Add \r to the list of non-unescaped characters to fix windows CI

[LilithHafner]

No description provided.

[codecov]

Codecov Report

Merging #14 (421f100) into lh/ci-old-julia (63ff132) will not change coverage.
Report is 1 commits behind head on lh/ci-old-julia.
The diff coverage is 100.00%.

@@               Coverage Diff                @@
##           lh/ci-old-julia      #14   +/-   ##
================================================
  Coverage            67.21%   67.21%           
================================================
  Files                    7        7           
  Lines                  796      796           
================================================
  Hits                   535      535           
  Misses                 261      261           

Files	Coverage Δ
src/stylemacro.jl	71.42% <100.00%> (ø)

[LilithHafner]

Whelp, that fixed it. Manually listing many escape characters is typically codesmell. It is notoriously difficult to get escaping and unescaping right. @tecosaur, how do we know we've got the right set of characters to not unescape?

[tecosaur]

Well, each one of those was added for a particular reason. Otherwise things like {red:\}} wouldn't work.

[tecosaur]

What I wonder with \r is whether we need to change the special handling with \n that allows for:

styled"some content \
       split across \
       lines"

(this is currently missing a test)

[LilithHafner]

Well, each one of those was added for a particular reason.

IIUC ['{', '}', ':', '$'] is a complete list of the special characters used in the styled macro and therefore must receive special treatment here. That makes sense.

The presence of \n and \r is weird, though.

[tecosaur]

\n was added because I wasn't able to make the \-continued lines work without it.

[LilithHafner]

I was going to say "shouldn't you just use the built-in support for strings without newlines split across multiple lines?" but it seems that Julia does not have support for this 😢

[tecosaur]

Everything from Julia strings that "just works" here has been manually implemented 😅.

[LilithHafner]

That's sad and error-prone. Julia should provide a way to reuse the existing string parsing machinery.

julia> "abc\
       def"
"abcdef"

julia> macro noop_str(str)
           println(str)
           str
       end
@noop_str (macro with 1 method)

julia> noop"abc\
       def"
abc\
def
"abc\\\ndef"

[LilithHafner]

It looks like Base strings identify and delete things that match the regex\\(\r|\n)\s* (i.e. backslack, followed by a newline or carriage return, and then any subsequent whitespace)

https://github.com/JuliaLang/JuliaSyntax.jl/blob/1b048aad2a3a3e2e9d7f5053ff51024f08ca308c/src/parser.jl#L3354-L3358

[tecosaur]

I thin we're just accomplishing the same effect via state machine instead of regexp.

[LilithHafner]

Oh, JuliaSyntax also doesn't use a Regex, that was just to state what's going on.

[LilithHafner]

Anyway, moving discussion to #16

[tecosaur]

I think this is superseded by #16 being merged.