BUG: preserve left frame order in left merge

[behzadnouri]

closes #7331

on master:

>>> left dates states 0 20140101 CA 1 20140102 NY 2 20140103 CA >>> right stateid states 0 1 CA 1 2 NY >>> pd.merge(left, right, how='left', on='states', sort=False) dates states stateid 0 20140101 CA 1 1 20140103 CA 1 2 20140102 NY 2 

DataFrame.join already works fine:

>>> left.join(right.set_index('states'), on='states', how='left') dates states stateid 0 20140101 CA 1 1 20140102 NY 2 2 20140103 CA 1 

on branch:

>>> pd.merge(left, right, how='left', on='states', sort=False) dates states stateid 0 20140101 CA 1 1 20140102 NY 2 2 20140103 CA 1 >>> pd.merge(left, right, how='left', on='states', sort=True) dates states stateid 0 20140101 CA 1 1 20140103 CA 1 2 20140102 NY 2 

[jreback]

is this the only case where it needs/should be preserved? (e.g. what about other how's)

[behzadnouri]

in pandas/src/join.pyx only left_outer_join accepts this extra argument sort, which was added in this commit to preserve left frame order.

I think it makes sense, because except for the case of left join preserving order is not well defined or I do not see an immediate use for it.

[jreback]

ok then!

btw, if you are looking for something to do......

#4088 ( I did this a while back, #6647 but it doesn't quite fix this)