1 from __future__ import print_function
2 from __future__ import absolute_import
3 from builtins import range
4 from past.builtins import basestring
5 from builtins import object
7 import arvados.commands.ws as ws
19 import arvados.commands._util as arv_cmd
20 import arvados.collection
22 from arvados._version import __version__
24 logger = logging.getLogger('arvados.arv-run')
25 logger.setLevel(logging.INFO)
27 arvrun_parser = argparse.ArgumentParser(parents=[arv_cmd.retry_opt])
28 arvrun_parser.add_argument('--dry-run', action="store_true",
29 help="Print out the pipeline that would be submitted and exit")
30 arvrun_parser.add_argument('--local', action="store_true",
31 help="Run locally using arv-run-pipeline-instance")
32 arvrun_parser.add_argument('--docker-image', type=str,
33 help="Docker image to use, otherwise use instance default.")
34 arvrun_parser.add_argument('--ignore-rcode', action="store_true",
35 help="Commands that return non-zero return codes should not be considered failed.")
36 arvrun_parser.add_argument('--no-reuse', action="store_true",
37 help="Do not reuse past jobs.")
38 arvrun_parser.add_argument('--no-wait', action="store_true",
39 help="Do not wait and display logs after submitting command, just exit.")
40 arvrun_parser.add_argument('--project-uuid', type=str,
41 help="Parent project of the pipeline")
42 arvrun_parser.add_argument('--git-dir', type=str, default="",
43 help="Git repository passed to arv-crunch-job when using --local")
44 arvrun_parser.add_argument('--repository', type=str, default="arvados",
45 help="repository field of component, default 'arvados'")
46 arvrun_parser.add_argument('--script-version', type=str, default="master",
47 help="script_version field of component, default 'master'")
48 arvrun_parser.add_argument('--version', action='version',
49 version="%s %s" % (sys.argv[0], __version__),
50 help='Print version and exit.')
51 arvrun_parser.add_argument('args', nargs=argparse.REMAINDER)
53 class ArvFile(object):
54 def __init__(self, prefix, fn):
59 return (self.prefix+self.fn).__hash__()
61 def __eq__(self, other):
62 return (self.prefix == other.prefix) and (self.fn == other.fn)
64 class UploadFile(ArvFile):
67 # Determine if a file is in a collection, and return a tuple consisting of the
68 # portable data hash and the path relative to the root of the collection.
69 # Return None if the path isn't with an arv-mount collection or there was is error.
70 def is_in_collection(root, branch):
74 fn = os.path.join(root, ".arvados#collection")
75 if os.path.exists(fn):
76 with file(fn, 'r') as f:
78 return (c["portable_data_hash"], branch)
80 sp = os.path.split(root)
81 return is_in_collection(sp[0], os.path.join(sp[1], branch))
82 except (IOError, OSError):
85 # Determine the project to place the output of this command by searching upward
86 # for arv-mount psuedofile indicating the project. If the cwd isn't within
87 # an arv-mount project or there is an error, return current_user.
88 def determine_project(root, current_user):
92 fn = os.path.join(root, ".arvados#project")
93 if os.path.exists(fn):
94 with file(fn, 'r') as f:
96 if 'writable_by' in c and current_user in c['writable_by']:
101 sp = os.path.split(root)
102 return determine_project(sp[0], current_user)
103 except (IOError, OSError):
106 # Determine if string corresponds to a file, and if that file is part of a
107 # arv-mounted collection or only local to the machine. Returns one of
108 # ArvFile() (file already exists in a collection), UploadFile() (file needs to
109 # be uploaded to a collection), or simply returns prefix+fn (which yields the
110 # original parameter string).
