mrocklin / dask / 1298

94%

master: 94%

Build Type

push

travis-ci

Committed by Matthew Rocklin

Commit Message

Avoid shuffles when merging with Pandas objects

Previously dask.dataframe would convert all Pandas objects into
dask.dataframe objects.  In some cases this causes unnecessary shuffles.

In particular when you join a non-index column of a large dask.dataframe
against a small pandas table the right thing to do is to just move the
small pandas dataframe to all tasks, turning the computation into an
embarassingly parallel one.

Run Details

27 of 27 new or added lines in 1 file covered. (100.0%)

8453 of 9035 relevant lines covered (93.56%)

0.94 hits per line