Stephen Karl Larroque

I stop here, I'll add unit tests if enough people are interested. (Adding unit tests shouldn't be too hard because we can reuse the tests of core tqdm but some...

I did some benchmark against different versions of tqdm: ``` bare: 100%|#################|100000000/100000000 [00:20

I pushed a class-based version of tqdm_bare the [branch tqdm_bare_class](https://github.com/tqdm/tqdm/tree/tqdm_bare_class), and did a few additional optimizations. Here are the results (manually averaged over 10 runs): ``` bare: 100%|#################|100000000/100000000 [00:18

For reference, core tqdm is currently 954 lines and 829 SLOC, so this is a drastic reduction of SLOC while keeping most features. This should greatly ease the porting to...

Ah sorry guys I didn't see your replies before posting my latest comments. @CrazyPython Thank you, that was more work than I intended to put on this issue :) However...

Here are the results without miniters tuning: ``` bare 2123341 v1.0 2075808 noclass 5175179 class 4942420 core 4427913 ``` Clearly there's a bug with bare calculation of dynamic_miniters. I'm going...

Ah no bug in fact, it's because I set smoothing=0 only for tqdm_bare (I can't remember why) in my benchmark, which disables dynamic_miniters. The same issue happens with core tqdm....

It's quite interesting to note that while `class` was the slowest with fixed miniters, it's the second fastest with dynamic_miniters.

@CrazyPython Fastest bar is original tqdm v1.0 with tuned miniters (because we basically skip time check, I think that setting mininterval=0 would give just the same result), with noclass being...

Ah! Interesting finding, while I was fixing the dynamic_miniters not working with smoothing=0 bug (I will merge this fix with #250), I found that the old way of calculating dynamic_miniters...