X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/3f726dd7022da6e2be816ba6cc493212596087f5..f4c3100bad26dff3c99ff4bb9fa19b0d9f7995c7:/sdk/python/arvados/commands/run.py diff --git a/sdk/python/arvados/commands/run.py b/sdk/python/arvados/commands/run.py index 6cebb44ed4..8815565460 100644 --- a/sdk/python/arvados/commands/run.py +++ b/sdk/python/arvados/commands/run.py @@ -1,6 +1,7 @@ #!/usr/bin/env python import arvados +import arvados.commands.ws as ws import argparse import json import re @@ -8,148 +9,314 @@ import os import stat import put import time -#import arvados.command.ws as ws import subprocess +import logging +import arvados.commands._util as arv_cmd -arvrun_parser = argparse.ArgumentParser() -arvrun_parser.add_argument('--dry-run', action="store_true") -arvrun_parser.add_argument('--local', action="store_true") -arvrun_parser.add_argument('--docker-image', type=str, default="arvados/jobs") -arvrun_parser.add_argument('--git-dir', type=str, default="") -arvrun_parser.add_argument('command') -arvrun_parser.add_argument('args', nargs=argparse.REMAINDER) +logger = logging.getLogger('arvados.arv-run') +logger.setLevel(logging.INFO) -needupload_files = [] +arvrun_parser = argparse.ArgumentParser(parents=[arv_cmd.retry_opt]) +arvrun_parser.add_argument('--dry-run', action="store_true", help="Print out the pipeline that would be submitted and exit") +arvrun_parser.add_argument('--local', action="store_true", help="Run locally using arv-run-pipeline-instance") +arvrun_parser.add_argument('--docker-image', type=str, help="Docker image to use, otherwise use instance default.") +arvrun_parser.add_argument('--ignore-rcode', action="store_true", help="Commands that return non-zero return codes should not be considered failed.") +arvrun_parser.add_argument('--no-reuse', action="store_true", help="Do not reuse past jobs.") +arvrun_parser.add_argument('--no-wait', action="store_true", help="Do not wait and display logs after submitting command, just exit.") +arvrun_parser.add_argument('--project-uuid', type=str, help="Parent project of the pipeline") +arvrun_parser.add_argument('--git-dir', type=str, default="", help="Git repository passed to arv-crunch-job when using --local") +arvrun_parser.add_argument('--repository', type=str, default="arvados", help="repository field of component, default 'arvados'") +arvrun_parser.add_argument('--script-version', type=str, default="master", help="script_version field of component, default 'master'") +arvrun_parser.add_argument('args', nargs=argparse.REMAINDER) class ArvFile(object): def __init__(self, prefix, fn): self.prefix = prefix self.fn = fn -def statfile(prefix, fn, pattern): +class UploadFile(ArvFile): + pass + +# Determine if a file is in a collection, and return a tuple consisting of the +# portable data hash and the path relative to the root of the collection. +# Return None if the path isn't with an arv-mount collection or there was is error. +def is_in_collection(root, branch): + try: + if root == "/": + return (None, None) + fn = os.path.join(root, ".arvados#collection") + if os.path.exists(fn): + with file(fn, 'r') as f: + c = json.load(f) + return (c["portable_data_hash"], branch) + else: + sp = os.path.split(root) + return is_in_collection(sp[0], os.path.join(sp[1], branch)) + except (IOError, OSError): + return (None, None) + +# Determine the project to place the output of this command by searching upward +# for arv-mount psuedofile indicating the project. If the cwd isn't within +# an arv-mount project or there is an error, return current_user. +def determine_project(root, current_user): + try: + if root == "/": + return current_user + fn = os.path.join(root, ".arvados#project") + if os.path.exists(fn): + with file(fn, 'r') as f: + c = json.load(f) + if 'writable_by' in c and current_user in c['writable_by']: + return c["uuid"] + else: + return current_user + else: + sp = os.path.split(root) + return determine_project(sp[0], current_user) + except (IOError, OSError): + return current_user + +# Determine if string corresponds to a file, and if that file is part of a +# arv-mounted collection or only local to the machine. Returns one of +# ArvFile() (file already exists in a collection), UploadFile() (file needs to +# be uploaded to a collection), or simply returns prefix+fn (which yields the +# original parameter string). +def statfile(prefix, fn): absfn = os.path.abspath(fn) if os.path.exists(absfn): - fn = os.path.abspath(fn) - st = os.stat(fn) + st = os.stat(absfn) if stat.S_ISREG(st.st_mode): - mount = os.path.dirname(fn)+"/.arvados#collection" - if os.path.exists(mount): - with file(mount, 'r') as f: - c = json.load(f) - return prefix+"$(file "+c["portable_data_hash"]+"/" + os.path.basename(fn) + ")" + sp = os.path.split(absfn) + (pdh, branch) = is_in_collection(sp[0], sp[1]) + if pdh: + return ArvFile(prefix, "$(file %s/%s)" % (pdh, branch)) else: - needupload_files.append(fn) - return ArvFile(prefix, fn[1:]) + # trim leading '/' for path prefix test later + return UploadFile(prefix, absfn[1:]) + if stat.S_ISDIR(st.st_mode): + sp = os.path.split(absfn) + (pdh, branch) = is_in_collection(sp[0], sp[1]) + if pdh: + return ArvFile(prefix, "$(dir %s/%s/)" % (pdh, branch)) + return prefix+fn +def uploadfiles(files, api, dry_run=False, num_retries=0, project=None, fnPattern="$(file %s/%s)"): + # Find the smallest path prefix that includes all the files that need to be uploaded. + # This starts at the root and iteratively removes common parent directory prefixes + # until all file pathes no longer have a common parent. + n = True + pathprefix = "/" + while n: + pathstep = None + for c in files: + if pathstep is None: + sp = c.fn.split('/') + if len(sp) < 2: + # no parent directories left + n = False + break + # path step takes next directory + pathstep = sp[0] + "/" + else: + # check if pathstep is common prefix for all files + if not c.fn.startswith(pathstep): + n = False + break + if n: + # pathstep is common parent directory for all files, so remove the prefix + # from each path + pathprefix += pathstep + for c in files: + c.fn = c.fn[len(pathstep):] + + orgdir = os.getcwd() + os.chdir(pathprefix) + + logger.info("Upload local files: \"%s\"", '" "'.join([c.fn for c in files])) + + if dry_run: + logger.info("$(input) is %s", pathprefix.rstrip('/')) + pdh = "$(input)" + else: + files = sorted(files, key=lambda x: x.fn) + collection = arvados.CollectionWriter(api, num_retries=num_retries) + stream = None + for f in files: + sp = os.path.split(f.fn) + if sp[0] != stream: + stream = sp[0] + collection.start_new_stream(stream) + collection.write_file(f.fn, sp[1]) + item = api.collections().create(body={"owner_uuid": project, "manifest_text": collection.manifest_text()}).execute() + pdh = item["portable_data_hash"] + logger.info("Uploaded to %s", item["uuid"]) + + for c in files: + c.fn = fnPattern % (pdh, c.fn) + + os.chdir(orgdir) + + def main(arguments=None): args = arvrun_parser.parse_args(arguments) - patterns = [re.compile("(--[^=]+=)(.*)"), - re.compile("(-[^=]+=)(.*)"), - re.compile("(-.)(.+)")] + if len(args.args) == 0: + arvrun_parser.print_help() + return - commandargs = [] + starting_args = args.args - for a in args.args: - if a[0] == '-': - matched = False - for p in patterns: - m = p.match(a) - if m: - commandargs.append(statfile(m.group(1), m.group(2), p)) - matched = True - break - if not matched: - commandargs.append(a) + reading_into = 2 + + # Parse the command arguments into 'slots'. + # All words following '>' are output arguments and are collected into slots[0]. + # All words following '<' are input arguments and are collected into slots[1]. + # slots[2..] store the parameters of each command in the pipeline. + # + # e.g. arv-run foo arg1 arg2 '|' bar arg3 arg4 '<' input1 input2 input3 '>' output.txt + # will be parsed into: + # [['output.txt'], + # ['input1', 'input2', 'input3'], + # ['foo', 'arg1', 'arg2'], + # ['bar', 'arg3', 'arg4']] + slots = [[], [], []] + for c in args.args: + if c.startswith('>'): + reading_into = 0 + if len(c) > 1: + slots[reading_into].append(c[1:]) + elif c.startswith('<'): + reading_into = 1 + if len(c) > 1: + slots[reading_into].append(c[1:]) + elif c == '|': + reading_into = len(slots) + slots.append([]) else: - commandargs.append(statfile('', a, None)) + slots[reading_into].append(c) - n = True - pathprefix = "/" - files = [c for c in commandargs if isinstance(c, ArvFile)] - if len(files) > 0: - while n: - pathstep = None - for c in files: - if pathstep is None: - sp = c.fn.split('/') - if len(sp) < 2: - n = False - break - pathstep = sp[0] + "/" - else: - if not c.fn.startswith(pathstep): - n = False - break - if n: - pathprefix += pathstep - for c in files: - c.fn = c.fn[len(pathstep):] - - os.chdir(pathprefix) - - if args.dry_run: - print("cd %s" % pathprefix) - print("arv-put \"%s\"" % '" "'.join([c.fn for c in files])) - pdh = "$(input)" + if slots[0] and len(slots[0]) > 1: + logger.error("Can only specify a single stdout file (run-command substitutions are permitted)") + return + + if not args.dry_run: + api = arvados.api('v1') + if args.project_uuid: + project = args.project_uuid else: - pdh = put.main(["--portable-data-hash"]+[c.fn for c in files]) - - commandargs = [("%s$(file %s/%s)" % (c.prefix, pdh, c.fn)) if isinstance(c, ArvFile) else c for c in commandargs] - - cut = None - i = -1 - stdio = [None, None] - for j in xrange(0, len(commandargs)): - c = commandargs[j] - if c == '<': - stdio[0] = [] - i = 0 - cut = j if cut is None else cut - elif c == '>': - stdio[1] = [] - i = 1 - cut = j if cut is None else cut - elif i > -1: - stdio[i].append(c) - - if cut is not None: - commandargs = commandargs[:cut] + project = determine_project(os.getcwd(), api.users().current().execute()["uuid"]) + + # Identify input files. Look at each parameter and test to see if there is + # a file by that name. This uses 'patterns' to look for within + # command line arguments, such as --foo=file.txt or -lfile.txt + patterns = [re.compile("([^=]+=)(.