
90 if os.path.exists(fn):
91 with file(fn, 'r') as f:
93 if 'writable_by' in c and current_user in c['writable_by']:
98 sp = os.path.split(root)
99 return determine_project(sp[0], current_user)
100 except (IOError, OSError):
103 # Determine if string corresponds to a file, and if that file is part of a
104 # arv-mounted collection or only local to the machine. Returns one of
105 # ArvFile() (file already exists in a collection), UploadFile() (file needs to
106 # be uploaded to a collection), or simply returns prefix+fn (which yields the
107 # original parameter string).
108 def statfile(prefix, fn, fnPattern="$(file %s/%s)", dirPattern="$(dir %s/%s/)"):
109 absfn = os.path.abspath(fn)
112 sp = os.path.split(absfn)
113 (pdh, branch) = is_in_collection(sp[0], sp[1])
115 if stat.S_ISREG(st.st_mode):
116 return ArvFile(prefix, fnPattern % (pdh, branch))
117 elif stat.S_ISDIR(st.st_mode):
118 return ArvFile(prefix, dirPattern % (pdh, branch))
120 raise Exception("%s is not a regular file or directory" % absfn)
122 # trim leading '/' for path prefix test later
123 return UploadFile(prefix, absfn[1:])
125 if e.errno == errno.ENOENT:
132 def write_file(collection, pathprefix, fn):
133 with open(os.path.join(pathprefix, fn)) as src:
134 dst = collection.open(fn, "w")
135 r = src.read(1024*128)
138 r = src.read(1024*128)
139 dst.close(flush=False)
141 def uploadfiles(files, api, dry_run=False, num_retries=0,
143 fnPattern="$(file %s/%s)",
145 # Find the smallest path prefix that includes all the files that need to be uploaded.
146 # This starts at the root and iteratively removes common parent directory prefixes
147 # until all file paths no longer have a common parent.
156 # no parent directories left
159 # path step takes next directory
160 pathstep = sp[0] + "/"
162 # check if pathstep is common prefix for all files
163 if not c.fn.startswith(pathstep):
167 # pathstep is common parent directory for all files, so remove the prefix
169 pathprefix += pathstep
171 c.fn = c.fn[len(pathstep):]
173 logger.info("Upload local files: \"%s\"", '" "'.join([c.fn for c in files]))
176 logger.info("$(input) is %s", pathprefix.rstrip('/'))
179 files = sorted(files, key=lambda x: x.fn)
180 collection = arvados.collection.Collection(api_client=api, num_retries=num_retries)
183 localpath = os.path.join(pathprefix, f.fn)
184 if prev and localpath.startswith(prev+"/"):
185 # If this path is inside an already uploaded subdirectory,
186 # don't redundantly re-upload it.
187 # e.g. we uploaded /tmp/foo and the next file is /tmp/foo/bar
188 # skip it because it starts with "/tmp/foo/"
191 if os.path.isfile(localpath):
192 write_file(collection, pathprefix, f.fn)
193 elif os.path.isdir(localpath):
194 for root, dirs, iterfiles in os.walk(localpath):
195 root = root[len(pathprefix):]
196 for src in iterfiles:
197 write_file(collection, pathprefix, os.path.join(root, src))
199 filters=[["portable_data_hash", "=", collection.portable_data_hash()],
200 ["name", "like", name+"%"]]
202 filters.append(["owner_uuid", "=", project])
204 exists = api.collections().list(filters=filters, limit=1).execute(num_retries=num_retries)
207 item = exists["items"][0]
208 pdh = item["portable_data_hash"]
209 logger.info("Using collection %s (%s)", pdh, item["uuid"])
211 collection.save_new(name=name, owner_uuid=project, ensure_unique_name=True)
212 pdh = collection.portable_data_hash()
213 logger.info("Uploaded to %s (%s)", pdh, collection.manifest_locator())
216 c.keepref = "%s/%s" % (pdh, c.fn)
217 c.fn = fnPattern % (pdh, c.fn)
220 def main(arguments=None):
221 args = arvrun_parser.parse_args(arguments)
223 if len(args.args) == 0:
224 arvrun_parser.print_help()
227 starting_args = args.args
231 # Parse the command arguments into 'slots'.
232 # All words following '>' are output arguments and are collected into slots[0].
233 # All words following '<' are input arguments and are collected into slots[1].
234 # slots[2..] store the parameters of each command in the pipeline.
