This data as json
html_url | id | node_id | tag_name | target_commitish | name | draft | author | prerelease | created_at | published_at | assets | body | repo |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|
https://github.com/simonw/csvs-to-sqlite/releases/tag/0.5 | 8575785 | MDc6UmVsZWFzZTg1NzU3ODU= | 0.5 | master | csvs-to-sqlite 0.5 | 0 | 9599 | 0 | 2017-11-19T05:11:27Z | 2017-11-19T05:53:25Z | [] | ## Now handles columns with integers and nulls in correctly Pandas does a good job of figuring out which SQLite column types should be used for a DataFrame - with one exception: due to a limitation of NumPy it treats columns containing a mixture of integers and NaN (blank values) as being of type float64, which means they end up as REAL columns in SQLite. http://pandas.pydata.org/pandas-docs/stable/gotchas.html#support-for-integer-na To fix this, we now check to see if a float64 column actually consists solely of NaN and integer-valued floats (checked using v.is_integer() in Python). If that is the case, we over-ride the column type to be INTEGER instead. See #5 - also a8ab524 and 0997b7b | 110509816 |