X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/014e73c96c56bc588a02bfdbdd653c10be30f80c..511e36db2a5ed0380752d15e9c60364483d0e9f0:/sdk/cwl/arvados_cwl/runner.py

diff --git a/sdk/cwl/arvados_cwl/runner.py b/sdk/cwl/arvados_cwl/runner.py
index 9f7b84a73a..155c6a0503 100644
--- a/sdk/cwl/arvados_cwl/runner.py
+++ b/sdk/cwl/arvados_cwl/runner.py
@@ -4,6 +4,7 @@ from functools import partial
 import logging
 import json
 import re
+from cStringIO import StringIO
 
 import cwltool.draft2tool
 from cwltool.draft2tool import CommandLineTool
@@ -13,6 +14,7 @@ from cwltool.load_tool import fetch_document
 from cwltool.pathmapper import adjustFileObjs, adjustDirObjs
 
 import arvados.collection
+import ruamel.yaml as yaml
 
 from .arvdocker import arv_docker_get_image
 from .pathmapper import ArvPathMapper
@@ -22,13 +24,32 @@ logger = logging.getLogger('arvados.cwl-runner')
 cwltool.draft2tool.ACCEPTLIST_RE = re.compile(r"^[a-zA-Z0-9._+-]+$")
 
 def upload_dependencies(arvrunner, name, document_loader,
-                        workflowobj, uri, keepprefix, loadref_run):
+                        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)
-        if joined not in loaded:
-            loaded.add(joined)
-            return document_loader.fetch(urlparse.urljoin(b, u))
+        defrg, _ = urlparse.urldefrag(joined)
+        if defrg not in loaded:
+            loaded.add(defrg)
+            # Use fetch_text to get raw file (before preprocessing).
+            text = document_loader.fetch_text(defrg)
+            if isinstance(text, bytes):
+                textIO = StringIO(text.decode('utf-8'))
+            else:
+                textIO = StringIO(text)
+            return yaml.safe_load(textIO)
         else:
             return {}
 
@@ -37,9 +58,15 @@ def upload_dependencies(arvrunner, name, document_loader,
     else:
         loadref_fields = set(("$import",))
 
-    sc = scandeps(uri, workflowobj,
+    scanobj = workflowobj
+    if "id" in workflowobj:
+        # Need raw file content (before preprocessing) to ensure
+        # that external references in $include and $mixin are captured.
+        scanobj = loadref("", workflowobj["id"])
+
+    sc = scandeps(uri, scanobj,
                   loadref_fields,
-                  set(("$include", "$schemas", "path", "location")),
+                  set(("$include", "$schemas", "location")),
                   loadref)
 
     files = []
@@ -55,8 +82,8 @@ def upload_dependencies(arvrunner, name, document_loader,
         files.append({"class": "File", "location": workflowobj["id"]})
 
     mapper = ArvPathMapper(arvrunner, files, "",
-                           keepprefix+"%s",
-                           keepprefix+"%s/%s",
+                           "keep:%s",
+                           "keep:%s/%s",
                            name=name)
 
     def setloc(p):
@@ -99,7 +126,6 @@ class Runner(object):
                                              self.tool.doc_loader,
                                              self.tool.tool,
                                              self.tool.tool["id"],
-                                             kwargs.get("keepprefix", ""),
                                              True)
 
         jobmapper = upload_dependencies(self.arvrunner,
@@ -107,7 +133,6 @@ class Runner(object):
                                         self.tool.doc_loader,
                                         self.job_order,
                                         self.job_order.get("id", "#"),
-                                        kwargs.get("keepprefix", ""),
                                         False)
 
         if "id" in self.job_order: