swap regex backend to librure (rust-lang/regex c abi) — 27x faster than posix regex.h

.github/workflows/ci.yml

@@ -44,13 +44,18 @@ jobs:
             libzstd-dev zlib1g-dev libsqlite3-dev libcurl4-openssl-dev libpq-dev
           sudo ln -sf /usr/bin/clang-21 /usr/bin/clang
           sudo ln -sf /usr/bin/llvm-config-21 /usr/bin/llvm-config
+          # Rust toolchain for vendor build of librure (rust-lang/regex C ABI).
+          # Build-time dep only — end users installing the release tarball
+          # receive a prebuilt librure.a and never need rustc.
+          curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y --default-toolchain stable --profile minimal
+          echo "$HOME/.cargo/bin" >> $GITHUB_PATH
       - run: npm install
       - uses: actions/cache@v4
         with:
           path: |
             vendor/
             c_bridges/*.o
-          key: vendor-${{ runner.os }}-${{ hashFiles('scripts/build-vendor.sh', 'scripts/vendor-pins.sh', 'c_bridges/*.c', 'c_bridges/*.cpp') }}
+          key: vendor-rure1-${{ runner.os }}-${{ hashFiles('scripts/build-vendor.sh', 'scripts/vendor-pins.sh', 'c_bridges/*.c', 'c_bridges/*.cpp') }}
       - run: bash scripts/build-vendor.sh
       - run: npm run build
       - name: Build tree-sitter TSX objects
@@ -121,7 +126,15 @@ jobs:
               return f' (\u2191{pct:.0f}%)'
           compute = {k: b for k, b in d['benchmarks'].items() if b.get('category') != 'cli'}
           cli = {k: b for k, b in d['benchmarks'].items() if b.get('category') == 'cli'}
-          for k, b in sorted(compute.items(), key=lambda x: x[1]['name']):
+          def rank_sort_key(b):
+              r = b['results']
+              lower = b.get('lower_is_better', True)
+              chad_v = r.get('chadscript', {}).get('value')
+              if chad_v is None:
+                  return (99, b['name'])
+              vals = sorted([v['value'] for v in r.values()], reverse=not lower)
+              return (vals.index(chad_v) + 1, b['name'])
+          for k, b in sorted(compute.items(), key=lambda x: rank_sort_key(x[1])):
               r = b['results']
               def g(lang):
                   v = r.get(lang, {}).get('label', '—')
@@ -132,7 +145,7 @@ jobs:
               cli_names = {'grep': 'grep', 'ripgrep': 'ripgrep', 'node': 'Node.js', 'xxd': 'xxd'}
               lines.append('')
               lines.append('### CLI Tool Benchmarks\n')
-              for k, b in sorted(cli.items(), key=lambda x: x[1]['name']):
+              for k, b in sorted(cli.items(), key=lambda x: rank_sort_key(x[1])):
                   r = b['results']
                   competitors = [l for l in r if l != 'chadscript']
                   chad_v = r.get('chadscript', {}).get('label', '—')
@@ -203,6 +216,11 @@ jobs:
             libzstd-dev zlib1g-dev libsqlite3-dev libcurl4-openssl-dev libpq-dev
           sudo ln -sf /usr/bin/clang-21 /usr/bin/clang
           sudo ln -sf /usr/bin/llvm-config-21 /usr/bin/llvm-config
+          # Rust toolchain for vendor build of librure (rust-lang/regex C ABI).
+          # Build-time dep only — end users installing the release tarball
+          # receive a prebuilt librure.a and never need rustc.
+          curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y --default-toolchain stable --profile minimal
+          echo "$HOME/.cargo/bin" >> $GITHUB_PATH
           clang --version 2>&1 | head -2
 
       - name: Install npm dependencies
@@ -214,7 +232,7 @@ jobs:
           path: |
             vendor/
             c_bridges/*.o
-          key: vendor-${{ runner.os }}-${{ hashFiles('scripts/build-vendor.sh', 'scripts/vendor-pins.sh', 'c_bridges/*.c', 'c_bridges/*.cpp') }}
+          key: vendor-rure1-${{ runner.os }}-${{ hashFiles('scripts/build-vendor.sh', 'scripts/vendor-pins.sh', 'c_bridges/*.c', 'c_bridges/*.cpp') }}
 
