10576: Working on fetch & url join for keep references.

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

diff --git a/sdk/cwl/arvados_cwl/fsaccess.py b/sdk/cwl/arvados_cwl/fsaccess.py
index e44e7a928218c04e377b04373d7ec44ea59b04a0..9d3fe1c1ba7b699f3491a3b69b30750051aadd33 100644 (file)
--- a/sdk/cwl/arvados_cwl/fsaccess.py
+++ b/sdk/cwl/arvados_cwl/fsaccess.py
@@ -1,6 +1,7 @@
 import fnmatch
 import os
 import errno
+import urlparse
 
 import cwltool.stdfsaccess
 from cwltool.pathmapper import abspath
@@ -9,12 +10,15 @@ import arvados.util
 import arvados.collection
 import arvados.arvfile
 
+from schema_salad.ref_resolver import DefaultFetcher
+
 class CollectionFsAccess(cwltool.stdfsaccess.StdFsAccess):
     """Implement the cwltool FsAccess interface for Arvados Collections."""
 
-    def __init__(self, basedir, api_client=None):
+    def __init__(self, basedir, api_client=None, keep_client=None):
         super(CollectionFsAccess, self).__init__(basedir)
         self.api_client = api_client
+        self.keep_client = keep_client
         self.collections = {}
 
     def get_collection(self, path):
@@ -22,7 +26,8 @@ class CollectionFsAccess(cwltool.stdfsaccess.StdFsAccess):
         if p[0].startswith("keep:") and arvados.util.keep_locator_pattern.match(p[0][5:]):
             pdh = p[0][5:]
             if pdh not in self.collections:
-                self.collections[pdh] = arvados.collection.CollectionReader(pdh, api_client=self.api_client)
+                self.collections[pdh] = arvados.collection.CollectionReader(pdh, api_client=self.api_client,
+                                                                            keep_client=self.keep_client)
             return (self.collections[pdh], "/".join(p[1:]))
         else:
             return (None, path)
@@ -118,3 +123,51 @@ class CollectionFsAccess(cwltool.stdfsaccess.StdFsAccess):
             return path
         else:
             return os.path.realpath(path)
+
+class CollectionFetcher(DefaultFetcher):
+    def __init__(self, cache, session, api_client=None, keep_client=None):
+        super(CollectionFetcher, self).__init__(cache, session)
+        self.fsaccess = CollectionFsAccess("", api_client=api_client, keep_client=keep_client)
+
+    def fetch_text(self, url):
+        if url.startswith("keep:"):
+            with self.fsaccess.open(url) as f:
+                return f.read()
+        return super(CollectionFetcher, self).fetch_text(url)
+
+    def check_exists(self, url):
+        if url.startswith("keep:"):
+            return self.fsaccess.exists(url)
+        return super(CollectionFetcher, self).check_exists(url)
+
+    def urljoin(self, base_url, url):
+        if not url:
+            return base_url
+
+        urlsp = urlparse.urlsplit(url)
+        if urlsp.scheme:
+            return url
+
+        basesp = urlparse.urlsplit(base_url)
+        if basesp.scheme == "keep":
+            if not basesp.path:
+                raise IOError(errno.EINVAL, "Invalid Keep locator", base_url)
+
+            baseparts = basesp.path.split("/")
+            urlparts = urlsp.path.split("/")
+
+            pdh = baseparts.pop(0)
+
+            if not arvados.util.keep_locator_pattern.match(pdh):
+                raise IOError(errno.EINVAL, "Invalid Keep locator", base_url)
+
+            if urlsp.path.startswith("/"):
+                baseparts = []
+
+            if baseparts and urlparts:
+                baseparts.pop()
+
+            path = "/".join([pdh] + baseparts + urlparts)
+            return urlparse.urlunsplit(("keep", "", path, "", urlsp.fragment))
+
+        return super(CollectionFetcher, self).urljoin(base_url, url)