CI: Add lgtm.com integration to pull requests

[jschendel]

Seems like lgtm.com integration could be useful. Was helpful in identifying an obscure potential issue fixed in #20563, and I've used it to find a couple other minor things.

Looks like both numpy and scikit-learn are using it as well. Not sure exactly what their experience with it has been, but at the very least scikit-learn has kept it around for a while now.

[gfyoung]

Haven't used it before, but could be interesting to look into! Perhaps a brief outline of the benefits (and how it has helped you previously) would be good.

[jhelie]

Hi @jschendel and @gfyoung and thanks for your interest in LGTM! I am one of LGTM devs and a big user of pandas so would love for you to give it a go.

In a nutshell: what sets us apart is our query language technology QL that allows us to find complex issues with a low false positive rate. You can write your own queries in QL to look for anything you might be interested in (see for instance here for a tutorial on points-to analysis).

At the moment PR integration means the standard set of queries is run but you can customise what is analysed and turn queries on and off. And very soon it will be possible to add custom queries to the PR checks.

In addition to scikit-learn and numpy, matplotlib is also using it - and I hope scipy will join soon too :)

I see @jschendel has already been using lgtm.com but I'm happy to answer any questions you have!

[jbrockmendel]

Is this worth pursuing?

[mroeschke]

Looks like this never gained much traction. Going to close due to the staleness but happy to revive if there's continued interest