Handle European decimal formats in to_csv

[adamklein]

Counterpart to issue #584

[wesm]

Need to build a low-level CSV-emitter at some point. tabling this to future release

[ghost]

closed by #584, 2 years ago.

[bertrandhaut]

Should this issue really be closed?

If I'm not wrong, the support of "," as decimal separtor is now (=pandas 0.14) only supported in "read_csv" and not in "to_csv".

DataFrame.to_csv only supports the float_format argument which does not allow to specify a particular decimal separtor.

[jreback]

well float format can be a function so you can do anything u want

that said

this issue was closed several years ago
if u would like to request this feature pls open a new issue with an example use case

pull requests are welcome.

[jorisvandenbossche]

Isn't the feature request quite clear? Just to be able to specify decimal in to_csv like it is available in read_csv. Eg:

df.to_csv('test.csv', sep=';', decimal=',')

The linked issues which would have closed this, did not yet actually close it, as that handled about the parser, not to write.

[bertrandhaut]

Indeed Joris has my point. For me the closed linked issue does not answer to the initial demand stated in this particular issue.

Concerning the ability of the float_format to handle a function, it seems a little too complex.

If I do a

def commaFloat(x):
    return str(x)

df.to_csv(fileName, sep=';', float_format=commaFloat() )

I get

TypeError: unsupported operand type(s) for %: 'function' and 'float'

I have to do the following in order to have it working

class commFloat:
    def __mod__(self, x):
        return str(x).replace('.',',')

Not very intuitive

[bertrandhaut]

I can try implement this feature.

Does anybody has a better proposal that just adding a:

if decimal:
       value.replace('.', decimal, 1)

after the application of the float_format function ?
Something more efficient?

[jreback]

start with adding the keyword to the top fo the call stack: https://github.com/pydata/pandas/blob/master/pandas/core/frame.py#L1071

(then need to propogate downwards)

add ulitamtely to FloatArrayFormatter. I belive you can do this all at once (its an array of float values). so no need to do indivdually. But I think the right idea.

[bertrandhaut]

I've created a pull request for this issue: #8448

It's my first contribution to pandas and first use of git: feel free to comment.

[jreback]

closed by #8448