nosj 0.2.0 → 0.3.1

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 78314f13c659fff6d0e146ac0b951564e66b7b10deeb439e317703754a0100ef
-  data.tar.gz: 993f8a38c4e04dcd68a6ec8ee2dc62c611d2a04c1bc0ff23fc31d67ad69dd585
+  metadata.gz: 7152dbc291ed6d291977555ba6a6b36f6d3d52dad57348a0baf8271212d1bb25
+  data.tar.gz: '083bd8ca6fac180455b4847c1634f4ac28455db8e6e85def4e14393700e3008f'
 SHA512:
-  metadata.gz: 6b5f09e43ed85c2e345a35d0fd681900e205d4743161542d82ba5c3d1c80879c62bff3acf88f12dac5836334c67d933cc6241ead7dbe6a7a820d5f672bccd3d4
-  data.tar.gz: 0b48922585bb4863f9ed71fb9500a9a60b4c1f857baeaaf79136ed4c9e752aa71c9c23235e85d3bab38b1f0967b7b7212ed6d771be57d62d87fcd39e91666cd8
+  metadata.gz: 254595b89839c50a48adedc03f8ec70498cc0cf6876390565f3dd188cc5dfca564c498443d0091c23bbf6562bb11b1724a25c073fa9e76986b8b1779cc91765d
+  data.tar.gz: 4d321fb92d1199649dc10f84853aa3124e40cc642b3da680772f392c315374116f04d09882f4fd95f2c445941b384f15e23ab89fc3c735edebd8868855cb8582

