DataFrame to_csv() ignores nanRep argument for float32 input

[scottza]

import numpy as np
import pandas

d = np.random.randn(8, 4)
df = pandas.DataFrame(d)
df[1][:] = np.nan
df.to_csv('pandas-success.csv', nanRep='999')

d = np.asarray(np.random.randn(8, 4), dtype=np.float32)
df = pandas.DataFrame(d)
df[1][:] = np.nan
df.to_csv('pandas-fail.csv', nanRep='999')

After running this with pandas 0.4.1 the file pandas-success.csv
contains the nanRep string '999' as defined, but the file
pandas-fail.csv contains the string 'nan' instead.

[wesm]

Fixed in the above commit. Will be part of the forthcoming 0.4.2 release

[scottza]

Thanks!
On Oct 1, 2011 9:24 PM, "Wes McKinney" <
reply@reply.github.com>
wrote:

Fixed in the above commit. Will be part of the forthcoming 0.4.2 release

Reply to this email directly or view it on GitHub:
https://github.com/wesm/pandas/issues/182#issuecomment-2258447