X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/d63f3ade27bb506da049164789679f9fed171005..0821f5481edd016a3744bb50d97a9e5b99cd1a0f:/sdk/cwl/arvados_cwl/runner.py

diff --git a/sdk/cwl/arvados_cwl/runner.py b/sdk/cwl/arvados_cwl/runner.py
index 570675bc4e..a4132ca3f7 100644
--- a/sdk/cwl/arvados_cwl/runner.py
+++ b/sdk/cwl/arvados_cwl/runner.py
@@ -21,10 +21,37 @@ 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):
+    """Upload the dependencies of the workflowobj document to Keep.
+
+    Returns a pathmapper object mapping local paths to keep references.  Also
+    does an in-place update of references in "workflowobj".
+
+    Use scandeps to find $import, $include, $schemas, run, File and Directory
+    fields that represent external references.
+
+    If workflowobj has an "id" field, this will reload the document to ensure
+    it is scanning the raw document prior to preprocessing.
+    """
+
     loaded = set()
     def loadref(b, u):
         joined = urlparse.urljoin(b, u)
@@ -54,28 +81,22 @@ def upload_dependencies(arvrunner, name, document_loader,
 
     sc = scandeps(uri, scanobj,
                   loadref_fields,
-                  set(("$include", "$schemas")),
+                  set(("$include", "$schemas", "location")),
                   loadref)
 
-    files = []
-    def visitFiles(path):
-        files.append(path)
-
-    adjustFileObjs(sc, visitFiles)
-    adjustDirObjs(sc, visitFiles)
-
-    normalizeFilesDirs(files)
+    normalizeFilesDirs(sc)
 
     if "id" in workflowobj:
-        files.append({"class": "File", "location": workflowobj["id"]})
+        sc.append({"class": "File", "location": workflowobj["id"]})
 
-    mapper = ArvPathMapper(arvrunner, files, "",
+    mapper = ArvPathMapper(arvrunner, sc, "",
                            "keep:%s",
                            "keep:%s/%s",
                            name=name)
 
     def setloc(p):
-        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)
 
@@ -123,6 +144,8 @@ class Runner(object):
                                         self.job_order.get("id", "#"),
                                         False)
 
+        adjustDirObjs(self.job_order, trim_listing)
+
         if "id" in self.job_order:
             del self.job_order["id"]