ncols parameter bug/inconsistency in tqdm.auto

[shun-lin]

Hi!

First I want to say this is a really really cool progress tool! I want to use tqdm to create a nicer looking progress bar for training models using keras and to make it compatible for most notebook and console I am importing tqdm using:

from tqdm.auto import tqdm

as per documentation. I found out that my progress bar is only using around half of the width in the output cell so I initialize my tqdm as

tqdm(ncols="100%")

This works great in notebook but when I tried to run it in console it breaks as ncols in console version is expecting an int instead of str. I was able to resolve the problem simply by setting dynamic_ncols to True instead of specifying ncols.

tqdm(dynamic_ncols=True)

Although my use-case is resolved (using 100% of width in notebook while having the same in console). The issue of ncols inconsistency still persists as in the notebook version of tqdm we also take in a string while standard tqdm doesn't. I just want to point this out here just in case someone also bump into the same issue. Let me know if you feel like this should be address (i.e. define behavior for passing in a string to ncols in standard tqdm) I am happy to help out!

Thank you so much for your time!

Environment:
python: 3.6.2
tqdm: 4.36.1

[casperdcl]

hmm yes, this is maybe something we should think about.

In future the arguments may diverge even more, and more frontends may be added.