Use extended brackets instead of box-drawing characters in log messages

[waldyrious]

This PR proposes using extended brackets instead of box-drawing characters in multi-line log messages. The box-drawing characters were first introduced in #25111.

This change improves formatting of multi-line log messages in two ways:

• They become semantically consistent with the single-line log messages (which use the left bracket character, [)
• They become more visually consistent with the single-line log messages, since the box drawing characters ┌ and └ have the horizontal section in the middle of the line, rather than at the top and the bottom like [.

With this change, the box drawing characters can now produce a left-flanking box that actually "wraps" the log message, (thus matching the style of the single-line log messages), rather than be slightly shorter than the text.

Current:

# Characters used:
# - '┌' Box Drawings Light Down and Right, U+250C
# - '│' Box Drawings Light Vertical, U+2502
# - '└' Box Drawings Light Up and Right, U+2514

┌ Info: parameters:
│   {UUID} = UUID("1a7ac041-bdf1-4da1-9e3a-91f6f6c88e92")
│   {PKGNAME} = "Foo"
│   {USERNAME} = "John Doe"
│   {USEREMAIL} = "[email protected]"
│   {GHUSER} = "jdoe"
└   {YEAR} = 2019
Confirm (Y)> 
[ Info: copy and substitute

Proposed:

# Characters used:
# - '⎡' Left Square Bracket Upper Corner, U+23A1
# - '⎜' Left Square Bracket Extension, U+23A2
# - '⎣' Left Square Bracket Lower Corner, U+23A3

⎡ Info: parameters:
⎜   {UUID} = UUID("1a7ac041-bdf1-4da1-9e3a-91f6f6c88e92")
⎜   {PKGNAME} = "Foo"
⎜   {USERNAME} = "John Doe"
⎜   {USEREMAIL} = "[email protected]"
⎜   {GHUSER} = "jdoe"
⎣   {YEAR} = 2019
Confirm (Y)> 
[ Info: copy and substitute

[waldyrious]

An alternative to make the printing of logs more consistent could be to go the other way and replace [ for single-line log messages with a simple hyphen -, or more appropriately, the box drawing character equivalent ╶:

# Characters used:
# - '┌' Box Drawings Light Down and Right, U+250C
# - '│' Box Drawings Light Vertical, U+2502
# - '└' Box Drawings Light Up and Right, U+2514
# - '╶' Box Drawings Light Right, U+2576

┌ Info: parameters:
│   {UUID} = UUID("1a7ac041-bdf1-4da1-9e3a-91f6f6c88e92")
│   {PKGNAME} = "Foo"
│   {USERNAME} = "John Doe"
│   {USEREMAIL} = "[email protected]"
│   {GHUSER} = "jdoe"
└   {YEAR} = 2019
Confirm (Y)> 
╶ Info: copy and substitute

But honestly I prefer the first approach to this one.

[waldyrious]

Pinging the author and reviewers of #25111: @fredrikekre, @c42f and @KristofferC.

[fredrikekre]

IIRC we tried this but those characters did not render nicely on some systems. Maybe it was some other characters we tried though.

[KristofferC]

Here is a screenshot on windows with the new characters to the right:

[waldyrious]

Ah, I feared that could happen. Alas, it's a font issue that we can't fix on Julia's side. 😞

Do you guys think it could be worth trying the alternate consolidation approach I mentioned above (or some other that may have been discussed somewhere)?

[Keno]

Just wanted to note that the box drawing characters have special support to look better on iTerm, so if we switch, somebody should validate that the alternative works there also.

[c42f]

I'm ok with a change of characters in principle, but I also think there's a high bar to pass in compatibility. If they don't render correctly out of the box on some systems I'd not want to change this.

I'm somewhat surprised the box drawing characters have worked out as well as they have TBH. Originally I had ascii for that reason; it's not pretty but it does work everywhere.

[DilumAluthge]

This PR has had no activity for five years+, so I'll close it as stale.

@waldyrious If you're still interested in working on this PR, let me know, and I can re-open it.

[waldyrious]

I'd be happy to drive this to the finish line. I believe the main concern was the font rendering issues, right? Perhaps @KristofferC can re-check whether the issues he noticed last time are still present? And of course, if anyone of the commenters who expressed this concern could also check on their systems, that would be great to give us reasonable confidence that we can move ahead with this.

If OTOH the rendering issues persist, I must regrettably agree to close the PR.

[DilumAluthge]

@waldyrious Can you rebase against the latest Julia master, and fix the merge conflicts? That will give us new PR builds for use with Juliaup, and then folks can test this out again on a variety of systems.

[waldyrious]

@DilumAluthge I've rebased the branch and fixed the conflicts.

[DilumAluthge]

Looks like there are some test failures in Logging:

Error in testset Logging:
Test Failed at /cache/build/tester-amdci5-14/julialang/julia-master/julia-32d0b45841/share/julia/stdlib/v1.14/Logging/test/runtests.jl:357
  Expression: total_entries == n * 3
   Evaluated: 0 == 30000
ERROR: LoadError: Test run finished with errors