Changes between Version 3 and Version 4 of Ticket #28444, comment 58


Ignore:
Timestamp:
09/06/19 06:59:27 (2 years ago)
Author:
SimonKing
Comment:

Legend:

Unmodified
Added
Removed
Modified
  • Ticket #28444, comment 58

    v3 v4  
    1414> If you want to make things work with 'bytes' encoding *as well as* 'latin1' encoding, I think you're just making extra work for yourself, and if you're making the infrastructure more lax about input types, I think you can end up with masking more errors in the end.
    1515
    16 Sure. That's why I said it is a proof of concept. I show that making `bytes_to_str` and `str_to_bytes` more tolerant about input and using it more commonly in fact is a way to unpickle with `encoding='bytes'`, showing that with some work there ''is'' a chance to unpickle complex data structures involving numpy arrays, without patching arrays.
     16Sure. That's why I said it is a proof of concept. I show that making `bytes_to_str` and `str_to_bytes` more tolerant about input and using it more commonly in fact is a way to unpickle with `encoding='bytes'`, showing that with some work there ''is'' a chance to unpickle complex data structures involving numpy arrays, without patching numpy.
    1717
    1818But probably, it would be easier to just drop the new commit, and then provide some unpickle override for numpy arrays only.