Skip to content

Commit c37039f

Browse files
tannevaledclaude
andcommitted
landing: v0.4.0, the measured corpus figure, and parse speed stated honestly
- The site named v0.2.0 (2026-09-21) in three places. The real latest tag is v0.4.0, cut 2026-09-26; main is exactly that tag. - "every one of the 68 files in the pinned ruby/spec language/ corpus now parses" was right in substance but wrong in its count: the pinned corpus has 80 _spec.rb files under language/, and go-ruby-parser parses all 80. Measured here by calling parser.Parse on each. - Added a second completeness figure a reader can check: of 47 top-level CRuby 4.0.5 stdlib files MRI accepts, this parser accepts 46. - Added parse speed, because the repository README's figures did not reproduce. On 46 real stdlib files both engines accept: ~1.9x quicker than Ripper.sexp, and about 1.6x slower than CRuby's own C parser. The landing page previously said nothing about speed, and the README's claim to beat the C parser is corrected in its own PR. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
1 parent 1ddb7d2 commit c37039f

2 files changed

Lines changed: 9 additions & 6 deletions

File tree

‎hugo.toml‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -9,4 +9,4 @@ disableKinds = ["taxonomy", "term", "RSS", "sitemap", "404"]
99
name = "parser"
1010
url = "https://github.com/go-ruby-parser/parser"
1111
status = "ready"
12-
blurb = "The front-end: `parser.Parse(src) (*ast.Program, error)`, with sub-packages `token`, `ast` and `lexer`. A stateful lexer (MRI `spaceSeen` semantics), a recursive-descent + Pratt expression parser with a scope stack, and a typed AST for a broad, practical subset of **Ruby 4.0** — literals, operators, control flow, `case`/`in` **pattern matching**, methods/blocks, classes/modules/metaprogramming. 100% covered (the gate runs on all three OS lanes; the suite also runs on all six 64-bit targets and on js/wasm and wasip1/wasm), differential-tested against MRI Ruby 4.0.5. Current release **v0.2.0** (2026-09-21): fourteen grammar gaps closed, and every one of the 68 files in the pinned ruby/spec `language/` corpus now parses."
12+
blurb = "The front-end: `parser.Parse(src) (*ast.Program, error)`, with sub-packages `token`, `ast` and `lexer`. A stateful lexer (MRI `spaceSeen` semantics), a recursive-descent + Pratt expression parser with a scope stack, and a typed AST for a broad, practical subset of **Ruby 4.0** — literals, operators, control flow, `case`/`in` **pattern matching**, methods/blocks, classes/modules/metaprogramming. 100% covered (the gate runs on all three OS lanes; the suite also runs on all six 64-bit targets and on js/wasm and wasip1/wasm), differential-tested against MRI Ruby 4.0.5. Current release **v0.4.0** (2026-09-26): all 80 `_spec.rb` files in the pinned ruby/spec `language/` corpus parse, and the three limitations the docs used to list — paren-less command calls with keyword/splat/block args, default block parameters, and the positional `Class(a)` find-pattern — all parse now. `BEGIN { }` / `END { }` remain the one refusal."

‎layouts/index.html‎

Lines changed: 8 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -88,7 +88,7 @@ <h1>go-ruby-parser</h1>
8888
<span class="pill">stateful lexer</span>
8989
<span class="pill">recursive-descent + Pratt</span>
9090
<span class="pill">typed AST</span>
91-
<span class="pill">v0.2.0 · 2026-09-21</span>
91+
<span class="pill">v0.4.0 · 2026-09-26</span>
9292
<span class="pill">Ruby 4.0 subset</span>
9393
<span class="pill">pattern matching</span>
9494
<span class="pill">heredocs</span>
@@ -119,10 +119,13 @@ <h1>go-ruby-parser</h1>
119119
<strong>100% coverage</strong> is enforced in CI on all three OS lanes, and the suite also runs
120120
on all <strong>six 64-bit targets</strong> — amd64 and arm64 natively, riscv64 / loong64 /
121121
ppc64le / s390x under QEMU — plus <code>js/wasm</code> and <code>wasip1/wasm</code>.
122-
The current release is <a href="https://github.com/go-ruby-parser/parser/tree/v0.2.0"><strong>v0.2.0</strong></a>
123-
(2026-09-21), which closed fourteen grammar gaps: every one of the 68 files in the pinned
124-
ruby/spec <code>language/</code> corpus now parses, thirteen of which had not parsed at all
125-
before.</p>
122+
The current release is <a href="https://github.com/go-ruby-parser/parser/tree/v0.4.0"><strong>v0.4.0</strong></a>
123+
(2026-09-26). Measured on 2026-09-26: <strong>all 80</strong> <code>_spec.rb</code> files in the
124+
pinned ruby/spec <code>language/</code> corpus parse, and of 47 top-level CRuby 4.0.5 stdlib
125+
files MRI accepts, this parser accepts 46. On parse speed it is roughly
126+
<strong>1.9× quicker than <code>Ripper.sexp</code></strong> and about
127+
<strong>1.6× slower than the C parser built into CRuby</strong>
128+
(<code>RubyVM::AbstractSyntaxTree.parse</code>).</p>
126129

127130
<section class="grid">
128131
{{ range .Site.Params.repos }}

0 commit comments

Comments
 (0)