to_csv locale dependent float_format

[Winand]

Python (3.4) doesn't support "%n"%9.0, have to use "{:n}".format(9.0) instead.
But pandas (0.17.0) to_native_types uses 1st type of string formatting, so locale dependent format cannot be used with to_csv .

[jreback]

can you show an example of where this is failing?

[Winand]

I set locale.setlocale(locale.LC_NUMERIC, '') and want floats to use local decimal point symbol. This can be achieved with "{:n}".format(value), but pandas uses %-formatting for strings (inside to_native_types function) which doesn't support %n format (as of Python 3.4 x64 Win 7)

This works:
df.to_csv(buf, sep='\t', header=header, index=False, float_format="%g")

These don't:
df.to_csv(buf, sep='\t', header=header, index=False, float_format="%n") (Python %-formatting doesn't know %n format)
df.to_csv(buf, sep='\t', header=header, index=False, float_format="{:n}") (pandas uses %-formatting)

P.S. maybe this refers to Python issues and this issue should be closed?

[jreback]

@Winand I think we should simply change to the new style formatting here.

Want to do a pull-request?

[Winand]

#12132 (though some things changed in 9302811)

[SeeminSyed]

Should this be closed as a result of work on #12132?

[Winand]

Should this be closed as a result of work on #12132?

I've never finished that PR, and now it needs to be revised, because those parts of pandas have changed significantly. But it still uses old-style formatting, afaik.