\0 in a byte string literal does not encode NUL

[mneumann]

Currently:

assert: b"foo\0_bar\0" == b"foo\x5c\x30_bar\x5c\x30"

Expected:

assert: b"foo\0_bar\0" == b"foo\x00_bar\x00"

Maybe there is good reason for this.

[mneumann]

Might also apply to Strings.

[jemc]

Maybe there is good reason for this.

No, there's no reason apart from me just not knowing that \0 was a common escape sequence that people expect to be able to use.

I've never used the escape sequence \0 in another language - just to avoid me needing to search around a lot for what you already know, what language(s) are you used to using this escape sequence in?

I have no problem with us adding this escape sequence if it's common enough in other languages that people expect it.

[mneumann]

\0 is used in quite some languages: C, C++, Ruby, Python, Zig, ...

You can roughly say, if \n is an escape sequence in that language, then \0 is.

I don't need \0 as escape sequence, as it's use is rather rare! But, I'd make everything that follows \ special, and if it is unknown, I'd rather emit a compiler error than encode it as is. Otherwise, some people will use \0 (like I did :) and if they don't have tests will wonder why they get two characters \ and 0 instead of NUL.

[jemc]

Given the large number of languages where it's present, I'm definitely okay with us adding \0 if it is desired.

I also agree with you that we should issue a compilation error when an unknown escape sequence is encountered, in a String or Bytes literal. I'll prepare a quick PR to add that as a compilation error.

[jemc]

PR #353 adds a nice compilation error on invalid escape characters.

I'll leave this ticket open in case you want to do a PR to add \0 as a valid escape sequence.