Use parsed date column as index

[changhiskhan]

Right now date conversion is post-processing so the index has already been constructed by then. We need to make it so you can use the newly generated columns as the index

(maybe as simple as a call to set_index)?

[timmie]

A question regarding usage:
The following works well:

df_in = pa.read_table(sample_file, delimiter=';',
                     date_parser=YMDT,
                     index_col = 'Year_Month_Day_Time', 
                     parse_dates=[[0, 1, 2, 3]])

Why not parse_dates_as_index_col?
The user does not know the future name such as 'Year_Month_Day_Time'....

[changhiskhan]

you can do the following without knowing the column name:

df_in = pa.read_table(sample_file, delimiter=';',
date_parser=YMDT,
index_col = 0,
parse_dates=[[0, 1, 2, 3]])

or

df_in = pa.read_table(sample_file, delimiter=';',
date_parser=YMDT,
index_col = 'dt',
parse_dates={'dt' : [0, 1, 2, 3]})

parse_dates_as_index_col would be slightly more convenient. Maybe we can just have it be syntactic sugar that's equivalent to the above.

[timmie]

Let's add your hint to the docs.

Maybe we can just have it be syntactic sugar that's equivalent to the above.

This would be nice to have in case (mostly actually) there's only one cloumn with dates.

[changhiskhan]

Would be very nice to have. Pull requests would be more than welcome :)

[changhiskhan]

Here's how you can contribute to the documentation:

http://pandas.pydata.org/developers.html

#1355 #736