code-gauge 6.0.0 → 7.0.0

package/README.md

@@ -375,14 +375,10 @@ supporting arbitrary grammars would mean shipping incomplete metrics for them.
 
 Parsing and every metric pass run in a Rust addon (tree-sitter); the thin TypeScript layer handles
 the CLI, cross-file matching, and the Halstead float derivations. Releases include prebuilt addons
-for Linux x64/arm64 (glibc and musl), macOS x64/arm64, and Windows x64. On other platforms,
-`postinstall` builds the addon from the bundled sources, which requires a
-[Rust toolchain](https://rustup.rs).
-
-If a prebuilt addon is unavailable, package managers that block dependency install scripts by
-default (recent npm versions, and Bun unless `code-gauge` is listed in `trustedDependencies`) skip
-the fallback build. Either approve `code-gauge`'s install script, or build the addon manually
-inside the installed package (the runtime error message points here too):
+for Linux x64/arm64 (glibc and musl), macOS x64/arm64, and Windows x64, and a release that
+includes them runs no install script. On other platforms, build the addon from the bundled sources inside the installed
+package, which requires a [Rust toolchain](https://rustup.rs) (the runtime error message points
+here too):
 
 ```sh
 node node_modules/code-gauge/scripts/buildNative.mjs

package/dist/nativeMetrics.cjs

@@ -1,4 +1,4 @@
 "use strict";let e=require("@willbooster/shared-lib");function t(e,t,n,i,a=!1){return JSON.parse(f().measureCodeNative(...r(e,t,n,i,a)))}function n(e,t,n,i,a=!1){return f().measureCodeNativeAsync(...r(e,t,n,i,a)).then(e=>JSON.parse(e))}function r(e,t,n,r,i){return[o(e),t,n,s(r?.minTokens),s(r?.maxGapTokens),s(r?.minSimilarityPercent),i]}function i(e,t,n){return JSON.parse(f().collectCrossFileDataNative(o(e),t,s(n)))}function a(e,t){return JSON.parse(f().collectFunctionTokenSequencesNative(o(e),t)).map(e=>Int32Array.from(e))}function o(e){return e.isWellFormed()?e:e.toWellFormed()}function s(t){return t===void 0||Number.isNaN(t)?void 0:(0,e.clamp)(Math.trunc(t),0,4294967295)}var c=class extends Error{};let l,u;function d(e){l=e}function f(){if(l)return l;if(u)throw u;let t=process.getBuiltinModule(`node:module`).createRequire(require("url").pathToFileURL(__filename).href),n=[`code-gauge-${p()}`,`../native/code-gauge.node`],r=[];for(let i of n){let n;try{n=t(i)}catch(t){r.push(`  ${i}: ${(0,e.getErrorMessage)(t).split(`
-`)[0]}`);continue}let a=n.payloadVersion?.();if(a!==8){r.push(`  ${i}: payload version ${a??`unknown`} does not match the expected 8; rebuild the addon with \`bun run build-native\``);continue}return l=n,n}throw u=new c(`The code-gauge native addon is not available for ${p()}. Build it with \`node scripts/buildNative.mjs\` in the code-gauge package directory (requires a Rust toolchain); when installing with npm, also allow install scripts for code-gauge so its postinstall build can run.\n${r.join(`
+`)[0]}`);continue}let a=n.payloadVersion?.();if(a!==8){r.push(`  ${i}: payload version ${a??`unknown`} does not match the expected 8; rebuild the addon with \`bun run build-native\``);continue}return l=n,n}throw u=new c(`The code-gauge native addon is not available for ${p()}. Build it with \`node scripts/buildNative.mjs\` in the code-gauge package directory (requires a Rust toolchain).\n${r.join(`
 `)}`),u}function p(){let e=`${process.platform}-${process.arch}`;return process.platform===`win32`?`${e}-msvc`:process.platform===`linux`?(process.report?.getReport())?.header?.glibcVersionRuntime?`${e}-gnu`:`${e}-musl`:e}exports.NativeAddonError=c,exports.collectCrossFileDataNative=i,exports.collectFunctionTokenSequencesNative=a,exports.measureCodeNative=t,exports.measureCodeNativeAsync=n,exports.setNativeBinding=d;