data/CHANGELOG.md

@@ -1,3 +1,115 @@
+## [0.3.1] - 2026-07-17
+
+- Fixed: `NOSJ.minify` / `NOSJ.reformat` produced unparseable output
+  for a float that overflows to Infinity—a huge-exponent literal like
+  `1e999`, or a ~300-digit integer with an exponent (such literals
+  parse to Infinity even in strict mode, matching the `json` gem).
+  The pipe emitted a bare `Infinity` token; it now raises the same
+  `GeneratorError` that `generate` would
+  (`"Infinity not allowed in JSON"`). With `allow_nan: true` the
+  literal still passes through as `Infinity`. Found by fuzzing.
+- Fixed (second manifestation, also found by fuzzing): the same
+  unparseable output could appear with the overflowing literal hidden
+  behind a duplicate object key. Because the reformat pipe
+  deliberately preserves duplicate-key entries, it now refuses such
+  documents with the same `GeneratorError` even though
+  `generate(parse(x))` succeeds there (last-key-wins parsing discards
+  the shadowed value)—a documented divergence.
+- Differential fuzzing for the native extension (`ext/nosj/fuzz`):
+  three cargo-fuzz targets—reformat, NDJSON framing, and
+  byte-splicing/JSON Patch—each drive the real entry points on an
+  embedded Ruby VM and compare every input against pure-Ruby
+  reference implementations, with committed seed corpora and a weekly
+  CI workflow (`fuzz.yml`).
+
+## [0.3.0] - 2026-07-17
+
+- Reformat without parsing. `NOSJ.minify(json, opts)` and
+  `NOSJ.reformat(json, opts)` (plus `NOSJ.reformat_file`) pipe the
+  parser's events straight into the emission kernels: zero Ruby
+  objects are allocated for the document, and output is exactly
+  `generate(parse(json))`—canonical numbers, normalized escapes, the
+  full set of `generate` formatting and escape options, with
+  `pretty: true` as a `pretty_generate` shorthand—except duplicate
+  object keys pass through and lone-surrogate string values re-escape
+  as `\uXXXX` instead of raising (the output must always reparse).
+  Acceptance options apply per `parse` (`allow_trailing_comma`
+  normalizes the commas away). Measured on the 631 KB twitter.json:
+  409µs, 3.4× faster than `NOSJ.generate(NOSJ.parse(x))`, 3.9× faster
+  than gem json's cycle, 5.3× faster than Oj's, and 1.4× the cost of
+  `NOSJ.valid?`.
+- Byte-splicing edits and JSON Patch. `NOSJ.splice(json, pointer =>
+  value, ...)` replaces values directly in the text: every target
+  resolves in one forward pass and the result is rebuilt copying all
+  bytes outside the target spans untouched (formatting, key order, and
+  number spellings elsewhere survive exactly). Measured on
+  twitter.json: 10× faster than parse-mutate-generate for a late
+  field, 51× for an early one. Missing targets raise KeyError,
+  overlapping targets ArgumentError. `NOSJ.patch(json, ops)` applies
+  RFC 6902 JSON Patch (add/remove/replace/move/copy/test, String or
+  Symbol op keys) to the raw string the same way, with structural ops
+  walking only the parent container's span; application failures raise
+  the new `NOSJ::PatchError`, malformed patch documents ArgumentError.
+  `NOSJ.merge_patch(json, patch)` applies RFC 7386 JSON Merge Patch
+  (semantic form). Inserted values are byte-identical to
+  `NOSJ.generate` and accept its options; the RFC 6902 appendix-A
+  suite and the full RFC 7386 test table are in the specs.
+- NDJSON / JSON Lines. `NOSJ.each_line(source, opts)` yields one
+  parsed value per line (Enumerator without a block, so
+  `.first(10)`/`.lazy` walk only what they consume), skipping blank
+  lines and enforcing one value per line; a malformed line raises the
+  rich `ParserError` whose `#line` is the physical line number in the
+  stream. `NOSJ.generate_lines(values, opts)` emits one compact
+  newline-terminated document per element in a single buffer pass
+  (measured 4.1× faster than the map-generate-join idiom on twitter
+  statuses; `each_line` is 1.6× faster than line-split +
+  `JSON.parse`), rejecting formatting options that would break line
+  framing. File forms: `NOSJ.each_line_file` streams over a read-only
+  memory map, `NOSJ.write_lines` generates straight to disk and
+  returns the byte count. Parse options apply per line; pass a frozen
+  string to `each_line` for zero-copy iteration.
+- `NOSJ.stats(source, opts)` / `NOSJ.stats_file(path, opts)`: document
+  statistics from one counting pass through the null-sink machinery,
+  answering "what is this 40 MB blob" without building any Ruby values
+  for the document (measured ~1.3× faster than a full parse). Reports
+  `byte_size`, `root` kind, `max_depth`, value counts by type, key
+  totals, a key histogram sorted by count, largest container sizes,
+  and string byte totals. Nesting is unlimited by default (pass
+  `max_nesting` to enforce a limit); `allow_nan` and
+  `allow_trailing_comma` are honored; malformed documents raise the
+  rich `ParserError`. The file form memory-maps, so the document never
+  enters Ruby at all.
+- Rich parse errors. Parse failures now raise `NOSJ::ParserError`
+  (previously a bare `RuntimeError`) carrying the failure position,
+  computed only when a parse fails: `#byte_offset`, 1-based `#line`,
+  character-based `#column`, and a caret `#snippet` showing the
+  offending line (windowed when the line is long, as minified JSON
+  usually is). Positions are absolute within the document you passed,
+  including through partial parsing (`dig`, `at_pointer`, batches),
+  lazy documents, and the file APIs. `#detailed_message` appends the
+  snippet, so unrescued errors print it. Failures with no position
+  (encoding refusals) leave the accessors nil. Exceeding `max_nesting`
+  during parsing now raises `NOSJ::NestingError`, matching the gem's
+  class (generation already did); rescues of the old `RuntimeError`
+  need updating to `NOSJ::ParserError`/`NOSJ::Error`.
+- Rails mode: `require "nosj/rails"` accelerates a Rails application
+  in both directions. It installs a nosj-backed ActiveSupport JSON
+  encoder, so `obj.to_json`, `render json:`, and `ActiveSupport::JSON.encode` walk the object tree natively—values recurse through `as_json` exactly
+  like ActiveSupport's own encoder. It also loads the `nosj/json` drop-in, so
+  `ActiveSupport::JSON.decode` and JSON request-body parsing take the
+  fast path (including on Rails 7.x, whose `quirks_mode` option the
+  drop-in now accepts; the drop-in also accepts valid-UTF-8 BINARY
+  strings now, which is what Rack delivers request bodies as). The
+  HTML-safety escaping is fused into the SIMD string-emission kernels,
+  so escaped output costs the same single pass as unescaped. Measured
+  against stock ActiveSupport encoding: ×1.7 on small documents up to
+  ×5.2 on large trees and ×14 on HTML-heavy content
+  (`rake bench:rails`). In a Rails Gemfile:
+  `gem "nosj", require: "nosj/rails"`.
+- `JSON::Fragment` values now splice their pre-rendered JSON
+  everywhere the `json` gem does: in default mode, under `strict:
+  true`, and through the Rails encoder.
+
 ## [0.2.0] - 2026-07-16
 
