tqdm.write 'int' object has no attribute 'decode'

[thomasaarholt]

• I have visited the source website, and in particular
  read the known issues
• I have searched through the issue tracker for duplicates
• I have mentioned version numbers, operating system and
  environment, where applicable:

  import tqdm, sys
  print(tqdm.__version__, sys.version, sys.platform)
  4.31.1 3.7.1 | packaged by conda-forge | (default, Mar 13 2019, 12:57:14) 
  [GCC 7.3.0] linux

I think the following behaviour can at least be solved by improving the tqdm.write docs:

Whentqdm.write gets non-unicode input (I've tested with writing a list as well, same error), it throws the following error:

>>> from tqdm import tqdm
>>> tqdm.write(2)

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-14-8f12b8905b28> in <module>
      1 from tqdm import tqdm
----> 2 tqdm.write(2)

/shared/users/thomasaar/py37/lib/python3.7/site-packages/tqdm/_tqdm.py in write(cls, s, file, end, nolock)
    524         with cls.external_write_mode(file=file, nolock=nolock):
    525             # Write the message
--> 526             fp.write(s)
    527             fp.write(end)
    528 

/shared/users/thomasaar/py37/lib/python3.7/site-packages/ipykernel/iostream.py in write(self, string)
    394             # Make sure that we're handling unicode
    395             if not isinstance(string, unicode_type):
--> 396                 string = string.decode(self.encoding, 'replace')
    397 
    398             is_child = (not self._is_master_process())

AttributeError: 'int' object has no attribute 'decode'

I have viewed tqdm.write as a tqdm replacement for print that doesn't break progress bar flow, and print can handle printing non-string types. As hinted, the docs aren't extensive, and suggest that the behaviour should imitate print.

My immediate suggestion it wrap whatever argument is fed to tqdm.write in a str(), since tqdm.write(str(non_string_object)) works fine. Alternatively, one could try the decode in the error message, and if that fails use the string representation instead. I'm not familiar with how python 2.6+ would handle this, however, and I'd gladly take other suggestions.

[casperdcl]

I think we should provide a tqdm.print() which actually behaves like print(). The current write() function has confused other people too

[thomasaarholt]

That would make sense, and also not break any current behaviour with tqdm.write.

[casperdcl]

duplicate of #342