*)"), + re.compile("(-[A-Za-z])(.+)")] + for j, command in enumerate(slots[1:]): + for i, a in enumerate(command): + if j > 0 and i == 0: + # j == 0 is stdin, j > 0 is commands + # always skip program executable (i == 0) in commands + pass + elif a.startswith('\\'): + # if it starts with a \ then don't do any interpretation + command[i] = a[1:] + else: + # See if it looks like a file + command[i] = statfile('', a) + + # If a file named command[i] was found, it would now be an + # ArvFile or UploadFile. If command[i] is a basestring, that + # means it doesn't correspond exactly to a file, so do some + # pattern matching. + if isinstance(command[i], basestring): + for p in patterns: + m = p.match(a) + if m: + command[i] = statfile(m.group(1), m.group(2)) + break + + files = [c for command in slots[1:] for c in command if isinstance(c, UploadFile)] + if files: + uploadfiles(files, api, dry_run=args.dry_run, num_retries=args.num_retries, project=project) + + for i in xrange(1, len(slots)): + slots[i] = [("%s%s" % (c.prefix, c.fn)) if isinstance(c, ArvFile) else c for c in slots[i]] component = { "script": "run-command", - "script_version": "3609-arv-run", - "repository": "arvados", + "script_version": args.script_version, + "repository": args.repository, "script_parameters": { - "command": [args.command]+commandargs }, - "runtime_constraints": { - "docker_image": args.docker_image - } + "runtime_constraints": {} } - if stdio[0]: - component["script_parameters"]["task.stdin"] = stdio[0][0] - if stdio[1]: - component["script_parameters"]["task.stdout"] = stdio[1][0] + if args.docker_image: + component["runtime_constraints"]["docker_image"] = args.docker_image + + task_foreach = [] + group_parser = argparse.ArgumentParser() + group_parser.add_argument('-b', '--batch-size', type=int) + group_parser.add_argument('args', nargs=argparse.REMAINDER) + + for s in xrange(2, len(slots)): + for i in xrange(0, len(slots[s])): + if slots[s][i] == '--': + inp = "input%i" % (s-2) + groupargs = group_parser.parse_args(slots[2][i+1:]) + if groupargs.batch_size: + component["script_parameters"][inp] = {"value": {"batch":groupargs.args, "size":groupargs.batch_size}} + slots[s] = slots[s][0:i] + [{"foreach": inp, "command": "$(%s)" % inp}] + else: + component["script_parameters"][inp] = groupargs.args + slots[s] = slots[s][0:i] + ["$(%s)" % inp] + task_foreach.append(inp) + break + if slots[s][i] == '\--': + slots[s][i] = '--' + + if slots[0]: + component["script_parameters"]["task.stdout"] = slots[0][0] + if slots[1]: + task_foreach.append("stdin") + component["script_parameters"]["stdin"] = slots[1] + component["script_parameters"]["task.stdin"] = "$(stdin)" + + if task_foreach: + component["script_parameters"]["task.foreach"] = task_foreach + + component["script_parameters"]["command"] = slots[2:] + if args.ignore_rcode: + component["script_parameters"]["task.ignore_rcode"] = args.ignore_rcode pipeline = { - "name": "", + "name": "arv-run " + " | ".join([s[0] for s in slots[2:]]), + "description": "@" + " ".join(starting_args) + "@", "components": { - args.command: component + "command": component }, - "state":"RunningOnServer" + "state": "RunningOnClient" if args.local else "RunningOnServer" } if args.dry_run: print(json.dumps(pipeline, indent=4)) - elif args.local: - subprocess.call(["arv-crunch-job", "--job", json.dumps(component), "--git-dir", args.git_dir]) else: - api = arvados.api('v1') - pi = api.pipeline_instances().create(body=pipeline).execute() - #ws.main(["--pipeline", pi["uuid"]]) + pipeline["owner_uuid"] = project + pi = api.pipeline_instances().create(body=pipeline, ensure_unique_name=True).execute() + logger.info("Running pipeline %s", pi["uuid"]) + + if args.local: + subprocess.call(["arv-run-pipeline-instance", "--instance", pi["uuid"], "--run-jobs-here"] + (["--no-reuse"] if args.no_reuse else [])) + elif not args.no_wait: + ws.main(["--pipeline", pi["uuid"]]) + + pi = api.pipeline_instances().get(uuid=pi["uuid"]).execute() + logger.info("Pipeline is %s", pi["state"]) + if "output_uuid" in pi["components"]["command"]: + logger.info("Output is %s", pi["components"]["command"]["output_uuid"]) + else: + logger.info("No output") if __name__ == '__main__': main()