Phil Quitslund

Great conversation! FWIW, I'm very much w/ @jakemac53. Exploring something like what @lrhn proposed sounds valuable to me (though a little adjacent to what @Hixie was after in the OP).

/cc @stereotype441 for thoughts on NNBD implications

/fyi @goderbauer

cc @bwilkerson

Aside: we currently do own own pubspec parsing, but perhaps we should consider moving over to using [pubspec_parse](https://github.com/dart-lang/pubspec_parse)? @kevmoo: is this the kind of use case you envisioned?

@bwilkerson: worth considering using `pubspec_parse` in analyzer?

I agree with Brian. This seems like a smell but probably a different lint.

Thanks for the fix @natebosch. I updated the example. I totally agree that generalizing is preferred and was kind of wanting to babystep but maybe that's too cautious. An experiment...

Thanks all for the thoughtful feedback! In our last conversation we discussed an experiment to vet the impact of a constrained generalization which lines up w/ @lrhn's proposal to focus...

See also: https://github.com/dart-lang/linter/issues/1123