 //# sourceMappingURL=nativeMetrics.cjs.map

package/dist/nativeMetrics.cjs.map

@@ -1 +1 @@
-{"version":3,"file":"nativeMetrics.cjs","names":["clamp","getErrorMessage"],"sources":["../src/nativeMetrics.ts"],"sourcesContent":["import { clamp, getErrorMessage } from '@willbooster/shared-lib';\nimport type { CrossFileDuplicateCandidate, Token, TokenRange } from './duplication.js';\nimport type { CodeMetrics, DuplicationOptions, FunctionMetrics } from './types.js';\n\n/**\n * Halstead counts measured natively; the derived float metrics (volume, effort, ...) are\n * computed in TypeScript because they involve transcendental functions (log2) whose last-bit\n * results can differ between V8 and Rust's libm, and results must not depend on the Rust side's\n * libm build.\n */\nexport interface NativeHalsteadCounts {\n  distinctOperators: number;\n  distinctOperands: number;\n  totalOperators: number;\n  totalOperands: number;\n}\n\nexport interface NativeFunctionMetricsPayload extends Omit<FunctionMetrics, 'halstead'> {\n  halsteadCounts: NativeHalsteadCounts;\n}\n\nexport interface NativeMetricsPayload extends Omit<CodeMetrics, 'halstead' | 'functions' | 'syntaxTree'> {\n  functions: NativeFunctionMetricsPayload[];\n  halsteadCounts: NativeHalsteadCounts;\n  syntaxTree?: string;\n  crossFileData?: NativeCrossFileDataPayload;\n}\n\n/** One file's cross-file clone-detection contribution as serialized by the native addon. */\nexport interface NativeCrossFileDataPayload {\n  candidates: CrossFileDuplicateCandidate[];\n  tokens: Token[];\n  containerStatements: TokenRange[][];\n  nearMissBlocks: TokenRange[];\n  /** 1-based lines that are neither blank nor comment-only, sorted ascending. */\n  codeLineNumbers: number[];\n}\n\ntype NativeMeasureArguments = [\n  code: string,\n  language: string,\n  includeSyntaxTree: boolean,\n  minTokens: number | undefined,\n  maxGapTokens: number | undefined,\n  minSimilarityPercent: number | undefined,\n  includeCrossFileData: boolean,\n];\n\nexport interface NativeBinding {\n  measureCodeNative(...args: NativeMeasureArguments): string;\n  measureCodeNativeAsync(...args: NativeMeasureArguments): Promise<string>;\n  collectCrossFileDataNative(code: string, language: string, minTokens?: number): string;\n  collectFunctionTokenSequencesNative(code: string, language: string): string;\n  payloadVersion?(): number;\n}\n\n/**\n * Must equal `payload_version` in native/src/lib.rs. A previously built addon survives a\n * `git pull` untouched, so without this handshake it would silently return payloads missing\n * newer fields, or lack newer binding functions, instead of failing with a clear rebuild message.\n */\nexport const expectedPayloadVersion = 8;\n\n/**\n * Measures one file via the native addon, returning the raw payload for assembly in metrics.ts;\n * with `includeCrossFileData`, the payload also carries the file's cross-file contribution.\n */\nexport function measureCodeNative(\n  code: string,\n  language: string,\n  includeSyntaxTree: boolean,\n  duplication?: DuplicationOptions,\n  includeCrossFileData = false\n): NativeMetricsPayload {\n  return JSON.parse(\n    loadBinding().measureCodeNative(\n      ...toNativeMeasureArguments(code, language, includeSyntaxTree, duplication, includeCrossFileData)\n    )\n  ) as NativeMetricsPayload;\n}\n\n/**\n * measureCodeNative on the addon's worker threads, so several files can be measured in parallel.