 - File APIs. `NOSJ.load_file(path, opts)` parses a file directly

data/Cargo.lock

@@ -189,16 +189,16 @@ dependencies = [
 
 [[package]]
 name = "nosj"
-version = "0.1.1"
+version = "0.2.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c6ecf0db23b7e9944ecee15ab3dd01b0bde0ab2a195442f363847739f667ac8d"
+checksum = "d25913644b7871b36095cc3da27daf95ebf9508189c5d8ccfb16da3bdb9f802f"
 dependencies = [
  "fast-float2",
 ]
 
 [[package]]
 name = "nosj_native"
-version = "0.2.0"
+version = "0.3.1"
 dependencies = [
  "ahash",
  "magnus",

data/README.md

@@ -13,11 +13,14 @@ faster than Yajl—[see Benchmarks](#benchmarks).
 - It has **lazy documents**: `NOSJ.lazy` wraps a document and parses a value only when you touch it—repeated access costs nanoseconds, and everything you never read is never parsed.
 - It has a **partial parsing mode**: JSON Pointer lookups that pull single values out of big documents in microseconds, skipping everything else.
 - It has **file APIs**: parse, generate, dig, and lazy-wrap files directly—no throwaway file-sized Ruby String, and the partial modes memory-map the file so unread pages never even leave the disk.
+- It does **byte-splicing edits**: `NOSJ.splice` changes a field inside a passing payload 10–51× faster than parse-mutate-generate, and RFC 6902 JSON Patch / RFC 7386 merge patch apply straight to the raw string.
+- It is **great to debug with**: parse errors carry line, column, and a caret snippet pointing at the break, and `NOSJ.stats` X-rays a mystery blob (depth, value counts, key histogram) faster than parsing it.
+- It accelerates a **Rails** application in both encoding and decoding.
 - It comes **precompiled** (platform gems built with per-platform optimizations,
 nothing to compile on install).
 - Otherwise, same API and option names as gem json.
 
-**And there's more**: validate documents without building a single Ruby object, resolve whole batches of paths in one pass, and accelerate an entire application with a one-line drop-in.
+**And there's more**: validate documents without building a single Ruby object, minify and pretty-print without parsing, resolve whole batches of paths in one pass, stream NDJSON / JSON Lines in both directions, and accelerate an entire application with a one-line drop-in.
 
 - [Requirements](#requirements)
 - [Getting started](#getting-started)
@@ -26,6 +29,7 @@ nothing to compile on install).
 - [Switching from the json gem](#switching-from-the-json-gem)
 - [How it works](#how-it-works)
 - [Development](#development)
+- [Acknowledgements](#acknowledgements)
 - [License](#license)
 
 ## Requirements
@@ -50,6 +54,8 @@ NOSJ.generate({"a" => [1, true]})  #=> '{"a":[1,true]}'
 
 That's it—if you know the `json` gem, you already know `nosj`.
 
