Ability to see errors in editor

[chadrik]

I had a quick play with this plugin and it's an improvement over using a shell, but it's not very tightly integrated into the editor. It would be great to see a bit of the approach used by https://github.com/leinardi/mypy-pycharm integrated here. That plugin looks like it is not actively maintained any longer, and just getting it to detect the mypy executable is a huge PITA (for some reason it's trying to use my system's python2 interpreter), but the big thing it has going for it is the ability to see errors in the editor, which is pretty slick. It gives you a bi-directional relationship between editor view and error view, whereas with this plugin it seems I have only a uni-directional relationship from error-to-editor.

[ilevkivskyi]

I actually thought about this. But this may be misleading, because the code is not re-checked immediately when you edit it. So it will be underlined as an error even after you fix the error.

In other words, this issue and #19 should go together. Btw, now that we have 300-500ms mypy daemon response time (even for huge code bases), this looks like a reasonable feature to have.

[chadrik]

Hi Ivan!

Btw, now that we have 300-500ms mypy daemon response time (even for huge code bases), this looks like a reasonable feature to have.

I'm finally on the verge of deploying mypy on our existing (moderately large) code base, but I've been struggling to get dmypy rerun times to a level where I feel comfortable running it automatically (e.g. prior to commit, or within a PyCharm plugin). With the mypyc update, regular mypy run times went from 1:30 to 30s (amazing!). With dmypy it's taking about 15s to rerun if I edit any files, and <1s if nothing has been edited. That's adequate but it's certainly not the 300-500ms that you're getting. I'm using 0.701. Do I need to be using the latest from master branch? Or are there some key techniques I should be using?

Thanks again for all your amazing contributions to this ecosystem.

[ilevkivskyi]

No, these are times for 0.701. Internally we are using a script that is basically a wrapper around dmypy recheck [--update|--remove]. We received a report from another user a while ago also reporting super slow daemon runs (10-20s for a moderately large codebase).

TBH this is a mystery that I would like to be solved, but I can't reproduce this. Maybe you could start the daemon with -vv and check for any anomalies in the daemon log file?

[chadrik]

Ok, I created an issue over at mypy.

[syastrov]

This would be an extremely useful feature. FWIW dmypy takes about 100ms to check 20k lines of code, so this would be perfectly acceptable for my use case. I have a wrapper script which uses git ls-files to find which files to check (so I can exclude things) and that takes 60ms itself, so in total, it takes 160ms.

[ilevkivskyi]

Yeah, this would make most sense together with #19.