Issue #19719 has been updated by Eregon (Benoit Daloze). IMO it's quite strange and not respectful of the Ruby development workflow to: * create an issue about a big feature but discuss literally none of the details. Notably the background, design and release management sections are all completely unrelated, showing there are many holes in this issue description and maybe in the design of this feature. * open this issue on the day of the meeting, at the last minute * merge [large](https://github.com/ruby/ruby/pull/7927) code changes (+ 6607 lines) with 0 reviews, 0 comments, 0 design discussions It feels like sneaking in changes to me. Also while replacing bison and making the parser more independent of CRuby internals is useful, who is going to maintain all that extra code? Right now I don't see anyone happy to maintain either of those, except Kaneko-san, which of course might not always be available to fix/evolve things. Maintainability seems a crucial aspect that was not discussed at all here. For instance, it seems pretty clear to me that Universal Parser wouldn't improve maintainability of the parser significantly (parse.y has many maintainability issues, I think this wouldn't solve most of them). It would certainly be good if more committers and contributors would feel comfortable modifying the parser, right now it seems everyone avoids it as much as possible, and I don't see Universal Parser helping there. ---------------------------------------- Feature #19719: Universal Parser https://bugs.ruby-lang.org/issues/19719#change-103539 * Author: yui-knk (Kaneko Yuichiro) * Status: Closed * Priority: Normal ---------------------------------------- # Background There are three use cases where we need a CRuby parser which is independent of other CRuby features like Object, GC. I call such a parser a Universal Parser. 1. Use Universal Parser from Ruby applications. For example RuboCop. RuboCop needs a parser to get AST of source code. Currently RuboCop uses parser gem. In this sense Universal Parser is a replacement of parser gem. 2. Use Universal Parser from C/C++ or other language. For example Sorbet. Sorbet is written in C++. It has its own parser. Universal Parser can be used in such scenario. 3. Use Universal Parser for other Ruby implementations. mruby, JRuby and other Ruby implementations will use Universal Parser so that they don’t need to develop & manage their own parser. # Design * Implement Universal Parser by changing CRuby source code, especially parse.y and node.c. Introduce `UNIVERSAL_PARSER` macro and implement Universal Parser by passing all necessary CRuby related functions via `struct rb_parser_config_struct`. In this step there are two build modes with/without Universal Parser. * Reduce CRuby related functions passed by `struct rb_parser_config_struct`. Some of them are copied into parse.y, e.g. `rb_isspace`. Other are reimplemented, e.g. `NODE_LIT` has new String struct instead of `VALUE`. * Once CRuby related functions needed for Universal Parser are minimized, replace rest CRuby function references with functions provided by `struct rb_parser_config_struct` and remove `UNIVERSAL_PARSER` macro. # Release management There are three options for releasing a binary for Universal Parser (“librubyparser”). 1. Release it as so/dll a. Include it into Ruby release process b. Create another repository for release management of "librubyparser" 2. Release it as gem (ruby/universal_parser) "librubyparser" has only pure C interface and data structure. If you want to use it from Ruby code, you need C extension code to translate "librubyparser" data and Ruby objects. I propose to create "universal_parser" gem for this purpose. I prefer #1-b to #1-a because it doesn’t increase tasks for Ruby release process. I want to make #2 as a first milestone. -- https://bugs.ruby-lang.org/