more verbose multi-line display(c) for Char

[stevengj]

show('$') prints a terse representation like '$'. I've been thinking it would be nice to have a more informative display in the REPL (and in other things like IJulia that use the 3-argument show via display). This PR does so, giving output like:

julia> '$'
'$': ASCII/Unicode U+0024 (category Sc: Symbol, currency)

julia> '⊗'
'⊗': Unicode U+2297 (category Sm: Symbol, math)

[JeffBezanson]

Maybe use 6 instead of 8?

[stevengj]

ok

[quinnj]

What is category Sc vs category Sm? Should it maybe just be category: Symbol, currency? What are the non-Symbol characters? control characters?

[stevengj]

@quinnj, both the two-letter abbreviations and the expanded names for the Unicode character categories are standard, so it seemed useful to give them both.

[TotalVerb]

Would the Unicode description, like

'$': Unicode U+0024 Dollar sign (category Sc ...)

be possible/desirable to add?

[stevengj]

@TotalVerb, not without storing the Unicode character-name database somewhere; it's about a MB uncompressed IIRC (see also #19725 and related issues), so I'm not sure this is a great thing for us to do.

[stevengj]

One option that occurred to me is to have a show(io, "text/html", c::Char) method that links to the complete description of the codepoint on e.g. www.fileformat.info, for example U+0302, but that doesn't help in the REPL.

[stevengj]

It also occurred to me to output the tab-completion info, which @ScottPJones also suggested on the mailing list, but this seemed too REPL-specific. It seemed better to just have this in the REPL help, e.g. if you type ?⊗ it should tell you that you can type ⊗ with \otimes<tab>, in addition to any other help.

Update: see #19858.

[stevengj]

I'm not understanding the Travis failures here... oh, it was #19848.

[stevengj]

Should be good to merge; the i686 travis failure seems to be an endemic unrelated problem at the moment.

[StefanKarpinski]

I like it. Should we make sure that repr(c) is still just 'x'?

[stevengj]

I added a test for repr. (repr uses the 2-argument show, which hasn't been changed.)

[stevengj]

Should be good to squash/merge; the Travis failure is the unrelated x86 problem we are having nowadays.