ordered_merge improvements

[wesm]

http://stackoverflow.com/questions/12322289/kdb-like-asof-join-for-timeseries-data-in-pandas/12336039#12336039

[leonbaum]

It would be great to be able to perform ordered left joins using a time-series index as well.

[yarivm]

+1 vote for "ordered left joins."

[ngcrawford]

+1 vote for "ordered left joins.

[chrisaycock]

Any thoughts on this? A kdb-style asof-join is very useful in time-series analysis. The rows of the right-hand table would just be picked via a DataFrame.asof (see #2941).

[jreback]

http://pandas.pydata.org/pandas-docs/stable/merging.html?highlight=ordered_merge#merging-ordered-data

(unless something else was meant here). This has been in for a while.

[chrisaycock]

@jreback Unlike the pd.ordered_merge(), the kdb aj() only keeps the rows from the left-hand table.

http://code.kx.com/wiki/Reference/aj

I.e., as with a regular left-join, the table doesn't get more rows; it just appends columns from the right-hand table.

Actually, it's very similar to a left-join, but instead of searching for equal keys, it searches for the greatest key from the right-hand table that is less-than-or-equal-to the key from the left-hand table. This is exactly what the asof function does in pandas, so we're part of the way there already.