BUG: (GH4626) Fix decoding based on a passed in non-default encoding in pd.read_stata

[jreback]

closes #4626

[jreback]

@jseabold can you give this a once over; I included your file as a test case, FYI

It appears to me that everything is encoded by the default cp1252, the 'encoding' parameter is merely to then decode the strings

[jreback]

cc @PKEuS

pls take a look as well

[jseabold]

One comment. It's not a showstopper, but Stata doesn't support unicode. So really the only valid encodings for reading in data are 'ascii' and 'latin-1' or equivalent. Maybe worth a mention in the docs?

[jreback]

so it's really not possible to actually encode the data (unless u somehow did it external to stata?, meaning u encoded the bytes themselves

[jseabold]

Yes, AFAIK, the user has no control over this. A great source of anguish to me lately whenever my data passes through Stata. The only extended characters supported are latin-1. Some info here (from third-party software): http://stattransfer.com/faq/encoding.html

[jreback]

ok I'll put in a doc mention