+### gem json compatibility
+
 Want the speedup without touching your code? One line reroutes
 `JSON.parse`, `JSON.generate`, `JSON.pretty_generate`, and `JSON.dump`
 through nosj:
@@ -65,18 +71,39 @@ Gemfile you can do this:
 gem "nosj", require: "nosj/json"
 ```
 
-Options nosj supports take the fast path; anything exotic
-(`create_additions`, `object_class`, `JSON::State`, procs, IO
-arguments) falls back to the original implementation, so `JSON.load`,
-`JSON.parse!`, and `JSON.load_file` keep their exact behavior.
-Exceptions re-raise as the JSON classes, so your rescue clauses keep
-working. Measured through the patch: parse 1.11×, generate 1.05× over
-the original gem. (A MultiJson adapter ships too:
-`require "nosj/multi_json"`, then `MultiJson.use NOSJ::MultiJsonAdapter`.)
+A MultiJson adapter ships too: 
+
+```ruby
+require "nosj/multi_json"
+```
+
+And then `MultiJson.use NOSJ::MultiJsonAdapter`.
+
+### Ruby on Rails
+
+In a Rails app, use this for "Rails mode":
+
+```ruby
+gem "nosj", require: "nosj/rails"
+```
+
+That installs a nosj-backed ActiveSupport JSON encoder, so `obj.to_json`, `render json:`, and `ActiveSupport::JSON.encode` walk the object tree natively:
+values that aren't JSON-native recurse through `as_json` exactly like
+ActiveSupport's own encoder, non-finite floats encode as `null`, and
+HTML-safety escaping (`escape_html_entities_in_json`) behaves
+identically—verified differentially against ActiveSupport's encoder.
+It loads the drop-in too, so `ActiveSupport::JSON.decode` and JSON
+request-body parsing ride the fast path. 
+
+Measured against ActiveSupport's own encoder: ×1.9 on small documents
+up to ×5.2 on large trees and ×14 on HTML-heavy content—see
+[Benchmarks → Rails mode](#rails-mode).
 
 ## What's in the box
 
-**The `json` gem API**, on the `NOSJ` module:
+### json API
+
+The `json` gem API, on the `NOSJ` module:
 
 ```ruby
 NOSJ.parse(src, symbolize_names: true)   # also: freeze, max_nesting,
@@ -85,7 +112,9 @@ NOSJ.generate(obj)                       # indent, space, object_nl, ...,
 NOSJ.pretty_generate(obj)                # ascii_only, script_safe, strict
 ```
 
-**Lazy documents.** `NOSJ.lazy` wraps a document in a lazy view: read
+### Lazy documents
+
+`NOSJ.lazy` wraps a document in a lazy view: read
 a field and only that path is parsed—containers stay lazy, scalars
 arrive as plain Ruby values, and repeated reads are cached:
 
@@ -101,8 +130,9 @@ Pass a frozen string and creating the view is practically free—
 nanoseconds, even on a megabyte document. Malformed content raises
 when it is first read, not at wrap time.
 
+### Partial parsing
 
-**Partial parsing.** Pull values out of a document without
+Pull values out of a document without
 materializing the rest—skipped content is stepped over at SIMD block
 speed, so a lookup costs what it skips, not what the document weighs:
 
@@ -122,7 +152,9 @@ at the far end of a 570 KB document costs ~71µs, still 13× faster
 than parse-then-dig. Misses return nil; matched subtrees materialize
 with the same options as `parse` (`symbolize_names:`, `freeze:`).
 
-**Files.** Every mode has a file-native form, so a document never
+### Files API
+
+Every mode has a file-native form, so a document never
 round-trips through a throwaway Ruby String:
 
 ```ruby
@@ -138,10 +170,89 @@ read are never loaded from disk. Missing files raise the usual
 `Errno` exceptions. Measured numbers live in
 [Benchmarks → File APIs](#file-apis).
 
