Make SavePrimaryDB faster

We avoid reading in the entire file by just reading the first
byte. Also, this reduces the amount of waiting between polls, so small
test runs complete noticeably faster.

As a side effect of streaming the file, we have to prevent the use of
toil caching. (readGlobalFileStream doesn't have a cache=False option.)

src/cactus/pipeline/cactus_workflow.py

@@ -300,11 +300,11 @@ def run(self, fileStore):
         stopKtserver(dbElem)
         # Wait for the file to appear in the right place. This may take a while
         while True:
-            path = fileStore.readGlobalFile(self.cactusWorkflowArguments.snapshotID, cache=False)
-            stat = os.stat(path)
-            if stat.st_size > 0:
-                break
-            time.sleep(60)
+            with fileStore.readGlobalFileStream(self.cactusWorkflowArguments.snapshotID) as f:
+                if f.read(1) != '':
+                    # The file is no longer empty
+                    break
+            time.sleep(10)
         # We have the file now
         intermediateResultsUrl = getattr(self.cactusWorkflowArguments, 'intermediateResultsUrl', None)
         if intermediateResultsUrl is not None:

src/cactus/progressive/cactus_progressive.py

@@ -442,7 +442,12 @@ def main():
     setLoggingFromOptions(options)
 
     options.cactusDir = os.path.abspath(options.cactusDir)
-
+
+    # Caching generally slows down the cactus workflow, plus some
+    # methods like readGlobalFileStream don't support forced
+    # reads directly from the job store rather than from cache.
+    options.disableCaching = True
+
     #Create the progressive cactus project 
     projWrapper = ProjectWrapper(options)
     projWrapper.writeXml()