X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/846a2728f3cfc80efc220a60536dea16b5f0fcbd..5030797bc2035f1feee5876e069c0282308270ef:/crunch_scripts/cwl-runner

diff --git a/crunch_scripts/cwl-runner b/crunch_scripts/cwl-runner
index 59f6817336..c786fc10b1 100755
--- a/crunch_scripts/cwl-runner
+++ b/crunch_scripts/cwl-runner
@@ -1,5 +1,13 @@
 #!/usr/bin/env python
 
+# 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
@@ -9,32 +17,80 @@ import cwltool.main
 import logging
 import os
 import json
+import argparse
 from arvados.api import OrderedJsonModel
+from cwltool.process import adjustFiles
+from cwltool.load_tool import load_tool
+
+# Print package versions
+logging.info(cwltool.main.versionstring())
 
 api = arvados.api("v1")
 
 try:
     job_order_object = arvados.current_job()['script_parameters']
 
-    print job_order_object
+    def keeppath(v):
+        if arvados.util.keep_locator_pattern.match(v):
+            return "file://%s/%s" % (os.environ['TASK_KEEPMOUNT'], v)
+        else:
+            return v
+
+    job_order_object["cwl:tool"] = keeppath(job_order_object["cwl:tool"])
 
     for k,v in job_order_object.items():
-        if arvados.util.keep_locator_pattern.match(v):
-            job_order_object[k] = "file://%s/%s" % (os.environ['TASK_KEEPMOUNT'], v)
+        if isinstance(v, basestring) and arvados.util.keep_locator_pattern.match(v):
+            job_order_object[k] = {
+                "class": "File",
+                "path": keeppath(v)
+            }
+
+    adjustFiles(job_order_object, keeppath)
 
     runner = arvados_cwl.ArvCwlRunner(api_client=arvados.api('v1', model=OrderedJsonModel()))
 
-    t = cwltool.main.load_tool(job_order_object, False, True, runner.arvMakeTool, True)
+    t = load_tool(job_order_object, runner.arvMakeTool)
+
+    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.arvExecutor(t, job_order_object, **vars(args))
+
+    files = {}
+    def capture(path):
+        sp = path.split("/")
+        col = sp[0][5:]
+        if col not in files:
+            files[col] = set()
+        files[col].add("/".join(sp[1:]))
+        return path
+
+    adjustFiles(outputObj, capture)
+
+    final = arvados.collection.Collection()
+
+    for k,v in files.iteritems():
+        with arvados.collection.Collection(k) as c:
+            for f in c:
+                final.copy(f, f, c, True)
+
+    def makeRelative(path):
+        return "/".join(path.split("/")[1:])
 
-    outputObj = runner.arvExecutor(t, job_order_object, "", None)
+    adjustFiles(outputObj, makeRelative)
 
-    c = arvados.collection.Collection()
-    with c.open("cwl.output.json", "w") as f:
+    with final.open("cwl.output.json", "w") as f:
         json.dump(outputObj, f, indent=4)
 
     api.job_tasks().update(uuid=arvados.current_task()['uuid'],
                                          body={
-                                             'output': c.save_new(create_collection_record=False),
+                                             'output': final.save_new(create_collection_record=False),
                                              'success': True,
                                              'progress':1.0
                                          }).execute()