-**Validation without parsing.** `NOSJ.valid?` runs the full
-parser—tokenizers, string decode, number validation—into a null sink
-and allocates no Ruby objects at all. It is 2-4× faster than
-`NOSJ.parse`, which already leads every parser above:
+### Reformat without parsing
+
+`NOSJ.minify` and `NOSJ.reformat` pipe the parser's events straight
+into the emission kernels: **zero Ruby objects** are built for the
+document (the specs assert it), SIMD in and SIMD out. Minifying the
+631 KB twitter.json takes 409µs—3.4× faster than a
+`generate(parse(x))` round-trip, 3.9× faster than gem json's cycle,
+and only 1.4× the cost of `valid?`:
+
+```ruby
+NOSJ.minify(json)                        # compact
+NOSJ.reformat(json, pretty: true)        # pretty_generate layout
+NOSJ.reformat(json, ascii_only: true)    # escape-transcode, no parse
+NOSJ.reformat_file("big.json")           # straight off a memory map
+```
+
+Output is exactly `generate(parse(json))`—canonical number spellings,
+normalized escapes, same formatting options—except duplicate keys pass
+through (a reformatter must not silently drop data) and lone-surrogate
+strings re-escape instead of raising. Acceptance options apply too:
+`minify(src, allow_trailing_comma: true)` normalizes the commas away.
+
+### Byte-splicing edits and JSON Patch
+
+Change a field in a passing payload without the parse → mutate →
+generate cycle: `NOSJ.splice` skips to the target spans (one pass for
+the whole batch) and rebuilds the string around them—every byte
+outside the targets is copied untouched, so formatting, key order,
+and number spellings survive exactly. On twitter.json this is
+**10×–51× faster** than parse-mutate-generate (129µs for a late
+field, 26µs for an early one, vs ~1.3ms):
+
+```ruby
+NOSJ.splice(json, "/config/timeout" => 30)
+NOSJ.splice(json, "/a" => 1, "/b/c" => [true])   # batch, one pass
+```
+
+On top of the same machinery: **RFC 6902 JSON Patch** applied to the
+raw string (`add`, `remove`, `replace`, `move`, `copy`, `test`—the
+whole appendix-A suite is in the specs), and **RFC 7386 JSON Merge
+Patch**:
+
+```ruby
+NOSJ.patch(json, [
+  {"op" => "test", "path" => "/a", "value" => 1},
+  {"op" => "replace", "path" => "/a", "value" => 2},
+  {"op" => "add", "path" => "/list/-", "value" => "x"}
+])
+NOSJ.merge_patch(json, {"config" => {"legacy" => nil, "timeout" => 30}})
+```
+
+Missing splice targets raise `KeyError` (use `patch` `add` to
+insert); failed patches raise `NOSJ::PatchError`; inserted values are
+byte-identical to `NOSJ.generate`'s output.
+
+### NDJSON / JSON Lines
+
+Logs, data pipelines, LLM batch files—one JSON document per line.
+`NOSJ.each_line` streams values out (1.6× faster than the
+line-split-and-`JSON.parse` idiom on twitter statuses), and
+`NOSJ.generate_lines` builds the whole stream in one buffer pass
+(4.1× faster than map-generate-join):
+
+```ruby
+NOSJ.each_line(log) { |event| ingest(event) }   # or an Enumerator:
+NOSJ.each_line(log).first(10)                   # walks only 10 lines
+
+NOSJ.generate_lines(events)          #=> %({"a":1}\n{"b":2}\n...)
+NOSJ.each_line_file("events.ndjson") { |e| }    # over a memory map
+NOSJ.write_lines("out.ndjson", events)          # straight to disk
+```
+
+Blank lines are skipped, one value per line is enforced, and a
+malformed line raises a [rich `ParserError`](#rich-parse-errors) whose
+`#line` is the physical line in the stream. Parse options apply per
+line—`NOSJ.each_line(log, freeze: true)` yields Ractor-shareable
+values.
+
+### Validation without parsing
+
+`NOSJ.valid?` runs the full parser—tokenizers, string decode, 
+number validation—into a null sink and allocates no Ruby objects
+at all. It is 2-4× faster than `NOSJ.parse`:
 
 ```ruby
 NOSJ.valid?('{"a":1}')                #=> true
@@ -149,6 +260,46 @@ NOSJ.valid?('{"a":}')                 #=> false
 NOSJ.valid?(src, max_nesting: false)  # same options as parse
 ```
 
