set_description takes no effect until .update() is called

[gwerbin]

Expected behavior: set_description() takes effect immediately when called

Actual behavior: set_description() takes effect only when .update() is called.

Steps to reproduce

import time

with tqdm(total=10, miniters=1, mininterval=0) as progress:
    for i in range(10):
        progress.set_description("step " + str(i + 1))
        time.sleep(0.5)
        progress.update()

Python version 3.5.2, TQDM version 4.10.0-10bd71

[lrq3000]

Hello @gwerbin,

You can use progress.refresh() to force the display refresh. Maybe we should force the refresh inside set_description(), we didn't give this much of a thought to be honest. The only thing is that refresh() is expensive because it then forces your bar display twice per iteration...

What you could rather do to be more computationally effective while having the same result would be to set the first iteration's description at tqdm initialization and move the call to set_description() to the end of the loop:

import time

with tqdm(total=10, miniters=1, mininterval=0, desc="step 0") as progress:
    for i in range(10):
        time.sleep(0.5)
        progress.set_description("step " + str(i + 1))
        progress.update()

[gwerbin]

@lrq3000 thanks, that makes sense. Maybe we can just call this a documentation issue then. There should be a "TQDM Cookbook" if there isn't already.

[ericfrederich]

I would consider this a bug. Just add a call to update() or refresh() within set_description().

It's interesting to note that the docs have this snippet of code:

pbar = tqdm(["a", "b", "c", "d"])
for char in pbar:
    pbar.set_description("Processing %s" % char)

You'll never see "Processing a", so it should say "Processed" instead of "Processing"

[olalonde]

This seems like a bug.. .refresh() doesn't fix the issue here.

[ChrisPVille]

I've also run into this issue. I've moved the progress bar into its own worker thread, and I still can't seem to get the postfix to update with .refresh() or .update(0). It usually does draw correctly after .update(1) or larger, but this obviously skews the results upward.

[ericfrederich]

Thanks @casperdcl ... I was surprised to see how far back this issue went. I agree though, perhaps refresh should be set to False