\n * A missing addon still throws synchronously, like measureCodeNative.\n */\nexport function measureCodeNativeAsync(\n  code: string,\n  language: string,\n  includeSyntaxTree: boolean,\n  duplication?: DuplicationOptions,\n  includeCrossFileData = false\n): Promise<NativeMetricsPayload> {\n  return loadBinding()\n    .measureCodeNativeAsync(\n      ...toNativeMeasureArguments(code, language, includeSyntaxTree, duplication, includeCrossFileData)\n    )\n    .then((json) => JSON.parse(json) as NativeMetricsPayload);\n}\n\n/** The binding arguments shared by the sync and async measurements, so both measure alike. */\nfunction toNativeMeasureArguments(\n  code: string,\n  language: string,\n  includeSyntaxTree: boolean,\n  duplication: DuplicationOptions | undefined,\n  includeCrossFileData: boolean\n): NativeMeasureArguments {\n  return [\n    toWellFormed(code),\n    language,\n    includeSyntaxTree,\n    clampToU32(duplication?.minTokens),\n    clampToU32(duplication?.maxGapTokens),\n    clampToU32(duplication?.minSimilarityPercent),\n    includeCrossFileData,\n  ];\n}\n\n/** Collects one file's cross-file clone-detection contribution via the native addon. */\nexport function collectCrossFileDataNative(\n  code: string,\n  language: string,\n  minTokens?: number\n): NativeCrossFileDataPayload {\n  return JSON.parse(\n    loadBinding().collectCrossFileDataNative(toWellFormed(code), language, clampToU32(minTokens))\n  ) as NativeCrossFileDataPayload;\n}\n\n/** Collects normalized token hash sequences of every function via the native addon. */\nexport function collectFunctionTokenSequencesNative(code: string, language: string): Int32Array[] {\n  const sequences = JSON.parse(\n    loadBinding().collectFunctionTokenSequencesNative(toWellFormed(code), language)\n  ) as number[][];\n  return sequences.map((sequence) => Int32Array.from(sequence));\n}\n\n/**\n * Lone surrogates cannot cross the N-API boundary losslessly, so ill-formed strings (invalid\n * UTF-16 occasionally present in real-world files) are measured with U+FFFD replacements — the\n * same code units V8's own UTF-8 conversion would substitute.\n */\nfunction toWellFormed(code: string): string {\n  return code.isWellFormed() ? code : code.toWellFormed();\n}\n\n/**\n * The duplication settings cross the boundary as u32, whose JavaScript conversion wraps modulo\n * 2^32 (2 ** 32 would become 0 and match everything). The public API accepts any safe integer, so\n * out-of-range values clamp to [0, u32::MAX] — no source can hold 2^32 tokens, so a clamped\n * threshold behaves identically to the requested one. NaN (e.g. `Number(unsetEnvVariable)`)\n * survives clamping arithmetic and would also convert to 0, so it is treated as an absent setting\n * instead.\n */\nfunction clampToU32(value: number | undefined): number | undefined {\n  return value === undefined || Number.isNaN(value) ? undefined : clamp(Math.trunc(value), 0, 0xFF_FF_FF_FF);\n}\n\n/**\n * Raised when no usable native addon can be loaded. Every measurement fails identically until the\n * addon is built, so callers measuring many files (the CLI scan) treat it as fatal for the whole\n * run instead of recording one \"skipped\" entry per file.\n */\nexport class NativeAddonError extends Error {}\n\nlet cachedBinding: NativeBinding | undefined;\nlet cachedFailure: NativeAddonError | undefined;\n\n/** Replaces the N-API addon, e.g. with the WebAssembly build on runtimes without N-API. */\nexport function setNativeBinding(binding: NativeBinding): void {\n  cachedBinding = binding;\n}\n\nfunction loadBinding(): NativeBinding {\n  if (cachedBinding) {\n    return cachedBinding;\n  }\n  // The failure is memoized too: resolution (including platformTriplet's diagnostic-report call\n  // on Linux) would otherwise repeat for every measured file of an already-failing run.