Merge branch '10025-arvbox-layers' closes #10025

[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 cf2f1db997b9db9688b52258994bc6179863783d..6d382450e2fa83f4c81b8567224f742c3c389db2 100644 (file)
--- a/sdk/cwl/arvados_cwl/runner.py
+++ b/sdk/cwl/arvados_cwl/runner.py
@@ -21,7 +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.
+    """
+
+    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):
@@ -80,8 +95,8 @@ def upload_dependencies(arvrunner, name, document_loader,
                            name=name)
 
     def setloc(p):
-        if not p["location"].startswith("_:") and not p["location"].startswith("keep:"):
-            p["location"] = mapper.mapper(p["location"]).target
+        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)
 
@@ -97,44 +112,49 @@ def upload_docker(arvrunner, tool):
         for s in tool.steps:
             upload_docker(arvrunner, s.embedded_tool)
 
+def upload_instance(arvrunner, name, tool, job_order):
+        upload_docker(arvrunner, tool)
+
+        workflowmapper = upload_dependencies(arvrunner,
+                                             name,
+                                             tool.doc_loader,
+                                             tool.tool,
+                                             tool.tool["id"],
+                                             True)
+
+        jobmapper = upload_dependencies(arvrunner,
+                                        os.path.basename(job_order.get("id", "#")),
+                                        tool.doc_loader,
+                                        job_order,
+                                        job_order.get("id", "#"),
+                                        False)
+
+        if "id" in job_order:
+            del job_order["id"]
+
+        return workflowmapper
+
 
 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
 
     def arvados_job_spec(self, *args, **kwargs):
-        upload_docker(self.arvrunner, self.tool)
-
         self.name = os.path.basename(self.tool.tool["id"])
-
-        workflowmapper = upload_dependencies(self.arvrunner,
-                                             self.name,
-                                             self.tool.doc_loader,
-                                             self.tool.tool,
-                                             self.tool.tool["id"],
-                                             True)
-
-        jobmapper = upload_dependencies(self.arvrunner,
-                                        os.path.basename(self.job_order.get("id", "#")),
-                                        self.tool.doc_loader,
-                                        self.job_order,
-                                        self.job_order.get("id", "#"),
-                                        False)
-
-        if "id" in self.job_order:
-            del self.job_order["id"]
-
+        workflowmapper = upload_instance(self.arvrunner, self.name, self.tool, self.job_order)
+        adjustDirObjs(self.job_order, trim_listing)
         return workflowmapper
 
-
     def done(self, record):
         if record["state"] == "Complete":
             if record.get("exit_code") is not None:
@@ -152,7 +172,11 @@ 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):