Refactor offline

This refactoring includes some slicing and performance improvement attempts.

The slicing is still experimental and has some flaws:

• fancy indexing over thousands of elements might be horrible slow, eg f.events.tracks[idx_list, 0] where idx_list is a huge list of indices
• trivial problem of iterating: for track in f.evens.tracks does not stop after the last element
• meaningless indexing: f.events[5][23][420] is simply f.events[420] since the index/slice is not chained but overwritten when propagated. We need custom checks whether it makes sense
  • events are flat lists, so [x][y] does not make sense, it should throw an error
  • f.events[5:23][4] should effectively be the same as f.events[9] due to chaining of slice/index

km3io/tools.py

@@ -19,6 +19,18 @@ class cached_property:
         return prop
 
 
+def _unfold_indices(obj, indices):
+    """Unfolds an index chain and returns the corresponding item"""
+    for depth, idx in enumerate(indices):
+        try:
+            obj = obj[idx]
+        except IndexError:
+            print("IndexError while accessing item '{}' at depth {} ({}) of "
+                  "the index chain {}".format(repr(obj), depth, idx, indices))
+            raise
+    return obj
+
+
 BranchMapper = namedtuple(
     "BranchMapper",
     ['name', 'key', 'extra', 'exclude', 'update', 'attrparser', 'flat'])