from schema_salad.sourceline import SourceLine
from arvados.errors import ApiError
-from cwltool.pathmapper import PathMapper, MapperEnt, abspath, adjustFileObjs, adjustDirObjs
+from cwltool.pathmapper import PathMapper, MapperEnt
+from cwltool.utils import adjustFileObjs, adjustDirObjs
+from cwltool.stdfsaccess import abspath
from cwltool.workflow import WorkflowException
from .http import http_to_keep
if obj.get("location", "").startswith("keep:") and "listing" in obj:
del obj["listing"]
+collection_pdh_path = re.compile(r'^keep:[0-9a-f]{32}\+\d+/.+$')
+collection_pdh_pattern = re.compile(r'^keep:([0-9a-f]{32}\+\d+)(/.*)?')
+collection_uuid_pattern = re.compile(r'^keep:([a-z0-9]{5}-4zz18-[a-z0-9]{15})(/.*)?$')
class ArvPathMapper(PathMapper):
"""Convert container-local paths to and from Keep collection ids."""
- pdh_path = re.compile(r'^keep:[0-9a-f]{32}\+\d+/.+$')
- pdh_dirpath = re.compile(r'^keep:[0-9a-f]{32}\+\d+(/.*)?$')
-
def __init__(self, arvrunner, referenced_files, input_basedir,
- collection_pattern, file_pattern, name=None, single_collection=False):
+ collection_pattern, file_pattern, name=None, single_collection=False,
+ optional_deps=None):
self.arvrunner = arvrunner
self.input_basedir = input_basedir
self.collection_pattern = collection_pattern
self.referenced_files = [r["location"] for r in referenced_files]
self.single_collection = single_collection
self.pdh_to_uuid = {}
+ self.optional_deps = optional_deps or []
super(ArvPathMapper, self).__init__(referenced_files, input_basedir, None)
def visit(self, srcobj, uploadfiles):
if "#" in src:
src = src[:src.index("#")]
- if isinstance(src, basestring) and ArvPathMapper.pdh_dirpath.match(src):
- self._pathmap[src] = MapperEnt(src, self.collection_pattern % urllib.parse.unquote(src[5:]), srcobj["class"], True)
- if arvados_cwl.util.collectionUUID in srcobj:
- self.pdh_to_uuid[src.split("/", 1)[0][5:]] = srcobj[arvados_cwl.util.collectionUUID]
-
debug = logger.isEnabledFor(logging.DEBUG)
+ if isinstance(src, basestring) and src.startswith("keep:"):
+ if collection_pdh_pattern.match(src):
+ self._pathmap[src] = MapperEnt(src, self.collection_pattern % urllib.parse.unquote(src[5:]), srcobj["class"], True)
+
+ if arvados_cwl.util.collectionUUID in srcobj:
+ self.pdh_to_uuid[src.split("/", 1)[0][5:]] = srcobj[arvados_cwl.util.collectionUUID]
+ elif not collection_uuid_pattern.match(src):
+ with SourceLine(srcobj, "location", WorkflowException, debug):
+ raise WorkflowException("Invalid keep reference '%s'" % src)
+
if src not in self._pathmap:
if src.startswith("file:"):
# Local FS ref, may need to be uploaded or may be on keep
if srcobj["class"] == "Directory" and "listing" not in srcobj:
raise WorkflowException("Directory literal '%s' is missing `listing`" % src)
elif src.startswith("http:") or src.startswith("https:"):
- keepref = http_to_keep(self.arvrunner.api, self.arvrunner.project_uuid, src)
- logger.info("%s is %s", src, keepref)
- self._pathmap[src] = MapperEnt(keepref, keepref, srcobj["class"], True)
+ try:
+ keepref = http_to_keep(self.arvrunner.api, self.arvrunner.project_uuid, src)
+ logger.info("%s is %s", src, keepref)
+ self._pathmap[src] = MapperEnt(keepref, keepref, srcobj["class"], True)
+ except Exception as e:
+ logger.warning(str(e))
else:
self._pathmap[src] = MapperEnt(src, src, srcobj["class"], True)
f.write(obj["contents"])
remap.append((obj["location"], path + "/" + obj["basename"]))
else:
+ for opt in self.optional_deps:
+ if obj["location"] == opt["location"]:
+ return
raise SourceLine(obj, "location", WorkflowException).makeError("Don't know what to do with '%s'" % obj["location"])
def needs_new_collection(self, srcobj, prefix=""):
loc = srcobj["location"]
if loc.startswith("_:"):
return True
- if prefix:
- if loc != prefix+srcobj["basename"]:
- return True
+
+ i = loc.rfind("/")
+ if i > -1:
+ loc_prefix = loc[:i+1]
+ if not prefix:
+ prefix = loc_prefix
+ # quote/unquote to ensure consistent quoting
+ suffix = urllib.parse.quote(urllib.parse.unquote(loc[i+1:]), "/+@")
else:
- i = loc.rfind("/")
- if i > -1:
- prefix = loc[:i+1]
- else:
- prefix = loc+"/"
+ # no '/' found
+ loc_prefix = loc+"/"
+ prefix = loc+"/"
+ suffix = ""
+
+ if prefix != loc_prefix:
+ return True
+
+ if "basename" in srcobj and suffix != urllib.parse.quote(srcobj["basename"], "/+@"):
+ return True
+
if srcobj["class"] == "File" and loc not in self._pathmap:
return True
for s in srcobj.get("secondaryFiles", []):
if self.needs_new_collection(s, prefix):
return True
if srcobj.get("listing"):
- prefix = "%s%s/" % (prefix, srcobj["basename"])
+ prefix = "%s%s/" % (prefix, urllib.parse.quote(srcobj.get("basename", suffix), "/+@"))
for l in srcobj["listing"]:
if self.needs_new_collection(l, prefix):
return True
packed=False)
for src, ab, st in uploadfiles:
- self._pathmap[src] = MapperEnt(urllib.parse.quote(st.fn, "/:+@"), self.collection_pattern % st.fn[5:],
+ self._pathmap[src] = MapperEnt(urllib.parse.quote(st.fn, "/:+@"), urllib.parse.quote(self.collection_pattern % st.fn[5:], "/:+@"),
"Directory" if os.path.isdir(ab) else "File", True)
for srcobj in referenced_files:
ab = self.collection_pattern % c.portable_data_hash()
self._pathmap[srcobj["location"]] = MapperEnt("keep:"+c.portable_data_hash(), ab, "Directory", True)
- elif srcobj["class"] == "File" and (srcobj.get("secondaryFiles") or
- (srcobj["location"].startswith("_:") and "contents" in srcobj)):
-
- # If all secondary files/directories are located in
- # the same collection as the primary file and the
- # paths and names that are consistent with staging,
- # don't create a new collection.