       - name: Build vendor libraries
         run: bash scripts/build-vendor.sh
@@ -225,7 +243,7 @@ jobs:
       - name: Verify vendor libraries
         run: |
           fail=0
-          for lib in vendor/bdwgc/libgc.a vendor/yyjson/libyyjson.a vendor/libuv/build/libuv.a vendor/picohttpparser/picohttpparser.o c_bridges/lws-bridge.o c_bridges/multipart-bridge.o c_bridges/regex-bridge.o c_bridges/child-process-bridge.o c_bridges/child-process-spawn.o c_bridges/trampoline-bridge.o c_bridges/os-bridge.o c_bridges/strlen-cache.o c_bridges/time-bridge.o c_bridges/base64-bridge.o c_bridges/url-bridge.o c_bridges/uri-bridge.o c_bridges/dotenv-bridge.o c_bridges/watch-bridge.o c_bridges/arena-bridge.o c_bridges/curl-bridge.o c_bridges/pg-bridge.o c_bridges/compress-bridge.o c_bridges/yaml-bridge.o c_bridges/string-ops-bridge.o c_bridges/llvm-bridge.o c_bridges/llvm-builder-bridge.o c_bridges/lld-bridge.o; do
+          for lib in vendor/bdwgc/libgc.a vendor/yyjson/libyyjson.a vendor/libuv/build/libuv.a vendor/picohttpparser/picohttpparser.o vendor/rure/librure.a c_bridges/lws-bridge.o c_bridges/multipart-bridge.o c_bridges/regex-bridge.o c_bridges/child-process-bridge.o c_bridges/child-process-spawn.o c_bridges/trampoline-bridge.o c_bridges/os-bridge.o c_bridges/strlen-cache.o c_bridges/time-bridge.o c_bridges/base64-bridge.o c_bridges/url-bridge.o c_bridges/uri-bridge.o c_bridges/dotenv-bridge.o c_bridges/watch-bridge.o c_bridges/arena-bridge.o c_bridges/curl-bridge.o c_bridges/pg-bridge.o c_bridges/compress-bridge.o c_bridges/yaml-bridge.o c_bridges/string-ops-bridge.o c_bridges/llvm-bridge.o c_bridges/llvm-builder-bridge.o c_bridges/lld-bridge.o; do
             if [ ! -f "$lib" ]; then
               echo "MISSING: $lib"
               fail=1
@@ -289,6 +307,7 @@ jobs:
           cp c_bridges/lws-bridge.o release/lib/
           cp c_bridges/multipart-bridge.o release/lib/
           cp c_bridges/regex-bridge.o release/lib/
+          cp vendor/rure/librure.a release/lib/
           cp c_bridges/child-process-bridge.o release/lib/
           cp c_bridges/child-process-spawn.o release/lib/
           cp c_bridges/trampoline-bridge.o release/lib/
@@ -330,6 +349,13 @@ jobs:
           echo "/opt/homebrew/opt/llvm/bin" >> $GITHUB_PATH
           echo "/opt/homebrew/opt/postgresql@16/bin" >> $GITHUB_PATH
           echo "/usr/local/opt/llvm/bin" >> $GITHUB_PATH
+          # Rust toolchain for vendor build of librure (rust-lang/regex C ABI).
+          # Build-time dep only — end users get a prebuilt librure.a in the
+          # release tarball and never need rustc.
+          if ! command -v cargo >/dev/null 2>&1; then
+            curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y --default-toolchain stable --profile minimal
+            echo "$HOME/.cargo/bin" >> $GITHUB_PATH
+          fi
 
       - name: Start and provision postgres
         run: |
@@ -356,7 +382,7 @@ jobs:
           path: |
             vendor/
             c_bridges/*.o
-          key: vendor-${{ runner.os }}-${{ hashFiles('scripts/build-vendor.sh', 'scripts/vendor-pins.sh', 'c_bridges/*.c', 'c_bridges/*.cpp') }}
+          key: vendor-rure1-${{ runner.os }}-${{ hashFiles('scripts/build-vendor.sh', 'scripts/vendor-pins.sh', 'c_bridges/*.c', 'c_bridges/*.cpp') }}
 
