The datashape docs are particularly sparse when it comes to documenting discovery. In particular, it would be useful to describe how discovery works for the various backends. e.g., for Pandas dataframes, discover is conservative, and since pandas dataframes with string columns could, at any time, mutate to have None entries, it always reports those columns as ?string.
The datashape docs are particularly sparse when it comes to documenting discovery. In particular, it would be useful to describe how discovery works for the various backends. e.g., for Pandas dataframes,
discover
is conservative, and since pandas dataframes with string columns could, at any time, mutate to have None entries, it always reports those columns as?string
.