API: consistent float_format specification

[jorisvandenbossche]

For now just some observations:

• The docstring of different funtions about float_format:
  • the documentation of to_csv (as well as to_excel) states: string, Format string for floating point numbers
    • Example of this usage: df.to_csv(float_format='%.3f')
  • The docstring of to_string (as well as to_html) says: one-parameter function, Formatter function to apply to columns’ elements if they are floats
    • Example of this usage: df.to_string(float_format='{:.3f}'.format)
  • The option display.float_format also accepts a callable following the docs
• The narrative docs are consistent with that:
  • http://pandas.pydata.org/pandas-docs/stable/io.html#writing-to-csv-format
  • http://pandas.pydata.org/pandas-docs/stable/io.html#writing-a-formatted-string

Question:

• Should we make this consistent? Eg that all methods can accept both a format string as a format function?
• Should we make the usage of '{:.3f}'.format) more convenient?

There are a lot of different issues already about formatting, but I didn't find anything about this in specific.

[jorisvandenbossche]

Aha, related issue: #2502 (about the format of the format string), but this was closed in favor of #4668

[jreback]

put he reference on #4668

[davetapley]

This is probably a separate issue, but:

There are a lot of different issues already about formatting

---

Just to be sure: there is still no way to provide a per column float_format, as styler offers?

If so, and if we're going to shoot for consistent float_format, then might I suggest it also aligns with styler?

---

See also:

• https://stackoverflow.com/questions/44789358/python-pandas-dataframe-error-when-export-style-object-to-csv
• https://stackoverflow.com/questions/13999850/how-to-specify-date-format-when-using-pandas-to-csv#comment55059146_22798849