Add left_by/right_by and left_index/right_index to merge_asof()

[chrisaycock]

The current merge_asof() has only one parameter for by, assuming that both DataFrames have the same column name. It does have left_on and right_on, so perhaps something similar for by would be helpful.

Also, there is no left_index or right_index parameter in merge_asof(). Adding these might be helpful too.

For example:

pd.merge_asof(df1, df2,
              left_index=True, right_on='timestamp',
              left_by='Ticker', right_by='ticker')

[jreback]

@chrisaycock can you show an example where they are meaningful?

I did have them initially, but I couldn't come up with a good use case. In that the merge-asof is quite restricted in what you are doing with it and the API area pretty much covers it.

[chrisaycock]

The example I'm working through now is that my left table is indexed by timestamp and has a column called 'Ticker', whereas my right table has columns called 'timestamp' and 'ticker'. These are passed in, so I don't have control over what I get. I have to reshape the tables temporarily before calling merge_asof(), and then I have to undo the reshaping in the returned result.

[jreback]

@chrisaycock ahh I c. ok, can you update the topsection with an example of using that API then?

[chrisaycock]

@jreback Done.