Benoit Daloze comments

Results 1300 comments of


                                            Benoit Daloze

Add test for line number in HEREDOC with method call

That's fine, it's already used in many places, and the CI will notice immediately if the number of lines in between changes.

Extend the tests in regexp/inspect_spec with over escaping combined with character classes

> I'm not that familiar with this project and how things are structured, can I just add a few tests to the inspect spec of regexp, and make a new...

Update Dockerfiles and readme file

Should we merge this?

TravisCI did not run recently

> I discussed this recently with @stanhu too and he recommended GitLab CI. :) I don't know much about GitLab CI, but since this project is on GitHub I'd expect...

Yes, this seems to happen to many people still using TravisCI recently it seems: https://twitter.com/smarr/status/1380195827591876612 https://twitter.com/sh0dan/status/1380555125714997253 Maybe they reduced open-source credits or so significantly? (and before that it was very...

Non-interactive `chruby-exec` fails with `zsh:1: command not found: chruby`

FWIW there is `~/.zshenv` which is always sourced. That's where I do PATH modifications. And my `~/.zshrc` is then for interative-only stuff like completion, etc.

Parser::Lexer#advance method is too big for JIT compilers

> 4. Extracting larger semantic actions into methods/lambdas would definitely help. Maybe we could experiment by doing that as postprocessing (e.g. using `parser` :D) or tweaking Ragel. Specifically we could...

Parser::Lexer#advance method is too big for JIT compilers

> True, but it will also make it slightly slower on MRI, right? It could, we'd have to measure it to know. Probably not much if it's extracted methods. We...

Parser::Lexer#advance method is too big for JIT compilers

> Right, that's correct. It's probably much harder to analyze if it's `goto`s everywhere, but ultimately it depends on the max number of live variables at a point in the...

Parser::Lexer#advance method is too big for JIT compilers

BTW, I noticed in Ripper, the lexing function seems to be `parser_yylex` and that's only 707 lines long: https://github.com/oracle/truffleruby/blob/c9c717bf822deb4d5592eeedc05d1842bf72b88d/src/main/c/ripper/ripper.c#L16309-L17016 That's about 10x smaller than Ragel-generated lexers, which is quite an...