Merge pull request #392 from scikit-hep/optimize-remote-chunking

Set chunkbytes to 1 MB and limitbytes to 100 MB, based on @chrisburr's tests.

uproot/source/http.py

@@ -22,7 +22,7 @@ def __init__(self, path, auth=None, *args, **kwds):
         self._size = None
         self.auth = auth
 
-    defaults = {"chunkbytes": 32*1024, "limitbytes": 32*1024**2, "parallel": 8*multiprocessing.cpu_count() if sys.version_info[0] > 2 else 1}
+    defaults = {"chunkbytes": 1024**2, "limitbytes": 100*1024**2, "parallel": 8*multiprocessing.cpu_count() if sys.version_info[0] > 2 else 1}
 
     def _open(self):
         try:

uproot/source/xrootd.py

@@ -20,7 +20,7 @@ def __init__(self, path, timeout=None, *args, **kwds):
         self.timeout = timeout
         super(XRootDSource, self).__init__(path, *args, **kwds)
 
-    defaults = {"timeout": None, "chunkbytes": 32*1024, "limitbytes": 32*1024**2, "parallel": False}
+    defaults = {"timeout": None, "chunkbytes": 1024**2, "limitbytes": 100*1024**2, "parallel": False}
 
     def _open(self):
         try:

uproot/tree.py

@@ -1140,7 +1140,7 @@ def basket_uncompressedbytes(self, i, keycache=None):
             self._tryrecover()
         return self._threadsafe_key(i, keycache, False)._fObjlen
 
-    def basket_compressedbytes(self, i):
+    def basket_compressedbytes(self, i, keycache=None):
         if self._recoveredbaskets is None:
             self._tryrecover()
         key = self._threadsafe_key(i, keycache, False)
@@ -2023,7 +2023,7 @@ def daskframe(path, treepath, branches=None, namedecode="utf-8", entrysteps=floa
 
 ################################################################ for quickly getting numentries
 
-def numentries(path, treepath, total=True, localsource=MemmapSource.defaults, xrootdsource=XRootDSource.defaults, httpsource=HTTPSource.defaults, executor=None, blocking=True, **options):
+def numentries(path, treepath, total=True, localsource=MemmapSource.defaults, xrootdsource={"timeout": None, "chunkbytes": 32*1024, "limitbytes": 1024**2, "parallel": False}, httpsource={"chunkbytes": 32*1024, "limitbytes": 1024**2, "parallel": False}, executor=None, blocking=True, **options):
     if isinstance(path, string_types):
         paths = _filename_explode(path)
     else: