string literal enhancement ideas

[andychu]

Related to expression language enhancements #865

(1) get rid of legacy backticks in double quotes, especially for "Pulp" dialect. These should be valid:

var x = "`"  
echo "`"  

(2) for generating Make or Ninja: percent interpolation instead of dollar interpolation

var mystr = %"string literal %var %{var} \% \" \\ "
echo %"hello %world"   # need special detection in command mode

for variant in opt asan gc_debug { 
  build %"out1 out2: $in.%variant" {
    var = 'value'
  }
}

Notes:

• An alternative design is to use doubling like %% and %", but I think that's too inconsistent and confusing.
• We should probably deprecate $"" in this case, because it's confusing? Or maybe it's consistent? It's just an alias for "" that emphases that $ is the interpolation character. But then that's confused with $'', which needs to be replaced with c''.

(Discarded idea: repurpose it for C-escaped strings? $"" and %"" could both be c-escaped with \n \r \t \0 \x00 \u{1234} ? Nah that conflicts with r'' vs c''. Too confusing)

[andychu]

• This came up again with user feedback -- c'' and $'' is confusing.
• other feedback: string escaping is important to avoid HTML injection, SQL injection, shell injection

[andychu]

Crazy idea for new syntax similar to #799

• echo ${'auto escaped string $foo ${bar} zzz '}
  • is this too close to $'foo\n' ?

Maybe Rust-like ?

• echo r#"foo"#

  • although the rules aren't exactly the same, so this could be confusing

• like symbolic tags in expression mode:

  • echo %c'foo'
  • echo %r"foo"
  • echo %raw'foo'
  • echo %html"<a href=$foo>"

I think the latter syntax might be compatible and not require a new lexer mode. You would have a token like Lit_SymbolLike and then look ahead for ' or ". And then THAT can enter a different mode if shopt --set parse_string_literals or --set parse_percent

• problem: %foo vs. %foo'bar' in expression mode. Does that cause a problem?

More ideas

• echo ${html'foo'}
• const x = ${html'foo'} -- this works with no change?

[andychu]

I think this is mostly obviated by the escaper or default_escaper mechanism.

#938

[andychu]

Yeah this is all too elaborate, I did something simple and effective with #940