Issue 11819 - Implement better diagnostics for unrecognized traits.

[ghost]

https://issues.dlang.org/show_bug.cgi?id=11819

I wish I could avoid duplicating the list of traits with the ones in the msgtable in idgen.c, and that I could avoid hardcoding 39 but I want to keep DDMD compatibility.

But any improvement tips are always welcome!

[yebblies]

Instead of using a hardcoded number, you can put a NULL sentinel at the end.

You could also do it with a linear search instead of a string table. I 39 items in a very specific error message is going to be a performance bottleneck, and then you could get rid of the init method.

The number of errors doesn't match up with the number of traits invocations, do you need to split it to get the full error list?

[yebblies]

You could also do it with a linear search instead of a string table. I 39 items in a very specific error message is going to be a performance bottleneck, and then you could get rid of the init method.

I just remembered that speller calls that method 1000s of times. Maybe a hash table is worthwhile after all.

[ghost]

The number of errors doesn't match up with the number of traits invocations, do you need to split it to get the full error list?

Oops, I didn't realize the limit was getting reached. Will fixup.

[ghost]

Updated.

[ghost]

I just remembered that speller calls that method 1000s of times. Maybe a hash table is worthwhile after all.

Yeah it definitely calls it a lot of times.

[yebblies]

Looks good

[yebblies]

Auto-merge toggled on