Enumerate Directory contents in output when --compute-checksum is

[arvados.git] / sdk / cwl / arvados_cwl / runner.py

diff --git a/sdk/cwl/arvados_cwl/runner.py b/sdk/cwl/arvados_cwl/runner.py
index 4e91d6628a337d7e549220ec76887c19ec60ccb5..e5b4e006e8cce7cac780436fc06c6dcd79882730 100644 (file)
--- a/sdk/cwl/arvados_cwl/runner.py
+++ b/sdk/cwl/arvados_cwl/runner.py
@@ -21,11 +21,22 @@ from .pathmapper import ArvPathMapper
 
 logger = logging.getLogger('arvados.cwl-runner')
 
-cwltool.draft2tool.ACCEPTLIST_RE = re.compile(r"^[a-zA-Z0-9._+-]+$")
+cwltool.draft2tool.ACCEPTLIST_RE = re.compile(r".*")
+
+def trim_listing(obj):
+    """Remove 'listing' field from Directory objects that are keep references.
+
+    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.
+    """
 
-def del_listing(obj):
     if obj.get("location", "").startswith("keep:") and "listing" in obj:
         del obj["listing"]
+    if obj.get("location", "").startswith("_:"):
+        del obj["location"]
 
 def upload_dependencies(arvrunner, name, document_loader,
                         workflowobj, uri, loadref_run):
@@ -84,7 +95,7 @@ def upload_dependencies(arvrunner, name, document_loader,
                            name=name)
 
     def setloc(p):
-        if not p["location"].startswith("_:") and not p["location"].startswith("keep:"):
+        if "location" in p and (not p["location"].startswith("_:")) and (not p["location"].startswith("keep:")):
             p["location"] = mapper.mapper(p["location"]).resolved
     adjustFileObjs(workflowobj, setloc)
     adjustDirObjs(workflowobj, setloc)
@@ -103,13 +114,15 @@ def upload_docker(arvrunner, tool):
 
 
 class Runner(object):
-    def __init__(self, runner, tool, job_order, enable_reuse):
+    def __init__(self, runner, tool, job_order, enable_reuse, output_name):
         self.arvrunner = runner
         self.tool = tool
         self.job_order = job_order
         self.running = False
         self.enable_reuse = enable_reuse
         self.uuid = None
+        self.final_output = None
+        self.output_name = output_name
 
     def update_pipeline_component(self, record):
         pass
@@ -133,7 +146,7 @@ class Runner(object):
                                         self.job_order.get("id", "#"),
                                         False)
 
-        adjustDirObjs(self.job_order, del_listing)
+        adjustDirObjs(self.job_order, trim_listing)
 
         if "id" in self.job_order:
             del self.job_order["id"]
@@ -158,7 +171,11 @@ class Runner(object):
         outputs = None
         try:
             try:
-                outc = arvados.collection.Collection(record["output"])
+                self.final_output = record["output"]
+                outc = arvados.collection.CollectionReader(self.final_output,
+                                                           api_client=self.arvrunner.api,
+                                                           keep_client=self.arvrunner.keep_client,
+                                                           num_retries=self.arvrunner.num_retries)
                 with outc.open("cwl.output.json") as f:
                     outputs = json.load(f)
                 def keepify(fileobj):