From: Peter Amstutz <peter.amstutz@curoverse.com>
Date: Mon, 17 Oct 2016 19:07:09 +0000 (-0400)
Subject: Merge branch 'master' into 10259-cwl-pipeline-null
X-Git-Tag: 1.1.0~654^2~1
X-Git-Url: https://git.arvados.org/arvados.git/commitdiff_plain/a85e7c7526d985f66f073320155ae5d70590be60?hp=f156def25b0720dac1e102727f3a3821b98fb551

Merge branch 'master' into 10259-cwl-pipeline-null
---

diff --git a/crunch_scripts/cwl-runner b/crunch_scripts/cwl-runner
index 5d7f553fec..ff2addf8fd 100755
--- a/crunch_scripts/cwl-runner
+++ b/crunch_scripts/cwl-runner
@@ -1,15 +1,30 @@
 #!/usr/bin/env python
 
-# Crunch script integration for running arvados-cwl-runner (importing
-# arvados_cwl module) inside a crunch job.
-#
+# Crunch script integration for running arvados-cwl-runner inside a crunch job.
+
+import arvados_cwl
+import sys
+
+try:
+    # Use the crunch script defined in the arvados_cwl package.  This helps
+    # prevent the crunch script from going out of sync with the rest of the
+    # arvados_cwl package.
+    import arvados_cwl.crunch_script
+    arvados_cwl.crunch_script.run()
+    sys.exit()
+except ImportError:
+    pass
+
+# When running against an older arvados-cwl-runner package without
+# arvados_cwl.crunch_script, fall back to the old code.
+
+
 # This gets the job record, transforms the script parameters into a valid CWL
 # input object, then executes the CWL runner to run the underlying workflow or
 # tool.  When the workflow completes, record the output object in an output
 # collection for this runner job.
 
 import arvados
-import arvados_cwl
 import arvados.collection
 import arvados.util
 import cwltool.main
diff --git a/sdk/cwl/arvados_cwl/crunch_script.py b/sdk/cwl/arvados_cwl/crunch_script.py
new file mode 100644
index 0000000000..e5c6d67ac9
--- /dev/null
+++ b/sdk/cwl/arvados_cwl/crunch_script.py
@@ -0,0 +1,102 @@
+# Crunch script integration for running arvados-cwl-runner (importing
+# arvados_cwl module) inside a crunch job.
+#
+# This gets the job record, transforms the script parameters into a valid CWL
+# input object, then executes the CWL runner to run the underlying workflow or
+# tool.  When the workflow completes, record the output object in an output
+# collection for this runner job.
+
+import arvados
+import arvados_cwl
+import arvados.collection
+import arvados.util
+import cwltool.main
+import logging
+import os
+import json
+import argparse
+import re
+import functools
+
+from arvados.api import OrderedJsonModel
+from cwltool.process import shortname, adjustFileObjs, adjustDirObjs, getListing, normalizeFilesDirs
+from cwltool.load_tool import load_tool
+
+logger = logging.getLogger('arvados.cwl-runner')
+
+def run():
+    # Print package versions
+    logger.info(arvados_cwl.versionstring())
+
+    api = arvados.api("v1")
+
+    arvados_cwl.add_arv_hints()
+
+    try:
+        job_order_object = arvados.current_job()['script_parameters']
+
+        pdh_path = re.compile(r'^[0-9a-f]{32}\+\d+(/.+)?$')
+
+        def keeppath(v):
+            if pdh_path.match(v):
+                return "keep:%s" % v
+            else:
+                return v
+
+        def keeppathObj(v):
+            v["location"] = keeppath(v["location"])
+
+        job_order_object["cwl:tool"] = "file://%s/%s" % (os.environ['TASK_KEEPMOUNT'], job_order_object["cwl:tool"])
+
+        for k,v in job_order_object.items():
+            if isinstance(v, basestring) and arvados.util.keep_locator_pattern.match(v):
+                job_order_object[k] = {
+                    "class": "File",
+                    "location": "keep:%s" % v
+                }
+
+        adjustFileObjs(job_order_object, keeppathObj)
+        adjustDirObjs(job_order_object, keeppathObj)
+        normalizeFilesDirs(job_order_object)
+        adjustDirObjs(job_order_object, functools.partial(getListing, arvados_cwl.fsaccess.CollectionFsAccess("", api_client=api)))
+
+        output_name = None
+        if "arv:output_name" in job_order_object:
+            output_name = job_order_object["arv:output_name"]
+            del job_order_object["arv:output_name"]
+
+        runner = arvados_cwl.ArvCwlRunner(api_client=arvados.api('v1', model=OrderedJsonModel()),
+                                          output_name=output_name)
+
+        t = load_tool(job_order_object, runner.arv_make_tool)
+
+        args = argparse.Namespace()
+        args.project_uuid = arvados.current_job()["owner_uuid"]
+        args.enable_reuse = True
+        args.submit = False
+        args.debug = True
+        args.quiet = False
+        args.ignore_docker_for_reuse = False
+        args.basedir = os.getcwd()
+        args.cwl_runner_job={"uuid": arvados.current_job()["uuid"], "state": arvados.current_job()["state"]}
+        outputObj = runner.arv_executor(t, job_order_object, **vars(args))
+
+        if runner.final_output_collection:
+            outputCollection = runner.final_output_collection.portable_data_hash()
+        else:
+            outputCollection = None
+
+        api.job_tasks().update(uuid=arvados.current_task()['uuid'],
+                                             body={
+                                                 'output': outputCollection,
+                                                 'success': True,
+                                                 'progress':1.0
+                                             }).execute()
+    except Exception as e:
+        logging.exception("Unhandled exception")
+        api.job_tasks().update(uuid=arvados.current_task()['uuid'],
+                                             body={
+                                                 'output': None,
+                                                 'success': False,
+                                                 'progress':1.0
+                                             }).execute()