Use the same $\mu$ character that the REPL produces with \mu-TAB

[ivarne]

Is this a breaking change, or does normalization hide the issue?

[timholy]

If users have units written to a disk file, this would be breaking. (I will be one such user affected by this change.)

However, after talking with Keno, we both wonder whether the better approach might be to change the \mu rendered in the REPL. For both him and me, the one that \mu-TAB produces is not the one that is suggested by our OS. So if you use the character generator rather than the LaTeX mode, you get conflicts.

CC @stevengj, @JeffBezanson.

[stevengj]

Technically, the micro sign µ in micrometers etc. is a different codepoint than the greek small mu that is produced by \mu. NFC does normalize one into the other; NFKC would normalize µ into μ but it was decided not to use NFKC (JuliaLang/julia#5434).

However, I think that this distinction is not worth preserving, and we are better off with using the same \mu codepoint for both.

[stevengj]

No, \mu should definitely not produce the micro codepoint. This would be inconsistent with how all of the other greek letters are treated.

Yes, typing option-m on MacOS produces the micro sign, not mu. But using Greek input mode on MacOS (which is what I used to use for Greek letters in Julia prior to the LaTeX substitutions) produces mu.

[stevengj]

As discussed in JuliaLang/julia#5903, it might be worthwhile to normalize µ (micro) to μ (mu) in identifiers, as one of a limited set of additions to NFC normalization.

[timholy]

I'll be happy with any decision that synchronizes these two.

[Keno]

Now that tab completion is more prevalent, I think this should be merged, even if it's not the µ you get from the keyboard.

[stevengj]

Why not just export both versions?

[Keno]

Oh, huh, somehow I had not considered that option ;). Great idea!

[Keno]

Done