\n  if (cachedFailure) {\n    throw cachedFailure;\n  }\n  // Resolved relative to this file, so both src/ (tests) and dist/ (build) find the addon.\n  // node:module is loaded lazily so that runtimes without it can bundle this module.\n  const requireNative = process.getBuiltinModule('node:module').createRequire(import.meta.url);\n  const specifiers = [\n    // A prebuilt platform package, when published for this platform.\n    `code-gauge-${platformTriplet()}`,\n    // A locally built addon (`bun run build-native`).\n    '../native/code-gauge.node',\n  ];\n  const failures: string[] = [];\n  for (const specifier of specifiers) {\n    let binding: NativeBinding;\n    try {\n      binding = requireNative(specifier) as NativeBinding;\n    } catch (error) {\n      failures.push(`  ${specifier}: ${getErrorMessage(error).split('\\n')[0]}`);\n      continue;\n    }\n    const version = binding.payloadVersion?.();\n    if (version !== expectedPayloadVersion) {\n      failures.push(\n        `  ${specifier}: payload version ${version ?? 'unknown'} does not match the expected ` +\n          `${expectedPayloadVersion}; rebuild the addon with \\`bun run build-native\\``\n      );\n      continue;\n    }\n    cachedBinding = binding;\n    return binding;\n  }\n  cachedFailure = new NativeAddonError(\n    `The code-gauge native addon is not available for ${platformTriplet()}. Build it with ` +\n      '`node scripts/buildNative.mjs` in the code-gauge package directory (requires a Rust ' +\n      'toolchain); when installing with npm, also allow install scripts for code-gauge so its ' +\n      `postinstall build can run.\\n${failures.join('\\n')}`\n  );\n  throw cachedFailure;\n}\n\n/**\n * The platform-package suffix in the napi-rs naming convention: Linux targets are qualified by\n * libc ABI (`linux-x64-gnu` / `linux-x64-musl`) because a glibc-linked addon cannot load on\n * Alpine/musl, and Windows by toolchain ABI (`win32-x64-msvc`), matching what napi-rs tooling\n * generates. Must match scripts/installNative.mjs and the build-native workflow's target list.\n */\nfunction platformTriplet(): string {\n  const base = `${process.platform}-${process.arch}`;\n  if (process.platform === 'win32') {\n    return `${base}-msvc`;\n  }\n  if (process.platform !== 'linux') {\n    return base;\n  }\n  // Musl builds of Node report no glibc runtime version; see napi-rs's isMusl detection.\n  const report = process.report?.getReport() as { header?: { glibcVersionRuntime?: string } } | undefined;\n  return report?.header?.glibcVersionRuntime ? `${base}-gnu` : `${base}-musl`;\n}\n"],"mappings":"sDAmEA,SAAgB,EACd,EACA,EACA,EACA,EACA,EAAuB,GACD,CACtB,OAAO,KAAK,MACV,EAAY,CAAC,CAAC,kBACZ,GAAG,EAAyB,EAAM,EAAU,EAAmB,EAAa,CAAoB,CAClG,CACF,CACF,CAMA,SAAgB,EACd,EACA,EACA,EACA,EACA,EAAuB,GACQ,CAC/B,OAAO,EAAY,CAAC,CACjB,uBACC,GAAG,EAAyB,EAAM,EAAU,EAAmB,EAAa,CAAoB,CAClG,CAAC,CACA,KAAM,GAAS,KAAK,MAAM,CAAI,CAAyB,CAC5D,CAGA,SAAS,EACP,EACA,EACA,EACA,EACA,EACwB,CACxB,MAAO,CACL,EAAa,CAAI,EACjB,EACA,EACA,EAAW,GAAa,SAAS,EACjC,EAAW,GAAa,YAAY,EACpC,EAAW,GAAa,oBAAoB,EAC5C,CACF,CACF,CAGA,SAAgB,EACd,EACA,EACA,EAC4B,CAC5B,OAAO,KAAK,MACV,EAAY,CAAC,CAAC,2BAA2B,EAAa,CAAI,EAAG,EAAU,EAAW,CAAS,CAAC,CAC9F,CACF,CAGA,SAAgB,EAAoC,EAAc,EAAgC,CAIhG,OAHkB,KAAK,MACrB,EAAY,CAAC,CAAC,oCAAoC,EAAa,CAAI,EAAG,CAAQ,CAEjE,CAAC,CAAC,IAAK,GAAa,WAAW,KAAK,CAAQ,CAAC,CAC9