111 def statfile(prefix, fn, fnPattern="$(file %s/%s)", dirPattern="$(dir %s/%s/)", raiseOSError=False):
112 absfn = os.path.abspath(fn)
115 sp = os.path.split(absfn)
116 (pdh, branch) = is_in_collection(sp[0], sp[1])
118 if stat.S_ISREG(st.st_mode):
119 return ArvFile(prefix, fnPattern % (pdh, branch))
120 elif stat.S_ISDIR(st.st_mode):
121 return ArvFile(prefix, dirPattern % (pdh, branch))
123 raise Exception("%s is not a regular file or directory" % absfn)
125 # trim leading '/' for path prefix test later
126 return UploadFile(prefix, absfn[1:])
128 if e.errno == errno.ENOENT and not raiseOSError:
135 def write_file(collection, pathprefix, fn):
136 with open(os.path.join(pathprefix, fn)) as src:
137 dst = collection.open(fn, "w")
138 r = src.read(1024*128)
141 r = src.read(1024*128)
142 dst.close(flush=False)
144 def uploadfiles(files, api, dry_run=False, num_retries=0,
146 fnPattern="$(file %s/%s)",
149 # Find the smallest path prefix that includes all the files that need to be uploaded.
150 # This starts at the root and iteratively removes common parent directory prefixes
151 # until all file paths no longer have a common parent.
161 # no parent directories left
164 # path step takes next directory
165 pathstep = sp[0] + "/"
167 # check if pathstep is common prefix for all files
168 if not c.fn.startswith(pathstep):
172 # pathstep is common parent directory for all files, so remove the prefix
174 pathprefix += pathstep
176 c.fn = c.fn[len(pathstep):]
178 logger.info("Upload local files: \"%s\"", '" "'.join([c.fn for c in files]))
181 logger.info("$(input) is %s", pathprefix.rstrip('/'))
184 files = sorted(files, key=lambda x: x.fn)
185 if collection is None:
186 collection = arvados.collection.Collection(api_client=api, num_retries=num_retries)
189 localpath = os.path.join(pathprefix, f.fn)
190 if prev and localpath.startswith(prev+"/"):
191 # If this path is inside an already uploaded subdirectory,
192 # don't redundantly re-upload it.
193 # e.g. we uploaded /tmp/foo and the next file is /tmp/foo/bar
194 # skip it because it starts with "/tmp/foo/"
197 if os.path.isfile(localpath):
198 write_file(collection, pathprefix, f.fn)
199 elif os.path.isdir(localpath):
200 for root, dirs, iterfiles in os.walk(localpath):
201 root = root[len(pathprefix):]
202 for src in iterfiles:
203 write_file(collection, pathprefix, os.path.join(root, src))
205 filters=[["portable_data_hash", "=", collection.portable_data_hash()]]
207 filters.append(["name", "like", name+"%"])
209 filters.append(["owner_uuid", "=", project])
211 exists = api.collections().list(filters=filters, limit=1).execute(num_retries=num_retries)
214 item = exists["items"][0]
215 pdh = item["portable_data_hash"]
216 logger.info("Using collection %s (%s)", pdh, item["uuid"])
217 elif len(collection) > 0:
218 collection.save_new(name=name, owner_uuid=project, ensure_unique_name=True)
219 pdh = collection.portable_data_hash()
220 logger.info("Uploaded to %s (%s)", pdh, collection.manifest_locator())
223 c.keepref = "%s/%s" % (pdh, c.fn)
224 c.fn = fnPattern % (pdh, c.fn)
227 def main(arguments=None):
228 args = arvrun_parser.parse_args(arguments)
230 if len(args.args) == 0:
231 arvrun_parser.print_help()
234 starting_args = args.args
238 # Parse the command arguments into 'slots'.
239 # All words following '>' are output arguments and are collected into slots[0].
240 # All words following '<' are input arguments and are collected into slots[1].
241 # slots[2..] store the parameters of each command in the pipeline.
