Chunkstore performance enhancements (pandas-dev#182)

- new serializer for ChunkStore
- supports by column serialization
- Significantly faster than the record serializer for this use case
- Supports DataFrames and Series only
- Changes to chunker that boost performance
- Ability to read subset of columns
- Also fixes pandas-dev#164

CHANGES.md

@@ -1,5 +1,10 @@
 ## Changelog
 
+### 1.27
+
+  * Bugfix: #187 Compatibility with latest version of pytest-dbfixtures
+  * Feature: #182 Improve ChunkStore read/write performance
+
 ### 1.26 (2016-07-20)
 
   * Bugfix: Faster TickStore querying for multiple symbols simultaneously

arctic/chunkstore/_chunker.py

@@ -1,3 +1,7 @@
+START = 's'
+END = 'e'
+
+
 class Chunker(object):
 
     def to_chunks(self, data, *args, **kwargs):
@@ -60,3 +64,15 @@ def exclude(self, data, range_obj):
         data, filtered by range_obj
         """
         raise NotImplementedError
+
+    def chunk_to_str(self, chunk_id):
+        """
+        Converts parts of a chunk range (start or end) to a string. These
+        chunk ids/indexes/markers are produced by to_chunks.
+        (See to_chunks)
+
+        returns
+        -------
+        string
+        """
+        raise NotImplementedError