Improve error messages for list elements with the wrong type #236
Conversation
Fixes #230 The type error for the list now returns the expression for the mismatched element instead of the entire list. This improves the source span displayed to the user to focus on the invalid element. Since the source span now matches the invalid element, there is no need to redisplay the same element in the detailed error message.
There was a problem hiding this comment.
Not related to this PR, but have you considered some golden tests the test each of these error messages? It also works as a nice design tool, as people can easily review all error messages and easily provide some commentary on how the messages could be improved.
|
The only reason I haven't done this yet is because the error messages are not part of the API contract and they are pretty brittle to maintain I think if I did this it would be to add functions for automating common error message patterns (i.e. wrapping code in ASCII boxes, wrapping to 80 columns) and testing those functions. |
|
Golden tests don't really imply that, they are just a mega easy way to test things which have a known output. That output can still change over time, of course, it's just trivial to say "yea yea, my changes are what I wanted". I've come to really enjoy them. In a sense, golden testing the error messages against Dhall expressions also stresses the type checker, too. |
|
Alright, but there's still one other thing that has to be implemented, which is an easy way to update the tests to accept the new golden output. For example, if you significantly refactor error messages and 50 tests start failing, it would be time-consuming to update all 50 of them manually. |
|
With golden testing, you have a runner that just accepts the new result. With |
|
Ah, I didn't know that |
Fixes #230
The type error for the list now returns the expression for the mismatched
element instead of the entire list. This improves the source span
displayed to the user to focus on the invalid element.
Since the source span now matches the invalid element, there is no need
to redisplay the same element in the detailed error message.