       - name: Build vendor libraries
         run: bash scripts/build-vendor.sh
@@ -367,7 +393,7 @@ jobs:
       - name: Verify vendor libraries
         run: |
           fail=0
-          for lib in vendor/bdwgc/libgc.a vendor/yyjson/libyyjson.a vendor/libuv/build/libuv.a vendor/picohttpparser/picohttpparser.o c_bridges/lws-bridge.o c_bridges/multipart-bridge.o c_bridges/regex-bridge.o c_bridges/child-process-bridge.o c_bridges/child-process-spawn.o c_bridges/trampoline-bridge.o c_bridges/os-bridge.o c_bridges/strlen-cache.o c_bridges/time-bridge.o c_bridges/base64-bridge.o c_bridges/url-bridge.o c_bridges/uri-bridge.o c_bridges/dotenv-bridge.o c_bridges/watch-bridge.o c_bridges/arena-bridge.o c_bridges/curl-bridge.o c_bridges/pg-bridge.o c_bridges/compress-bridge.o c_bridges/yaml-bridge.o c_bridges/string-ops-bridge.o c_bridges/llvm-bridge.o c_bridges/llvm-builder-bridge.o c_bridges/lld-bridge.o; do
+          for lib in vendor/bdwgc/libgc.a vendor/yyjson/libyyjson.a vendor/libuv/build/libuv.a vendor/picohttpparser/picohttpparser.o vendor/rure/librure.a c_bridges/lws-bridge.o c_bridges/multipart-bridge.o c_bridges/regex-bridge.o c_bridges/child-process-bridge.o c_bridges/child-process-spawn.o c_bridges/trampoline-bridge.o c_bridges/os-bridge.o c_bridges/strlen-cache.o c_bridges/time-bridge.o c_bridges/base64-bridge.o c_bridges/url-bridge.o c_bridges/uri-bridge.o c_bridges/dotenv-bridge.o c_bridges/watch-bridge.o c_bridges/arena-bridge.o c_bridges/curl-bridge.o c_bridges/pg-bridge.o c_bridges/compress-bridge.o c_bridges/yaml-bridge.o c_bridges/string-ops-bridge.o c_bridges/llvm-bridge.o c_bridges/llvm-builder-bridge.o c_bridges/lld-bridge.o; do
             if [ ! -f "$lib" ]; then
               echo "MISSING: $lib"
               fail=1
@@ -435,6 +461,7 @@ jobs:
           cp c_bridges/lws-bridge.o release/lib/
           cp c_bridges/multipart-bridge.o release/lib/
           cp c_bridges/regex-bridge.o release/lib/
+          cp vendor/rure/librure.a release/lib/
           cp c_bridges/child-process-bridge.o release/lib/
           cp c_bridges/child-process-spawn.o release/lib/
           cp c_bridges/trampoline-bridge.o release/lib/
@@ -512,6 +539,7 @@ jobs:
             lws-bridge.o \
             multipart-bridge.o \
             regex-bridge.o \
+            librure.a \
             child-process-bridge.o \
             child-process-spawn.o \
             trampoline-bridge.o \

.github/workflows/cross-compile.yml

@@ -25,6 +25,12 @@ jobs:
           sudo apt-get install -y llvm clang lld cmake make gcc g++ \
             autoconf automake libtool linux-headers-generic git \
             libzstd-dev zlib1g-dev libsqlite3-dev libcurl4-openssl-dev file
+          # Rust for librure (rust-lang/regex C ABI). Build-time dep only.
+          # NOTE: this builds librure for the SDK's host arch (linux-x64).
+          # Cross-arch SDKs (e.g. linux-arm64) will need `rustup target add`
+          # — tracked as a follow-up; current cross-compile path is x64-only.
+          curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y --default-toolchain stable --profile minimal
+          echo "$HOME/.cargo/bin" >> $GITHUB_PATH
 
       - name: Install npm dependencies
         run: npm install
@@ -35,7 +41,7 @@ jobs:
           path: |
             vendor/
             c_bridges/*.o
-          key: vendor-${{ runner.os }}-${{ hashFiles('scripts/build-vendor.sh', 'scripts/vendor-pins.sh', 'c_bridges/*.c') }}
+          key: vendor-rure1-${{ runner.os }}-${{ hashFiles('scripts/build-vendor.sh', 'scripts/vendor-pins.sh', 'c_bridges/*.c') }}
 
       - name: Build vendor libraries
         run: bash scripts/build-vendor.sh
@@ -81,6 +87,11 @@ jobs:
           echo "/usr/local/opt/llvm/bin" >> $GITHUB_PATH
           echo "/opt/homebrew/opt/lld/bin" >> $GITHUB_PATH
           echo "/usr/local/opt/lld/bin" >> $GITHUB_PATH
+          # Rust for librure (rust-lang/regex C ABI). Build-time dep only.
+          if ! command -v cargo >/dev/null 2>&1; then
+            curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y --default-toolchain stable --profile minimal
+            echo "$HOME/.cargo/bin" >> $GITHUB_PATH
+          fi
 
       - name: Install npm dependencies
         run: npm install
@@ -91,7 +102,7 @@ jobs:
           path: |
             vendor/
             c_bridges/*.o
-          key: vendor-${{ runner.os }}-${{ hashFiles('scripts/build-vendor.sh', 'scripts/vendor-pins.sh', 'c_bridges/*.c') }}
+          key: vendor-rure1-${{ runner.os }}-${{ hashFiles('scripts/build-vendor.sh', 'scripts/vendor-pins.sh', 'c_bridges/*.c') }}
 
