TST: Tests and fix for unhandled data types in Excel writers.

[jmcnamara]

Added tests and a fix for unhandled numpy data types in the
Excel writers. Issue #3122.

[jmcnamara]

Some background to this PR:

1. The original issue ExcelWriter fails with columns with int32 dtype #3122 is due to the fact that xlwt whitelists the types that it handles and raises an exception for unhandled types. The fact that certain numpy types aren't identified correctly may be a numpy bug but their devs have rejected that as an issue.
2. The pandas/io/excel.py module contains a _conv_value() method to map pandas/numpy types to types handled by xlwt/openpyxl.
3. This PR extends/strengthens the _conv_value() function to handle more generic types.
4. This PR adds more extensive, although non-exhaustive, tests for numpy data types.
5. XlwsWriter doesn't require the _conv_value() method since it has its own type conversion methods. So I've removed that from the xlsxwriter engine.

[jreback]

@jtratner ?

[jtratner]

Going to take a look tonight or tomorrow.
On Oct 7, 2013 7:36 PM, "jreback" [email protected] wrote:

@jtratner https://github.com/jtratner ?

—
Reply to this email directly or view it on GitHubhttps://github.com//pull/5122#issuecomment-25854366
.

[jreback]

@jtratner this ok for merging?

[jtratner]

Looks good.

[jtratner]

Thanks for fixing this!

[jtratner]

And 👍 for not doing _conv_value where it's not needed.