+### Document statistics
+
+`NOSJ.stats` answers "what is this 40 MB blob": one counting pass 
+through the same null-sink machinery—no Ruby value is built for 
+the document, and it costs _less_ than a parse (~1.3× faster on `twitter.json`):
+
+```ruby
+s = NOSJ.stats(blob)          # or NOSJ.stats_file("huge.json")
+s[:byte_size]                 #=> 631514
+s[:root]                      #=> :object
+s[:max_depth]                 #=> 10
+s[:values]                    #=> {total: 13914, objects: 1264, arrays: 1050,
+                              #    strings: 4754, integers: 2108, ...}
+s[:keys]                      #=> {total: 13345, unique: 94}
+s[:key_histogram].first(3)    #=> [["id", 447], ["id_str", 447], ["urls", 364]]
+s[:containers]                #=> {max_object_entries: 40, max_array_length: 100}
+s[:strings]                   #=> {bytes: 200716, max_bytes: 463}
+```
+
+Nesting is unlimited by default (a deep blob is exactly what a
+diagnostic should describe); malformed documents raise the usual rich
+`ParserError`.
+
+### Rich parse errors
+
+A failed parse raises `NOSJ::ParserError`
+carrying where the document broke—`#byte_offset`, `#line`, `#column`,
+and a caret `#snippet`—computed only when a parse fails, so success
+pays nothing. Positions stay absolute through partial parsing, lazy
+documents, and the file APIs; unrescued errors print the snippet:
+
+```ruby
+NOSJ.parse(%({\n  "a": 1,\n  "b": }))
+# NOSJ::ParserError: unexpected character at byte 19
+#   e.line     #=> 3
+#   e.column   #=> 8
+#   e.snippet  #=>   "b": }
+#                         ^
+```
+
 ## Benchmarks
 
 Every installed JSON gem, benchmark-ips: AWS EC2 c7a.2xlarge (AMD EPYC 9R14, Zen 4), Ruby 4.0.6 + YJIT, json 2.21.1, Oj 3.17.4, RapidJSON 0.4.0, FastJsonparser 0.6.0, Yajl 1.4.3, PGO build, 2026-07-16. `×N` = times slower than nosj. 
@@ -209,6 +360,27 @@ Silicon dev box, Ruby 4.0.6 + YJIT, PGO build, 2026-07-16):
 too much for an honest multiplier; what it saves is the intermediate
 file-sized Ruby String.
 
+### Rails mode
+
+`ActiveSupport::JSON.encode` with the nosj encoder installed, against
+stock ActiveSupport (Apple Silicon dev box, Ruby 4.0.6 + YJIT,
+activesupport 8.1, medians of 5 interleaved per-process rounds,
+outputs verified byte-identical first, 2026-07-17;
+`rake bench:rails`):
+
+| workload | nosj (i/s) | vs ActiveSupport |
+|---|---:|---|
+| twitter tree (570 KB) | 4.5k | ×5.2 |
+| 100-record index (with timestamps) | 33.7k | ×1.7 |
+| HTML-heavy user content | 177.7k | ×14.2 |
+| Time/Date/BigDecimal hash | 1.1M | ×3.0 |
+| small API hash | 4.2M | ×1.9 |
+| small hash `to_json` | 4.1M | ×1.9 |
+
+The HTML-safety escaping that dominates stock encodes of
+user-generated content is fused into the SIMD string-emission kernels
+here: escaped output costs the same single pass as unescaped.
+
 Reproduce with `rake bench` (the parity-gated comparison, after a PGO retrain—the shipping configuration) or `rake bench:ips` (the multi-gem shoot-out).
 
 ## Switching from the json gem
@@ -222,8 +394,10 @@ You mostly don't have to do anything. Some differences:
   UTF-8) follow the strict semantics instead.
 - Unlike `Array#dig`, negative indices in `NOSJ.dig` return nil (JSON
   Pointer has no equivalent).
-- Parse error *messages* use byte offsets rather than the gem's
-  phrasing (classes match).
+- Parse errors raise `NOSJ::ParserError` (`NOSJ::NestingError` past
+  `max_nesting`, like the gem); messages use byte offsets rather than
+  the gem's phrasing, and the exception carries `#line`, `#column`,
+  and a caret `#snippet`.
 
