Give better error when matching too many fields #13062
Conversation
| else args | ||
| case (argType : AppliedType) :: Nil | ||
| if (defn.isTupleClass(argType.tycon.typeSymbol)) && | ||
| (args.lengthCompare(1) > 0 && Feature.autoTuplingEnabled) => |
There was a problem hiding this comment.
Och yeah, we should remove those. Thanks!
There was a problem hiding this comment.
Better to move isTupleClass after args.lengthCompare since it is a more expensive test.
There was a problem hiding this comment.
It would be good to add a check file for i10757 that tests that the error message is correct.
The easiest way to do that is to generate a dummy check file i10757.check first and then follow instructions when running the test (it will suggest you a command to move the correct check file i10757.check.out that it generated to the actual check file i10757.check.
There was a problem hiding this comment.
It would be good to add a check file for i10757 that tests that the error message is correct.
The easiest way to do that is to generate a dummy check file i10757.check first and then follow instructions when running the test (it will suggest you a command to move the correct check file i10757.check.outthat it generated to the actual check filei10757.check`.
In fact #13063 seems to be one step closer to the finish line, so I think we can continue with this one.
| else args | ||
| case (argType : AppliedType) :: Nil | ||
| if (defn.isTupleClass(argType.tycon.typeSymbol)) && | ||
| (args.lengthCompare(1) > 0 && Feature.autoTuplingEnabled) => |
There was a problem hiding this comment.
Better to move isTupleClass after args.lengthCompare since it is a more expensive test.
I must say it's a bit discouraging that we worked on it during the spree for the effort to be for nothing. Would be nice to work on the communication aspect a bit, so that this doesn't happen again. |
|
@tgodzik I was also surprised to see the double PR. Agreed we should avoid that in the future! |
|
I'm sorry that this happened, but communication was very literally the problem: it seemed Tomasz couldn't hear me in Spacial Chat even when my avatar was right on top of his! So I rage quit when I realised that all the comments I'd made were actually me speaking to myself for half an hour, like an idiot. It's a shame because I really like the concept of the chat, but it sucks that the implementation let's it down (including the CPU hogging...). It is at least reassuring that we both reached the same conclusion on how to fix the issue though. |
|
That explains what happened, I saw your avatar, but I thought that you might be muted on purpose and haven't realized you were talking to us. Glad to see the particular issue fixed, but next time we should make sure that we additionally communicate on the issue itself, I think that would help here and I haven't properly communicated there. |
|
That is a fraught story. Footnote, the other PR doesn't have "convert to pattern guard" with reduced parens. PR stands for "paren reduction". I was also curious whether the name-based symbol test or the type test is better in this context. Probably the type test in the other PR, but in such cases where there are 2 ways, it's not obvious if there is a downside (forcing info or inducing circularity). |
Fixes #10757