       - name: Build vendor libraries
         run: bash scripts/build-vendor.sh

.github/workflows/update-benchmarks.yml

@@ -30,13 +30,16 @@ jobs:
             libzstd-dev zlib1g-dev libsqlite3-dev libcurl4-openssl-dev
           sudo ln -sf /usr/bin/clang-21 /usr/bin/clang
           sudo ln -sf /usr/bin/llvm-config-21 /usr/bin/llvm-config
+          # rustc/cargo for librure (rust-lang/regex C ABI) vendor build.
+          curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y --default-toolchain stable --profile minimal
+          echo "$HOME/.cargo/bin" >> $GITHUB_PATH
       - run: npm install
       - uses: actions/cache@v4
         with:
           path: |
             vendor/
             c_bridges/*.o
-          key: vendor-${{ runner.os }}-${{ hashFiles('scripts/build-vendor.sh', 'scripts/vendor-pins.sh', 'c_bridges/*.c', 'c_bridges/*.cpp') }}
+          key: vendor-rure1-${{ runner.os }}-${{ hashFiles('scripts/build-vendor.sh', 'scripts/vendor-pins.sh', 'c_bridges/*.c', 'c_bridges/*.cpp') }}
       - run: bash scripts/build-vendor.sh
       - run: npm run build
       - name: Build tree-sitter TSX objects

BUILDING.md

@@ -29,6 +29,18 @@ sudo dnf install cmake autoconf automake libtool nodejs npm
 brew install cmake autoconf automake libtool node
 ```
 
+**Rust toolchain** (only needed when (re)building vendor libraries):
+
+ChadScript's regex engine is `librure` — the C ABI for Rust's `regex`
+crate. Vendor builds compile it from source via cargo. End users
+installing via the release tarball receive a prebuilt `librure.a` and
+do **not** need Rust.
+
+```bash
+# Any OS (one-time, ~60s):
+curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y --default-toolchain stable --profile minimal
+```
+
 ## Build
 