D,CAOA,SAAS,EAAa,EAAsB,CAC1C,OAAO,EAAK,aAAa,EAAI,EAAO,EAAK,aAAa,CACxD,CAUA,SAAS,EAAW,EAA+C,CACjE,OAAO,IAAU,IAAA,IAAa,OAAO,MAAM,CAAK,EAAI,IAAA,IAAA,EAAYA,EAAAA,MAAAA,CAAM,KAAK,MAAM,CAAK,EAAG,EAAG,UAAa,CAC3G,CAOA,IAAa,EAAb,cAAsC,KAAM,CAAC,EAE7C,IAAI,EACA,EAGJ,SAAgB,EAAiB,EAA8B,CAC7D,EAAgB,CAClB,CAEA,SAAS,GAA6B,CACpC,GAAI,EACF,OAAO,EAIT,GAAI,EACF,MAAM,EAIR,IAAM,EAAgB,QAAQ,iBAAiB,aAAa,CAAC,CAAC,cAAA,QAAA,KAAA,CAAA,CAAA,cAAA,UAAA,CAAA,CAAA,IAA6B,EACrF,EAAa,CAEjB,cAAc,EAAgB,IAE9B,2BACF,EACM,EAAqB,CAAC,EAC5B,IAAK,IAAM,KAAa,EAAY,CAClC,IAAI,EACJ,GAAI,CACF,EAAU,EAAc,CAAS,CACnC,OAAS,EAAO,CACd,EAAS,KAAK,KAAK,EAAU,KAAA,EAAIC,EAAAA,gBAAAA,CAAgB,CAAK,CAAC,CAAC,MAAM;CAAI,CAAC,CAAC,IAAI,EACxE,QACF,CACA,IAAM,EAAU,EAAQ,iBAAiB,EACzC,GAAI,IAAA,EAAoC,CACtC,EAAS,KACP,KAAK,EAAU,oBAAoB,GAAW,UAAU,gFAE1D,EACA,QACF,CAEA,MADA,GAAgB,EACT,CACT,CAOA,KANA,GAAgB,IAAI,EAClB,oDAAoD,EAAgB,EAAE,2NAGrC,EAAS,KAAK;CAAI,GACrD,EACM,CACR,CAQA,SAAS,GAA0B,CACjC,IAAM,EAAO,GAAG,QAAQ,SAAS,GAAG,QAAQ,OAS5C,OARI,QAAQ,WAAa,QAChB,GAAG,EAAK,OAEb,QAAQ,WAAa,SAIV,QAAQ,QAAQ,UAAU,EAAA,EAC1B,QAAQ,oBAAsB,GAAG,EAAK,MAAQ,GAAG,EAAK,OAJ5D,CAKX"}
+{"version":3,"file":"nativeMetrics.cjs","names":["clamp","getErrorMessage"],"sources":["../src/nativeMetrics.ts"],"sourcesContent":["import { clamp, getErrorMessage } from '@willbooster/shared-lib';\nimport type { CrossFileDuplicateCandidate, Token, TokenRange } from './duplication.js';\nimport type { CodeMetrics, DuplicationOptions, FunctionMetrics } from './types.js';\n\n/**\n * Halstead counts measured natively; the derived float metrics (volume, effort, ...) are\n * computed in TypeScript because they involve transcendental functions (log2) whose last-bit\n * results can differ between V8 and Rust's libm, and results must not depend on the Rust side's\n * libm build.\n */\nexport interface NativeHalsteadCounts {\n  distinctOperators: number;\n  distinctOperands: number;\n  totalOperators: number;\n  totalOperands: number;\n}\n\nexport interface NativeFunctionMetricsPayload extends Omit<FunctionMetrics, 'halstead'> {\n  halsteadCounts: NativeHalsteadCounts;\n}\n\nexport interface NativeMetricsPayload extends Omit<CodeMetrics, 'halstead' | 'functions' | 'syntaxTree'> {\n  functions: NativeFunctionMetricsPayload[];\n  halsteadCounts: NativeHalsteadCounts;\n  syntaxTree?: string;\n  crossFileData?: NativeCrossFileDataPayload;\n}\n\n/** One file's cross-file clone-detection contribution as serialized by the native addon. */\nexport interface NativeCrossFileDataPayload {\n  candidates: CrossFileDuplicateCandidate[];\n  tokens: Token[];\n  containerStatements: TokenRange[][];\n  nearMissBlocks: TokenRange[];\n  /** 1-based lines that are neither blank nor comment-only, sorted ascending. */\n  codeLineNumbers: number[];\n}\n\ntype NativeMeasureArguments = [\n  code: string,\n  language: string,\n  includeSyntaxTree: boolean,\n  minTokens: number | undefined,\n  maxGapTokens: number | undefined,\n  minSimilarityPercent: number | undefined,\n  includeCrossFileData: boolean,\n];\n\nexport interface NativeBinding {\n  measureCodeNative(...args: NativeMeasureArguments): string;\n  measureCodeNativeAsync(...args: NativeMeasureArguments): Promise<string>;\n  collectCrossFileDataNative(code: string, language: string, minTokens?: number): string;\n  collectFunctionTokenSequencesNative(code: string, language: string): string;\n  payloadVersion?(): number;\n}\n\n/**\n * Must equal `payload_version` in native/src/lib.rs. A previously built addon survives a\n * `git pull` untouched, so without this handshake it would silently return payloads missing\n * newer fields, or lack newer binding functions, instead of failing with a clear rebuild message.