Remove control character stripping.

[uri-canva]

The logic to strip away ansi control sequences only works if they're at the end of the line, if the control sequences are at the beginning of the line the whole line will be stripped away, and all progress information will be lost.

[Byron]

strip() will not mutate the string. Thus it should be sline = sline.strip().
If it is important, do you think it's possible to have a test for it? Otherwise, maybe that line is not needed at all.

[Byron]

I just saw that the line stripping was done previously as well, let's just re-add it then as in sline = sline.rstrip(). Apparently there is no test that would need it, and it's tech-debt I wouldn't ask you to pay either 😅.

[uri-canva]

Absolutely, good catch!

[Byron]

(Sorry, hit one-line comment by accident)

Thanks a lot for your contribution! I would love to merge it as soon as possible!

[codecov-io]

Codecov Report

Merging #926 into master will not change coverage.
The diff coverage is n/a.

@@ Coverage Diff @@## master #926 +/- ##
=======================================
Coverage 93.62% 93.62% =======================================
Files 59 59 Lines 9800 9800 =======================================
Hits 9175 9175 Misses 625 625

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 9121f62...7a5de47. Read the comment docs.

[uri-canva]

Hmm, tests are failing, it is a bit different in behaviour, before if there was abc<control sequence>def the resulting string would be abc, now it's abcdef. Will have to test it further.

[uri-canva]

Seems to have been a bad rebase on my part, now that it's rebased on master after the other PRs were merged it's working fine.

[Byron]

It looks like the strip() call is not yet reassigned, but looks great otherwise!
Thanks so much for your efforts!

[Byron]

Thank you! Once line = line.strip() is back I can merge it.

[uri-canva]

Sorry, I must have dropped the change during the rebasing. Because the tests only work on master I lose track of where the commits are sometimes. 😆

[uri-canva]

Ah it's the stripping that breaks the tests.

[uri-canva]

Changed it back to rstrip instead. I don't really understand what is going on, it works on my machine both in the macos host and the linux docker image with strip, so I couldn't debug it. I guess leaving it as it is is the safest option.

[Byron]

Thanks so much for your continued efforts, it's much appreciated :)!

[uri-canva]

Thank you for the wonderful library!

[Byron]

About 10 years later, I am not quite sure if it's wonderful anymore ;).

Somehow now it breaks on master in a spot that seems unrelated. Fascinating. Could it be the difference between strip() and rstrip()?

[uri-canva]

It does seem completely unrelated: it doesn't have anything to do with progress. I understand if it would break tests related to progress. But anyway definitely revert it if it's the problem!

[Byron]

It looks like the rstrip() call was the problem. Probably it should worry me that I have no idea how this could affect that particular test :/.
The conservative voice tells me to roll back the commit entirely, but prefer to trust the respective tests and rather add additional ones as needed.

[uri-canva]

Yeah I think rolling back makes sense, then figure out how to repro the failure on the PR with a new test, then fix it. Without a new test that fails on the PR there's no way to tell if the problem will reappear after the PR is merged.

[Byron]

Thanks, @uri-canva , the respective commit was reverted.

I agree, probably it would be beneficial to have a test for the initial motivation of this PR, i.e. the code removing too much of the line if the escape sequence appears in an unexpected spot, and then provide a fix.

[uri-canva]

Oh yes, that too. I meant it mostly the other way around: none of the tests failed on this PR, but then some tests failed on master, so it would be good for that hole to be filled.