write() not printing

[ntomita]

While I was investigating a problem in my program where it doesn't show a progress bar, I found another problem so I report it.

I will show up when I use tqdm in REPL, try running following code.

from time import sleep
for i in tqdm(range(10)):
        sleep(1)
        tqdm.write(i)

Sure, it fails since I'm passing an integer. I fix it and run it again,

for i in tqdm(range(10)):
        sleep(1)
        tqdm.write(str(i))

Now it shows just a progress bar, no numbers printed at all.
Empirically speaking It will come back after I do tqdm().clear() and run some extra loops, like in this way...

... 
                                                                                      
                                                                                      
                                                                                      
  0%|                                                          | 0/10 [01:19<?, ?it/s]
3                                                                                     
4                                                                                     
5                                                                                     
6                                                                                     
7                                                                                     
8                                                                                     
9                                                                                     
100%|█████████████████████████████████████████████████| 10/10 [00:10<00:00,  1.00s/it]
>>> 

So, I'm assuming something wrong with buffer handling. Have you ever observed this issue?
I'm using Python3 on mac with newest OS. tqdm have installed via pip3. I use standard terminal for REPL.

[casperdcl]

can't reproduce this:

C:\> tqdm --version
4.11.2
C:\> python
Python 2.7.13 |Anaconda 4.2.0 (64-bit)| (default, Dec 19 2016, 13:29:36) [MSC v.
1500 64 bit (AMD64)] on win32

>>> from time import sleep
>>> from tqdm import tqdm, trange
>>> for i in trange(10):
...   sleep(1)
...   tqdm.write(str(i))
...
0
1
2
3
4
5
6
7
8
9
100%|##########################################| 10/10 [00:10<00:00,  1.00s/it]
>>>

What OS and terminal are you using?

[ntomita]

Thanks for the reply!

I'm using Python3 on mac with newest OS. tqdm have installed via pip3. I use standard terminal for REPL.

More specifically, I use version 2.7.1 of Terminal app, OS is macOS Sierra ver 10.12.2.
Also, I use Python 3.5.2 with tqdm 4.11.2

You need to fail inside a loop at least once in order to reproduce the behavior. That occurs when you keep using tqdm with second loop after you got some error from first loop, like I did above.

[casperdcl]

ah. ok. so the problem is you write invalid code (writing a non-string). I don't think tqdm (or any other library) should silently work properly in such a case. we could cast all input of tqdm.write to a string but that way lies dragons.

[ntomita]

I see. Ya, as you say automatically converting the type of input is obviously a bad idea. Instead, I think the module should be re-initialized when it encounters an exception. I never had such issue when using other modules that's why I reported this issue. I think it is still beneficial to users, but I understand the policy you have.
Anyway, you guys implements an awesome library, many thanks!

[CentralLT]

@ntomita let's close the issue. I think your question / problem has been resolved.

[casperdcl]

I think the module should be re-initialized when it encounters an exception

this might now be the case but I haven't checked. I recall fixing exceptions within iterations but not sure about when calling write().