236 # e.g. arv-run foo arg1 arg2 '|' bar arg3 arg4 '<' input1 input2 input3 '>' output.txt
237 # will be parsed into:
239 # ['input1', 'input2', 'input3'],
240 # ['foo', 'arg1', 'arg2'],
241 # ['bar', 'arg3', 'arg4']]
244 if c.startswith('>'):
247 slots[reading_into].append(c[1:])
248 elif c.startswith('<'):
251 slots[reading_into].append(c[1:])
253 reading_into = len(slots)
256 slots[reading_into].append(c)
258 if slots[0] and len(slots[0]) > 1:
259 logger.error("Can only specify a single stdout file (run-command substitutions are permitted)")
263 api = arvados.api('v1')
264 if args.project_uuid:
265 project = args.project_uuid
267 project = determine_project(os.getcwd(), api.users().current().execute()["uuid"])
269 # Identify input files. Look at each parameter and test to see if there is
270 # a file by that name. This uses 'patterns' to look for within
271 # command line arguments, such as --foo=file.txt or -lfile.txt
272 patterns = [re.compile("([^=]+=)(.*)"),
273 re.compile("(-[A-Za-z])(.+)")]
274 for j, command in enumerate(slots[1:]):
275 for i, a in enumerate(command):
277 # j == 0 is stdin, j > 0 is commands
278 # always skip program executable (i == 0) in commands
280 elif a.startswith('\\'):
281 # if it starts with a \ then don't do any interpretation
284 # See if it looks like a file
285 command[i] = statfile('', a)
287 # If a file named command[i] was found, it would now be an
288 # ArvFile or UploadFile. If command[i] is a basestring, that
289 # means it doesn't correspond exactly to a file, so do some
291 if isinstance(command[i], basestring):
295 command[i] = statfile(m.group(1), m.group(2))
298 files = [c for command in slots[1:] for c in command if isinstance(c, UploadFile)]
300 uploadfiles(files, api, dry_run=args.dry_run, num_retries=args.retries, project=project)
302 for i in xrange(1, len(slots)):
303 slots[i] = [("%s%s" % (c.prefix, c.fn)) if isinstance(c, ArvFile) else c for c in slots[i]]
306 "script": "run-command",
307 "script_version": args.script_version,
308 "repository": args.repository,
309 "script_parameters": {
311 "runtime_constraints": {}
314 if args.docker_image:
315 component["runtime_constraints"]["docker_image"] = args.docker_image
318 group_parser = argparse.ArgumentParser()
319 group_parser.add_argument('-b', '--batch-size', type=int)
320 group_parser.add_argument('args', nargs=argparse.REMAINDER)
322 for s in xrange(2, len(slots)):
323 for i in xrange(0, len(slots[s])):
324 if slots[s][i] == '--':
325 inp = "input%i" % (s-2)
326 groupargs = group_parser.parse_args(slots[2][i+1:])
327 if groupargs.batch_size:
328 component["script_parameters"][inp] = {"value": {"batch":groupargs.args, "size":groupargs.batch_size}}
329 slots[s] = slots[s][0:i] + [{"foreach": inp, "command": "$(%s)" % inp}]
331 component["script_parameters"][inp] = groupargs.args
332 slots[s] = slots[s][0:i] + ["$(%s)" % inp]
333 task_foreach.append(inp)
335 if slots[s][i] == '\--':
339 component["script_parameters"]["task.stdout"] = slots[0][0]
341 task_foreach.append("stdin")
342 component["script_parameters"]["stdin"] = slots[1]
343 component["script_parameters"]["task.stdin"] = "$(stdin)"
346 component["script_parameters"]["task.foreach"] = task_foreach
348 component["script_parameters"]["command"] = slots[2:]
349 if args.ignore_rcode:
350 component["script_parameters"]["task.ignore_rcode"] = args.ignore_rcode
353 "name": "arv-run " + " | ".join([s[0] for s in slots[2:]]),
354 "description": "@" + " ".join(starting_args) + "@",
358 "state": "RunningOnClient" if args.local else "RunningOnServer"
362 print(json.dumps(pipeline, indent=4))
364 pipeline["owner_uuid"] = project
365 pi = api.pipeline_instances().create(body=pipeline, ensure_unique_name=True).execute()
366 logger.info("Running pipeline %s", pi["uuid"])
369 subprocess.call(["arv-run-pipeline-instance", "--instance", pi["uuid"], "--run-jobs-here"] + (["--no-reuse"] if args.no_reuse else []))
370 elif not args.no_wait:
371 ws.main(["--pipeline", pi["uuid"]])
373 pi = api.pipeline_instances().get(uuid=pi["uuid"]).execute()
374 logger.info("Pipeline is %s", pi["state"])
375 if "output_uuid" in pi["components"]["command"]:
376 logger.info("Output is %s", pi["components"]["command"]["output_uuid"])
378 logger.info("No output")
380 if __name__ == '__main__':