 Everything else—including the gem's exact float formatting, which is
 not the shortest-round-trip form most libraries emit—matches
@@ -260,6 +434,33 @@ bundle exec rake bench:fast             # the sweep without retraining
 bundle exec rake "bench:ips[twitter]"   # multi-gem shoot-out (benchmark-ips); no args = full corpus
 ```
 
+The native layer is differentially fuzzed on an embedded VM
+(reformat pipe, NDJSON framing, splice/patch span arithmetic against
+pure-Ruby reference implementations)—see
+[ext/nosj/fuzz/README.md](ext/nosj/fuzz/README.md).
+
+## Acknowledgements
+
+Thanks to [Jean Boussier](https://github.com/byroot), [Florian Frank](https://github.com/flori), [Hiroshi Shibata](https://github.com/hsbt), [Nobuyoshi Nakada](https://github.com/nobu), [Étienne Barrié](https://github.com/etiennebarrie), and the other authors and maintainers of the [json gem](https://github.com/ruby/json)—for their work on the gem itself, for optimization ideas, and for some of the JSON documents in the benchmark corpus.
+
+Thanks to [Mat Sadler](https://github.com/matsadler) for [magnus](https://github.com/matsadler/magnus).
+
+Thanks to the authors of the projects credited in [NOTICE](NOTICE) and the
+crate's NOTICE: [fpconv](https://github.com/night-shift/fpconv) and
+Florian Loitsch's Grisu2,
+[fast_float](https://github.com/fastfloat/fast_float) and
+[fast_double_parser](https://github.com/lemire/fast_double_parser)
+(Daniel Lemire et al.),
+[simdjson](https://github.com/simdjson/simdjson) (Daniel Lemire, Geoff
+Langdale, and contributors), and the architectural lineage of
+[yyjson](https://github.com/ibireme/yyjson),
+[simd-json](https://github.com/simd-lite/simd-json), and
+[sonic-rs](https://github.com/cloudwego/sonic-rs).
+
+## Assisted by
+
+Claude Fable 5, Claude Opus 4.8.
+
 ## License
 
 MIT. The underlying Rust crate is `MIT AND BSL-1.0 AND Apache-2.0`; its

data/ext/nosj/Cargo.toml

@@ -5,7 +5,7 @@
 # required by lib/nosj.rb as "nosj/nosj" (Init_nosj set explicitly in
 # lib.rs). Gem name, module, and API are plain nosj.
 name = "nosj_native"
-version = "0.2.0"
+version = "0.3.1"
 edition = "2021"
 authors = ["Yaroslav Markin <yaroslav@markin.net>"]
 license = "MIT"
@@ -25,4 +25,4 @@ memmap2 = "0.9"
 # First-party SIMD JSON parse/generate library (github.com/yaroslav/nosj).
 # For coordinated crate+gem work, temporarily flip to
 # { path = "../../../nosj" } and restore before any commit or release.
-nosj = "0.1.1"
+nosj = "0.2.0"

data/ext/nosj/fuzz/Cargo.toml

@@ -0,0 +1,41 @@
+[package]
+name = "nosj-ruby-fuzz"
+version = "0.0.0"
+publish = false
+edition = "2021"
+
+[package.metadata]
+cargo-fuzz = true
+
+[dependencies]
+libfuzzer-sys = "0.4"
+# Must match the extension's own magnus line so cargo unifies the
+# rb-sys link (one libruby, one VM).
+magnus = { git = "https://github.com/matsadler/magnus", features = ["embed", "rb-sys"] }
+# The extension as an rlib; its lib target is named `nosj` (the bundle
+# name), so rename the dependency to keep fuzz code readable.
+nosj_ext = { path = "..", package = "nosj_native" }
+
+# Prevent this from being interpreted as part of a parent workspace.
+[workspace]
+
+[[bin]]
+name = "reformat"
+path = "fuzz_targets/reformat.rs"
