Encoding fails when output is longer than input about python-iconv HOT 6 CLOSED

I submitted a pull request that seems to fix the issue, along with a few other problems I noticed while messing around with it. I think the main problem was that the indexing and retries didn't change when the encode() implementation was changed to use byte strings internally.

from python-iconv.

Thank you for reporting this issue.
I cannot fix this quickly though.

When running locally with python 3.9, I cannot even execute the existing unit tests successfully.
Loading the ASCII//TRANSLIT codec is not possible at all there. Though iconv on the command line works fine with it.
(So I’m not able to reproduce the problem right now ^^)
I’ll probably have to do some tests with pyenv / asdf and different python versions.
The version I originally used was probably 3.7.
Maybe there was some change in the c-api, but I can’t be sure.

If you could contribute something that would of course be highly appreciated.
Setting up a local dev environment is as simple as:

python -m venv env
source env/bin/activate
pip install -e .
python -m unittest

from python-iconv.

I'll see if I can find time to mess around with it in the next few days but no promises. I've never needed to use the buffer protocol before and after a quick glance at the documentation, I'm not overly optimistic that I'll be able to master it in a short timeframe. In particular, I didn't see an obvious way to translate the number of bytes converted into the number of characters converted. Maybe it's obvious to someone who has used that API before; I'll tinker a little more.

from python-iconv.

Haha, auto-closed on merge :-D
Thank you so much for your contribution.

Do you have other specific example strings that caused errors previously?
I would like to add some more unit tests.

from python-iconv.

I published v1.1.1 with the fix to pypi.

from python-iconv.

The original bug was in cases where the output string was longer than the input string. The easiest example is converting any string containing "™" (U+2122) from unicode (where it is 3 bytes) to ASCII//TRANSLIT, where it is rendered as "(TM)" (4 bytes). My other tinkering fixed bugs when errors was 'replace' or 'ignore'. I think any conversion that triggers those should do.

from python-iconv.