243 # e.g. arv-run foo arg1 arg2 '|' bar arg3 arg4 '<' input1 input2 input3 '>' output.txt
244 # will be parsed into:
246 # ['input1', 'input2', 'input3'],
247 # ['foo', 'arg1', 'arg2'],
248 # ['bar', 'arg3', 'arg4']]
251 if c.startswith('>'):
254 slots[reading_into].append(c[1:])
255 elif c.startswith('<'):
258 slots[reading_into].append(c[1:])
260 reading_into = len(slots)
263 slots[reading_into].append(c)
265 if slots[0] and len(slots[0]) > 1:
266 logger.error("Can only specify a single stdout file (run-command substitutions are permitted)")
270 api = arvados.api('v1')
271 if args.project_uuid:
272 project = args.project_uuid
274 project = determine_project(os.getcwd(), api.users().current().execute()["uuid"])
276 # Identify input files. Look at each parameter and test to see if there is
277 # a file by that name. This uses 'patterns' to look for within
278 # command line arguments, such as --foo=file.txt or -lfile.txt
279 patterns = [re.compile("([^=]+=)(.*)"),
280 re.compile("(-[A-Za-z])(.+)")]
281 for j, command in enumerate(slots[1:]):
282 for i, a in enumerate(command):
284 # j == 0 is stdin, j > 0 is commands
285 # always skip program executable (i == 0) in commands
287 elif a.startswith('\\'):
288 # if it starts with a \ then don't do any interpretation
291 # See if it looks like a file
292 command[i] = statfile('', a)
294 # If a file named command[i] was found, it would now be an
295 # ArvFile or UploadFile. If command[i] is a basestring, that
296 # means it doesn't correspond exactly to a file, so do some
298 if isinstance(command[i], basestring):
302 command[i] = statfile(m.group(1), m.group(2))
305 files = [c for command in slots[1:] for c in command if isinstance(c, UploadFile)]
307 uploadfiles(files, api, dry_run=args.dry_run, num_retries=args.retries, project=project)
309 for i in range(1, len(slots)):
310 slots[i] = [("%s%s" % (c.prefix, c.fn)) if isinstance(c, ArvFile) else c for c in slots[i]]
313 "script": "run-command",
314 "script_version": args.script_version,
315 "repository": args.repository,
316 "script_parameters": {
318 "runtime_constraints": {}
321 if args.docker_image:
322 component["runtime_constraints"]["docker_image"] = args.docker_image
325 group_parser = argparse.ArgumentParser()
326 group_parser.add_argument('-b', '--batch-size', type=int)
327 group_parser.add_argument('args', nargs=argparse.REMAINDER)
329 for s in range(2, len(slots)):
330 for i in range(0, len(slots[s])):
331 if slots[s][i] == '--':
332 inp = "input%i" % (s-2)
333 groupargs = group_parser.parse_args(slots[2][i+1:])
334 if groupargs.batch_size:
335 component["script_parameters"][inp] = {"value": {"batch":groupargs.args, "size":groupargs.batch_size}}
336 slots[s] = slots[s][0:i] + [{"foreach": inp, "command": "$(%s)" % inp}]
338 component["script_parameters"][inp] = groupargs.args
339 slots[s] = slots[s][0:i] + ["$(%s)" % inp]
340 task_foreach.append(inp)
342 if slots[s][i] == '\--':
346 component["script_parameters"]["task.stdout"] = slots[0][0]
348 task_foreach.append("stdin")
349 component["script_parameters"]["stdin"] = slots[1]
350 component["script_parameters"]["task.stdin"] = "$(stdin)"
353 component["script_parameters"]["task.foreach"] = task_foreach
355 component["script_parameters"]["command"] = slots[2:]
356 if args.ignore_rcode:
357 component["script_parameters"]["task.ignore_rcode"] = args.ignore_rcode
360 "name": "arv-run " + " | ".join([s[0] for s in slots[2:]]),
361 "description": "@" + " ".join(starting_args) + "@",
365 "state": "RunningOnClient" if args.local else "RunningOnServer"
369 print(json.dumps(pipeline, indent=4))
371 pipeline["owner_uuid"] = project
372 pi = api.pipeline_instances().create(body=pipeline, ensure_unique_name=True).execute()
373 logger.info("Running pipeline %s", pi["uuid"])
376 subprocess.call(["arv-run-pipeline-instance", "--instance", pi["uuid"], "--run-jobs-here"] + (["--no-reuse"] if args.no_reuse else []))
377 elif not args.no_wait:
378 ws.main(["--pipeline", pi["uuid"]])
380 pi = api.pipeline_instances().get(uuid=pi["uuid"]).execute()
381 logger.info("Pipeline is %s", pi["state"])
382 if "output_uuid" in pi["components"]["command"]:
383 logger.info("Output is %s", pi["components"]["command"]["output_uuid"])
385 logger.info("No output")
387 if __name__ == '__main__':