+test = false
+doc = false
+bench = false
+
+[[bin]]
+name = "lines"
+path = "fuzz_targets/lines.rs"
+test = false
+doc = false
+bench = false
+
+[[bin]]
+name = "patch"
+path = "fuzz_targets/patch.rs"
+test = false
+doc = false
+bench = false

data/ext/nosj/fuzz/fuzz_targets/lines.rs

@@ -0,0 +1,9 @@
+//! NDJSON framing vs one parse per line, plus a generate_lines
+//! round-trip over whatever the streaming side yielded.
+#![no_main]
+
+use libfuzzer_sys::fuzz_target;
+
+fuzz_target!(|data: &[u8]| {
+    nosj_ruby_fuzz::drive("lines_case", data);
+});

data/ext/nosj/fuzz/fuzz_targets/patch.rs

@@ -0,0 +1,10 @@
+//! Byte-splicing edits and RFC 6902 patch vs a pure-Ruby tree
+//! reference. Input is "document \0 spec": an object spec drives
+//! splice, an array spec drives patch.
+#![no_main]
+
+use libfuzzer_sys::fuzz_target;
+
+fuzz_target!(|data: &[u8]| {
+    nosj_ruby_fuzz::drive("patch_case", data);
+});

data/ext/nosj/fuzz/fuzz_targets/reformat.rs

@@ -0,0 +1,10 @@
+//! Reformat pipe vs whole-document parse: acceptance must agree
+//! (valid?/stats included), minify must round-trip and be idempotent,
+//! pretty must round-trip.
+#![no_main]
+
+use libfuzzer_sys::fuzz_target;
+
+fuzz_target!(|data: &[u8]| {
+    nosj_ruby_fuzz::drive("reformat_case", data);
+});

data/ext/nosj/fuzz/src/lib.rs

@@ -0,0 +1,48 @@
+//! Shared harness for the extension fuzz targets: boot one embedded
+//! Ruby VM per process, register the extension exactly as `require
+//! "nosj/nosj"` would, and load the Ruby-side differential checks.
+//!
+//! The targets exercise the gem-side byte-manipulation layers the
+//! parser crate's own fuzzers never see (reformat, NDJSON framing,
+//! splice/patch span arithmetic). The checking logic lives in
+//! prelude.rb; a raised Ruby exception surfaces here as an Err and
+//! aborts the run.
+
+use std::sync::Once;
+
+use magnus::prelude::*;
+use magnus::{RModule, Ruby, Value};
+
+static VM: Once = Once::new();
+
+const PRELUDE: &str = include_str!("prelude.rb");
+
+/// Run one fuzz iteration with the VM up and the checks defined.
+/// libFuzzer drives every input on the same thread, which is the
+/// thread the VM was initialized on.
+pub fn with_vm<F: FnOnce(&Ruby, RModule)>(f: F) {
+    VM.call_once(|| {
+        let cluster = unsafe { magnus::embed::init() };
+        // The VM must outlive every future iteration.
+        std::mem::forget(cluster);
+        let ruby = Ruby::get().expect("VM just initialized on this thread");
+        unsafe { nosj_ext::Init_nosj() };
+        ruby.eval::<Value>(PRELUDE).expect("fuzz prelude must load");
+    });
+    let ruby = Ruby::get().expect("fuzz iterations run on the VM thread");
+    let checks = ruby
+        .define_module("NOSJFuzz")
+        .expect("prelude defines NOSJFuzz");
+    f(&ruby, checks);
+}
+
+/// Feed raw bytes to a single-argument check; any raised exception
+/// (an assertion failure or an unexpected error class) aborts.
+pub fn drive(check: &str, data: &[u8]) {
+    with_vm(|ruby, checks| {
+        let input = ruby.str_from_slice(data);
+        if let Err(e) = checks.funcall::<_, _, Value>(check, (input,)) {
+            panic!("{check}: {e}");
+        }
+    });
+}