Can we fix toLower/toUpper?

[pbiggar]

The .NET toUpper/toLower do not handle some unicode issues as well as the OCaml code. I believe .NET handles "simple" Case Mapping, but not "full" Case Mapping.

This comment describes what I know.

It appears that someone wrote some code that might solve this: dotnet/runtime#30960 (comment) (links to https://gist.github.com/tarekgh/de1a1b29b03d048ad580c0a338c44fbb). We should determine if this solves it, and if so, incorporate it in String::toUpper and String::toLower

[StachuDotNet]

We should determine if this solves it

To check this, I:

• created a new .sln locally, with C# src and F# test projects in it
• copied the C# code from the gist above into the src project
• wrote a few lines of F# to test the exposed fn there, against the relevant OCAMLONLY test cases here

And seemingly this doesn't not solve our situation? Or perhaps I am misinterpreting.

[pbiggar]

I am quite surprised it didn't work - the comments seemed to indicate a high degree of confidence that it would. But you are indeed testing the right thing afaict.

[StachuDotNet]

I was surprised too.
For what it's worth, I created a separate project in case you'd like to clone: https://github.com/StachuDotNet/StringCasingFix

Maybe it's worth mentioning in the other thread? Maybe it's specific to .NET 6?

[pbiggar]

Maybe it's worth mentioning in the other thread? Maybe it's specific to .NET 6?

For sure!

[StachuDotNet]

Maybe it's worth mentioning in the other thread? Maybe it's specific to .NET 6?

For sure!

I have a draft and would appreciate some feedback on the below, @pbiggar - still finding myself a bit lost here!

---

I'm attempting to use the provided gist to resolve the same (or seemingly such) issue of surprising ToUpper results noted here, in #3264, linked just before this comment.

Namely, we're hoping to find a solution that will ToUpper strings like "և" and "fifl" properly (to "ԵՒ" and "FIFL", respectively).

I've copied the code into a new .net6 project, only to find that they're not working the same as noted in the linked comment. My solution may be found here: https://github.com/StachuDotNet/StringCasingFix

This is a bit tangential, so hope I'm not distracting much, but was wondering if I'm missing something - should I expect this to work? Has something changed, perhaps related to .NET versions?

[pbiggar]

Here's what I'd write:

---

I tried the gist in .NET 6 and found that strings "և" and "fifl" did not convert to upper case properly (I was expecting "ԵՒ" and "FIFL", respectively). Here's a repo that replicates the issue.

Should I expect this to work?

---

And maybe tag tarekgh.

That said..

This isn't the exact same input than they were discussing, right? So there's two potential issues:

• their code doesn't work in NET6 or some other condition in the standalone repo
• their code never covered the input we are testing.

So it may be worth establishing which of these is the case, as that might help us to ask a better question and get a better response.

[StachuDotNet]

Hm. Their example of ffi, in my solution, yields ffi (no-op).

So, either this is an environment thing [my OS + .NET version + ???], or the gist and that comment aren't in sync, and I/we were mistakenly assuming they were.

[StachuDotNet]

FWIW, on .net core 3.1, I get the same results:

Expected: 'TEST'; Actual: 'TEST'
Expected: 'FIFL'; Actual: 'fifl'
Expected: 'FFL'; Actual: 'ffi'
Expected: 'ԵՒ'; Actual: 'և'

I could set up .NET on my Windows machine and try there to rule out OS? Seems somewhat overkill, though, and now just thinking the gist doesn't match what we're looking for.

[pbiggar]

Yeah. Maybe let them know and we can move on if they don't respond.

[StachuDotNet]

We got a response dotnet/runtime#30960 (comment)
Sounds like we misinterpreted things

[pbiggar]

Whoops!

What should we do here then? Close this issue?

[StachuDotNet]

I've scoured the internet the best I can to find another solution here, but haven't found anything.
Theoretically, this is still possible, I presume, so closing feels a bit wrong, but likely not blocking for an F# release. So maybe let's leave the conversation here and leave it for "some day"?

[pbiggar]

Makes sense, thanks!

[StachuDotNet]

Should this be closed in favor of #3437? (or the other way around)