+# Copyright (C) The Arvados Authors. All rights reserved.
+#
+# SPDX-License-Identifier: Apache-2.0
+
import os
import urlparse
from functools import partial
import cwltool.workflow
from cwltool.process import get_feature, scandeps, UnsupportedRequirement, normalizeFilesDirs, shortname
from cwltool.load_tool import fetch_document
-from cwltool.pathmapper import adjustFileObjs, adjustDirObjs
+from cwltool.pathmapper import adjustFileObjs, adjustDirObjs, visit_class
from cwltool.utils import aslist
from cwltool.builder import substitute
from cwltool.pack import pack
import ruamel.yaml as yaml
from .arvdocker import arv_docker_get_image
-from .pathmapper import ArvPathMapper
+from .pathmapper import ArvPathMapper, trim_listing
from ._version import __version__
from . import done
logger = logging.getLogger('arvados.cwl-runner')
-def trim_listing(obj):
- """Remove 'listing' field from Directory objects that are keep references.
+def trim_anonymous_location(obj):
+ """Remove 'location' field from File and Directory literals.
+
+ To make internal handling easier, literals are assigned a random id for
+ 'location'. However, when writing the record back out, this can break
+ reproducibility. Since it is valid for literals not have a 'location'
+ field, remove it.
- When Directory objects represent Keep references, it redundant and
- potentially very expensive to pass fully enumerated Directory objects
- between instances of cwl-runner (e.g. a submitting a job, or using the
- RunInSingleContainer feature), so delete the 'listing' field when it is
- safe to do so.
"""
- if obj.get("location", "").startswith("keep:") and "listing" in obj:
- del obj["listing"]
if obj.get("location", "").startswith("_:"):
del obj["location"]
+def remove_redundant_fields(obj):
+ for field in ("path", "nameext", "nameroot", "dirname"):
+ if field in obj:
+ del obj[field]
+
+def find_defaults(d, op):
+ if isinstance(d, list):
+ for i in d:
+ find_defaults(i, op)
+ elif isinstance(d, dict):
+ if "default" in d:
+ op(d)
+ else:
+ for i in d.itervalues():
+ find_defaults(i, op)
+
def upload_dependencies(arvrunner, name, document_loader,
workflowobj, uri, loadref_run, include_primary=True):
"""Upload the dependencies of the workflowobj document to Keep.
for s in workflowobj["$schemas"]:
sc.append({"class": "File", "location": s})
+ def capture_default(obj):
+ remove = [False]
+ def add_default(f):
+ if "location" not in f and "path" in f:
+ f["location"] = f["path"]
+ del f["path"]
+ if "location" in f and not arvrunner.fs_access.exists(f["location"]):
+ # Remove from sc
+ sc[:] = [x for x in sc if x["location"] != f["location"]]
+ # Delete "default" from workflowobj
+ remove[0] = True
+ visit_class(obj["default"], ("File", "Directory"), add_default)
+ if remove[0]:
+ del obj["default"]
+
+ find_defaults(workflowobj, capture_default)
+
mapper = ArvPathMapper(arvrunner, sc, "",
"keep:%s",
"keep:%s/%s",
- name=name)
+ name=name,
+ single_collection=True)
def setloc(p):
if "location" in p and (not p["location"].startswith("_:")) and (not p["location"].startswith("keep:")):
raise SourceLine(docker_req, "dockerOutputDirectory", UnsupportedRequirement).makeError(
"Option 'dockerOutputDirectory' of DockerRequirement not supported.")
arv_docker_get_image(arvrunner.api, docker_req, True, arvrunner.project_uuid)
+ else:
+ arv_docker_get_image(arvrunner.api, {"dockerPull": "arvados/jobs"}, True, arvrunner.project_uuid)
elif isinstance(tool, cwltool.workflow.Workflow):
for s in tool.steps:
upload_docker(arvrunner, s.embedded_tool)
packed["http://schema.org/version"] = githash
-def upload_job_order(arvrunner, name, tool, job_order):
- """Upload local files referenced in the input object and return updated input
- object with 'location' updated to the proper keep references.
- """
-
- for t in tool.tool["inputs"]:
+def discover_secondary_files(inputs, job_order):
+ for t in inputs:
def setSecondary(fileobj):
if isinstance(fileobj, dict) and fileobj.get("class") == "File":
if "secondaryFiles" not in fileobj:
if shortname(t["id"]) in job_order and t.get("secondaryFiles"):
setSecondary(job_order[shortname(t["id"])])
+def upload_job_order(arvrunner, name, tool, job_order):
+ """Upload local files referenced in the input object and return updated input
+ object with 'location' updated to the proper keep references.
+ """
+
+ discover_secondary_files(tool.tool["inputs"], job_order)
+
jobmapper = upload_dependencies(arvrunner,
name,
tool.doc_loader,
return job_order
-def upload_workflow_deps(arvrunner, tool):
+def upload_workflow_deps(arvrunner, tool, override_tools):
# Ensure that Docker images needed by this workflow are available
upload_docker(arvrunner, tool)
False,
include_primary=False)
document_loader.idx[deptool["id"]] = deptool
+ override_tools[deptool["id"]] = json.dumps(deptool)
tool.visit(upload_tool_deps)
def __init__(self, runner, tool, job_order, enable_reuse,
output_name, output_tags, submit_runner_ram=0,
- name=None, on_error=None, submit_runner_image=None):
+ name=None, on_error=None, submit_runner_image=None,
+ intermediate_output_ttl=0):
self.arvrunner = runner
self.tool = tool
self.job_order = job_order
self.running = False
+ if enable_reuse:
+ # If reuse is permitted by command line arguments but
+ # disabled by the workflow itself, disable it.
+ reuse_req, _ = get_feature(self.tool, "http://arvados.org/cwl#ReuseRequirement")
+ if reuse_req:
+ enable_reuse = reuse_req["enableReuse"]
self.enable_reuse = enable_reuse
self.uuid = None
self.final_output = None
self.name = name
self.on_error = on_error
self.jobs_image = submit_runner_image or "arvados/jobs:"+__version__
+ self.intermediate_output_ttl = intermediate_output_ttl
if submit_runner_ram:
self.submit_runner_ram = submit_runner_ram
else:
- self.submit_runner_ram = 1024
+ self.submit_runner_ram = 3000
if self.submit_runner_ram <= 0:
raise Exception("Value of --submit-runner-ram must be greater than zero")
projects / arvados.git / blobdiff