DataFrame -> ColumnDataSource erases non-trivial index

[mrocklin]

By default we ignore the index. We should probably only do this if it's the trivial ordering 0, 1, 2, ... and treat it as data otherwise.

cc @brittainhard

[femtotrader]

Support for datetime index (as primary key / unique) will be great.

Here is a sample.

You might install pandas_datareader

$ pip install pandas_datareader

Here is code:

import datetime
import pandas as pd
from pandas_datareader import data, wb

import blaze as bz
from blaze import odo

symbol = "AAPL"
df = data.DataReader(symbol, 'yahoo', datetime.datetime(2011, 1, 1), datetime.datetime(2015, 1, 1))
table_uri = 'dialect+driver://user:[email protected]:port/symbols::%s' % symbol
odo(df, table_uri)

[jdmcbr]

I just ran into this issue when converting a dataframe to a table in a SQL database. Rather than checking for a non-trivial index, what about checking if the index is named or not?

[jdmcbr]

My first thought was to set index=True in the appropriate places in odo.convert when there is a non-trivial index. However, I'm not seeing a way to have datashape.discover recognize an index. Calling reset_index in cases of a non-trivial index doesn't feel right. Any thoughts?