Hacker News Reader

> [edit]: flake8, remarkably, dropped roughly an order of magnitude in performance between version 2 and 3. So that didn't help.

Sorry for the drive by, but did you try bisecting the flake8 change that slowed it down? Unless there were thousands of commits between releases, it would only take <10 runs to bisect, an automated bisect should be under an hour with a test case that takes a few minutes. Could be an interesting thing to try?

View on HN

Doubt it would be helpful. It appears that the model changed dramatically between those two versions. For instance, the whole API (which we relied on) was dropped[0] and some legacy API was glued in place.

In either case, the information wouldn't be all that useful. We run our stuff off of Debian packages (yes, we're old school), and we're not particularly interested in maintaining our own fork.

[0] https://flake8.pycqa.org/en/latest/user/python-api.html#lega...

Reply on news.ycombinator.com