Skip to content
This repository has been archived by the owner on Apr 8, 2024. It is now read-only.

Latest commit

 

History

History
103 lines (79 loc) · 4.95 KB

meeting-notes.md

File metadata and controls

103 lines (79 loc) · 4.95 KB
Raw

2019-03-06

  • parallel test harness runner landed!

  • we reviewed the content for creating a PR for the Grammar-WG to rust-lang/lang-team

  • @CAD is still working on the %% PR

  • @qmx is taking unit-testing as his next task

    We've had a brief discussion on disambiguation, precedence tables as an option to solve this. Since GLL gives us parse forests, we have to figure out a way to pick one of the trees.

    We have $production(.$random_description?)(.$number_in_sequence)?.expected.

    We test that all of the .expected files are in the parse forest. If $number_in_sequence is omitted then this asserts uniqueness (the forest is a tree).

    Once we have a way to test the grammar with expectations, we can experiment with post-processing the parse forest and remove ambiguities; from that we hope to gain a declarative way of specifying disambiguation in .rs files. Eventually that may be converted into the .lyg format.

2019-01-23

We're making slow, but steady progress! PRs from @CAD97, @cramertj, and @ehuss!

  1. lykenware/gll maintenance.
  • The overall consensus is that we need to wrap our brains around it ASAP, and reduce the bus factor.
  • @cramertj started refactoring/cleaning up some bits, we should probably do the same until we get more clarity around how gll works.
  • It's clear that we all miss @eddyb :D
  1. Removing the existing grammar from rust-lang/rust: #9
  • We punted the topic for the next meeting, there was a suggestion of just moving the old grammar into our repo and linking to it.

2019-01-09

tentative agenda

  • deal with the in-flight PRs
  • setup travis for the repo
  • mini-roadmap discussion

action items

  • @qmx parallelization with rayon, travis fixes, blacklist => toml.
  • @CAD %% and |=
  • @cramertj inline unit tests
  • @centril removing grammar redundancies + spicing up repo structure
  • @ehuss Making a PR to close issue #1 (charter), cleaning up + enhancing lexical spec in reference, use that to close #3 (lexical spec).

2018-11-28

Summary for today's meeting

  • @eddyb#0426 had nothing to report since no progress has been made
  • @Centril#2862 worked on an RFC rust-lang/rfcs#2602 and tested out the GLL syntax, it felt good, https://github.com/Centril/rfcs/blob/rfc/attributes-galore/text/0000-attributes-galore.md#grammar particularly with the Rule |= Additions; formatting. @qmx#4460 agreed it was pleasant and that it was "BNF done right".
  • We call the notation in GLL: "lyg".
    • Change file ending to .lyg after merging PR.
  • @Centril#2862 felt that parameterized grammars could reduce some duplication with e.g. Bounds<P> = elems:P* % "+" "+"?; and similar things.
  • Cargo features should be tried out for nightly features
  • @qmx#4460 needs to: "setup the initial test harness with rust-lang/rust vendored into the grammar repo as a submodule"
  • @eddyb#0426 needs to do some stuff on the GLL side;
  • @qmx#4460 will work on adding %% to GLL
  • Test harness is top prio;
  • @Centril#2862 will work on improving @eddyb#0426's PR wrt. style and such things.
  • @ehus might split some work with @qmx.

2018-11-14

Last wednesday the wg-grammar team had it's biweekly meeting - here are the principal points:

  1. Progress report on @eddyb's PR
  • how to deal with unstable features
    • We're going to go with separate files and using |= for defining alternate rules.
    • when to merge @eddyb's PR?
    • There's a blocker where GLL uses a lot of ram for parsing a single file - once @eddyb fixes this we'll merge the PR.
  1. testing
  • inline unit tests
    • We've decided to go with tests in separate files, mimicking the grammar file tree structure, since this brings us the advantage of being able to even had AST/parse tree expectation files
  • The initial proposed test file name scheme is <Production>.<random_description>.<ext>
  • testing against external corpus
    • The initial testing corpus will be rust-lang/rust, once we get to a good place we'll extend this to maybe top 100 crates.io (stable)
  1. how to deal with ambiguity (declarative...)
  • eddyb thinks that's too early to discuss this, we're punting this to the near future
  1. GLL
  • formatting
    • we asked for small changes on the overall feel for GLL grammars, GLL PR with the proposed changes.
FnArgs =
  | Regular:FnArg+ % ","
  | Variadic:"..."
  | RegularAndVariadic:{ args:FnArg+ % "," "," "..." }
  ;

We're using leading | and moving the other | to the start of the line.

  • syntax changes
    • there were some discussions about using BNF-like ::= for attribution of new rules but this went nowhere at least for now.
    • there were also discussions about having parameterized grammars