 ```bash
@@ -38,7 +50,7 @@ bash scripts/build-vendor.sh
 npm run build
 ```
 
-`scripts/build-vendor.sh` clones and builds static archives for libgc, cJSON, libuv, tree-sitter, and libwebsockets into `vendor/`. It's idempotent — re-running skips already-built libraries.
+`scripts/build-vendor.sh` clones and builds static archives for libgc, cJSON, libuv, tree-sitter, libwebsockets, and `librure` (Rust regex C ABI) into `vendor/`. It's idempotent — re-running skips already-built libraries.
 
 ## Run
 

benchmarks/assemble_json.py

@@ -51,8 +51,10 @@
     "stringops":     {"name": "String Manipulation", "desc": "Concatenate 100K strings, split, toUpperCase, join.", "metric": "s", "lower_is_better": True},
     "fileio":        {"name": "File I/O", "desc": "Write and read ~100MB to /tmp.", "metric": "s", "lower_is_better": True},
     "binarytrees":   {"name": "Binary Trees", "desc": "Build/check/discard binary trees of depth 18.", "metric": "s", "lower_is_better": True},
-    "json":          {"name": "JSON Parse/Stringify", "desc": "Parse 10K JSON objects, stringify back.", "metric": "s", "lower_is_better": True},
+    "json":          {"name": "JSON Parse/Stringify", "desc": "Parse 100K JSON objects, stringify back.", "metric": "s", "lower_is_better": True},
     "stringsearch":  {"name": "String Search", "desc": "Recursive directory search for 'console.log' in src/.", "metric": "s", "lower_is_better": True},
+    "regex_match":   {"name": "Regex Match", "desc": "100K matches of an anchored pattern with one capture group.", "metric": "s", "lower_is_better": True},
+    "map_lookup":    {"name": "Hash Map Lookup", "desc": "100K-entry Map<string,number>, 1M random .get() lookups.", "metric": "s", "lower_is_better": True},
     "cligrep":       {"name": "Recursive Grep", "desc": "cgrep vs grep — search for 'function' across 5x copies of src/.", "metric": "s", "lower_is_better": True, "category": "cli"},
     "clihex":        {"name": "Hex Dump", "desc": "chex vs xxd — hex dump a 5MB binary file.", "metric": "s", "lower_is_better": True, "category": "cli"},
 }

benchmarks/json/bench.c

@@ -4,7 +4,7 @@
 #include <time.h>
 #include "yyjson.h"
 
-#define COUNT 10000
+#define COUNT 100000
 
 typedef struct {
     int id;

benchmarks/json/chadscript.ts

@@ -1,4 +1,8 @@
-const COUNT = 10000;
+// Parse 100k JSON objects (~10MB total throughput), then stringify them
+// all back. Exercises the parser/serializer hot paths at a scale where
+// SIMD-friendly engines (yyjson, V8) actually pull away from naive ones.
+
+const COUNT = 100000;
 
 interface Item {
   id: number;

benchmarks/json/json_bench.go

@@ -6,7 +6,7 @@ import (
 	"time"
 )
 
-const COUNT = 10000
+const COUNT = 100000
 
 type Item struct {
 	ID     int     `json:"id"`

benchmarks/json/node.mjs

@@ -1,4 +1,4 @@
-const COUNT = 10000;
+const COUNT = 100000;
 
 const jsonStrings = [];
 for (let i = 0; i < COUNT; i++) {

benchmarks/map_lookup/bench.c

@@ -0,0 +1,72 @@
+// C reference: small open-addressing hash map (FNV-1a, linear probe).
+// String keys, int values. Same N/Q as the other implementations.
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <time.h>
+
+#define N 100000
+#define Q 1000000
+#define CAP (N * 2)
+
+typedef struct {
+    char *key;
+    int   val;
+    int   used;
+} Slot;
+
+static Slot table[CAP];
+
+static unsigned long fnv1a(const char *s) {
+    unsigned long h = 1469598103934665603UL;
+    while (*s) {
+        h ^= (unsigned char)*s++;
+        h *= 1099511628211UL;
+    }
+    return h;
+}
+
+static void map_set(const char *k, int v) {
+    unsigned long h = fnv1a(k) % CAP;
+    while (table[h].used) {
+        if (strcmp(table[h].key, k) == 0) { table[h].val = v; return; }
+        h = (h + 1) % CAP;
+    }
+    table[h].key = strdup(k);
+    table[h].val = v;
+    table[h].used = 1;
+}
+
+static int map_get(const char *k, int *found) {
+    unsigned long h = fnv1a(k) % CAP;
+    while (table[h].used) {
+        if (strcmp(table[h].key, k) == 0) { *found = 1; return table[h].val; }
+        h = (h + 1) % CAP;
+    }
+    *found = 0;
+    return 0;
+}
+
+int main(void) {
+    char buf[32];
+    for (int i = 0; i < N; i++) {
+        snprintf(buf, sizeof(buf), "key%d", i);
+        map_set(buf, i);
+    }
+
+    struct timespec t0, t1;
+    clock_gettime(CLOCK_MONOTONIC, &t0);
+    long long sum = 0;
+    for (int q = 0; q < Q; q++) {
+        snprintf(buf, sizeof(buf), "key%d", q % N);
+        int found;
+        int v = map_get(buf, &found);
+        if (found) sum += v;
+    }
+    clock_gettime(CLOCK_MONOTONIC, &t1);
+    double elapsed = (t1.tv_sec - t0.tv_sec) + (t1.tv_nsec - t0.tv_nsec) / 1e9;
+
+    printf("Sum:      %lld\n", sum);
+    printf("Time:     %.6fs\n", elapsed);
+    return 0;
+}

benchmarks/map_lookup/chadscript.ts

@@ -0,0 +1,22 @@
+const N = 100000;
+const Q = 1000000;
+
+const m = new Map<string, number>();
+let i = 0;
+while (i < N) {
+  m.set("key" + i, i);
+  i = i + 1;
+}
+
+const start = Date.now();
+let sum = 0;
+let q = 0;
+while (q < Q) {
+  const v = m.get("key" + (q % N));
+  if (v !== undefined) sum = sum + v;
+  q = q + 1;
+}
+const elapsed = (Date.now() - start) / 1000;
+
+console.log("Sum:      " + sum);
+console.log("Time:     " + elapsed + "s");

benchmarks/map_lookup/map_lookup.go

@@ -0,0 +1,30 @@
+package main
+
+import (
+	"fmt"
+	"strconv"
+	"time"
+)
+
+const N = 100000
+const Q = 1000000
+
+func main() {
+	m := make(map[string]int, N)
+	for i := 0; i < N; i++ {
+		m["key"+strconv.Itoa(i)] = i
+	}
+
+	start := time.Now()
+	sum := 0
+	for q := 0; q < Q; q++ {
+		v, ok := m["key"+strconv.Itoa(q%N)]
+		if ok {
+			sum += v
+		}
+	}
+	elapsed := time.Since(start).Seconds()
+
+	fmt.Printf("Sum:      %d\n", sum)
+	fmt.Printf("Time:     %.6fs\n", elapsed)
+}