- if not self.needs_new_collection(srcobj):
- continue
-
+ elif srcobj["class"] == "File" and self.needs_new_collection(srcobj):
c = arvados.collection.Collection(api_client=self.arvrunner.api,
keep_client=self.arvrunner.keep_client,
- num_retries=self.arvrunner.num_retries )
+ num_retries=self.arvrunner.num_retries)
self.addentry(srcobj, c, ".", remap)
container = arvados_cwl.util.get_current_container(self.arvrunner.api, self.arvrunner.num_retries, logger)
class StagingPathMapper(PathMapper):
+ # Note that StagingPathMapper internally maps files from target to source.
+ # Specifically, the 'self._pathmap' dict keys are the target location and the
+ # values are 'MapperEnt' named tuples from which we use the 'resolved' attribute
+ # as the file identifier. This makes it possible to map an input file to multiple
+ # target directories. The exception is for file literals, which store the contents of
+ # the file in 'MapperEnt.resolved' and are therefore still mapped from source to target.
+
_follow_dirs = True
def __init__(self, referenced_files, basedir, stagedir, separateDirs=True):
def visit(self, obj, stagedir, basedir, copy=False, staged=False):
# type: (Dict[unicode, Any], unicode, unicode, bool) -> None
loc = obj["location"]
+ stagedir = obj.get("dirname") or stagedir
tgt = os.path.join(stagedir, obj["basename"])
basetgt, baseext = os.path.splitext(tgt)
+
+ def targetExists():
+ return tgt in self.targets and ("contents" not in obj) and (self._pathmap[tgt].resolved != loc)
+ def literalTargetExists():
+ return tgt in self.targets and "contents" in obj
+
n = 1
- if tgt in self.targets and (self.reversemap(tgt)[0] != loc):
+ if targetExists() or literalTargetExists():
while tgt in self.targets:
n += 1
tgt = "%s_%i%s" % (basetgt, n, baseext)
self.targets.add(tgt)
if obj["class"] == "Directory":
if obj.get("writable"):
- self._pathmap[loc] = MapperEnt(loc, tgt, "WritableDirectory", staged)
+ self._pathmap[tgt] = MapperEnt(loc, tgt, "WritableDirectory", staged)
else:
- self._pathmap[loc] = MapperEnt(loc, tgt, "Directory", staged)
+ self._pathmap[tgt] = MapperEnt(loc, tgt, "Directory", staged)
if loc.startswith("_:") or self._follow_dirs:
self.visitlisting(obj.get("listing", []), tgt, basedir)
elif obj["class"] == "File":
- if loc in self._pathmap:
+ if tgt in self._pathmap:
return
if "contents" in obj and loc.startswith("_:"):
self._pathmap[loc] = MapperEnt(obj["contents"], tgt, "CreateFile", staged)
else:
if copy or obj.get("writable"):
- self._pathmap[loc] = MapperEnt(loc, tgt, "WritableFile", staged)
+ self._pathmap[tgt] = MapperEnt(loc, tgt, "WritableFile", staged)
else:
- self._pathmap[loc] = MapperEnt(loc, tgt, "File", staged)
+ self._pathmap[tgt] = MapperEnt(loc, tgt, "File", staged)
self.visitlisting(obj.get("secondaryFiles", []), stagedir, basedir)
+ def mapper(self, src): # type: (Text) -> MapperEnt.
+ # Overridden to maintain the use case of mapping by source (identifier) to
+ # target regardless of how the map is structured interally.
+ def getMapperEnt(src):
+ for k,v in viewitems(self._pathmap):
+ if (v.type != "CreateFile" and v.resolved == src) or (v.type == "CreateFile" and k == src):
+ return v
+
+ if u"#" in src:
+ i = src.index(u"#")
+ v = getMapperEnt(src[i:])
+ return MapperEnt(v.resolved, v.target + src[i:], v.type, v.staged)
+ return getMapperEnt(src)
+
class VwdPathMapper(StagingPathMapper):
def setup(self, referenced_files, basedir):