I believe part of this is already covered by https://github.com/phpstan/phpstan/blob/master/src/Rules/Arrays/NonexistentOffsetInArrayDimFetchRule.php and the other part would be covered by my suggestion in #1307. But it's always good to at least recycle these tests to corresponding rules once they're implemented so I'm leaving this open.

See https://phpstan.org/r/4b9affb1252f4ccab55ef797703101cb - it's already handled.

It does not show up here https://phpstan.org/r/2445f50bdd5edf1fd8cee57bf45fd38d because $foo becomes an array on line 7.

I don't understand. This should report and error:

https://phpstan.org/r/dbda208449aa9798e2901be22bfba409

So should this:

https://phpstan.org/r/9f1b618077bf1b9ca1db1c6051378eb1

There's no rule for checking that setOffsetValueType returns ErrorType. I suggested that here:
#1307 (comment)

To summarize - there's already a rule for hasOffsetValueType. It can be improved by grabbing relevant testcases from this PR and fixing those cases.

I also suggested in #1307 to add a rule that checks whether setOffsetValueType return ErrorType in cases of $foo['bar'] = 'baz';

@ondrejmirtes I see. Sorry for the confusion 😆

I've recycled your tests and used them with the current rules and typesystem in this commit: 7a630b8

Thanks!

@ondrejmirtes Cool, thanks 🙂