\n */\nexport const expectedPayloadVersion = 8;\n\n/**\n * Measures one file via the native addon, returning the raw payload for assembly in metrics.ts;\n * with `includeCrossFileData`, the payload also carries the file's cross-file contribution.\n */\nexport function measureCodeNative(\n  code: string,\n  language: string,\n  includeSyntaxTree: boolean,\n  duplication?: DuplicationOptions,\n  includeCrossFileData = false\n): NativeMetricsPayload {\n  return JSON.parse(\n    loadBinding().measureCodeNative(\n      ...toNativeMeasureArguments(code, language, includeSyntaxTree, duplication, includeCrossFileData)\n    )\n  ) as NativeMetricsPayload;\n}\n\n/**\n * measureCodeNative on the addon's worker threads, so several files can be measured in parallel.\n * A missing addon still throws synchronously, like measureCodeNative.\n */\nexport function measureCodeNativeAsync(\n  code: string,\n  language: string,\n  includeSyntaxTree: boolean,\n  duplication?: DuplicationOptions,\n  includeCrossFileData = false\n): Promise<NativeMetricsPayload> {\n  return loadBinding()\n    .measureCodeNativeAsync(\n      ...toNativeMeasureArguments(code, language, includeSyntaxTree, duplication, includeCrossFileData)\n    )\n    .then((json) => JSON.parse(json) as NativeMetricsPayload);\n}\n\n/** The binding arguments shared by the sync and async measurements, so both measure alike. */\nfunction toNativeMeasureArguments(\n  code: string,\n  language: string,\n  includeSyntaxTree: boolean,\n  duplication: DuplicationOptions | undefined,\n  includeCrossFileData: boolean\n): NativeMeasureArguments {\n  return [\n    toWellFormed(code),\n    language,\n    includeSyntaxTree,\n    clampToU32(duplication?.minTokens),\n    clampToU32(duplication?.maxGapTokens),\n    clampToU32(duplication?.minSimilarityPercent),\n    includeCrossFileData,\n  ];\n}\n\n/** Collects one file's cross-file clone-detection contribution via the native addon. */\nexport function collectCrossFileDataNative(\n  code: string,\n  language: string,\n  minTokens?: number\n): NativeCrossFileDataPayload {\n  return JSON.parse(\n    loadBinding().collectCrossFileDataNative(toWellFormed(code), language, clampToU32(minTokens))\n  ) as NativeCrossFileDataPayload;\n}\n\n/** Collects normalized token hash sequences of every function via the native addon. */\nexport function collectFunctionTokenSequencesNative(code: string, language: string): Int32Array[] {\n  const sequences = JSON.parse(\n    loadBinding().collectFunctionTokenSequencesNative(toWellFormed(code), language)\n  ) as number[][];\n  return sequences.map((sequence) => Int32Array.from(sequence));\n}\n\n/**\n * Lone surrogates cannot cross the N-API boundary losslessly, so ill-formed strings (invalid\n * UTF-16 occasionally present in real-world files) are measured with U+FFFD replacements — the\n * same code units V8's own UTF-8 conversion would substitute.\n */\nfunction toWellFormed(code: string): string {\n  return code.isWellFormed() ? code : code.toWellFormed();\n}\n\n/**\n * The duplication settings cross the boundary as u32, whose JavaScript conversion wraps modulo\n * 2^32 (2 ** 32 would become 0 and match everything). The public API accepts any safe integer, so\n * out-of-range values clamp to [0, u32::MAX] — no source can hold 2^32 tokens, so a clamped\n * threshold behaves identically to the requested one. NaN (e.g. `Number(unsetEnvVariable)`)\n * survives clamping arithmetic and would also convert to 0, so it is treated as an absent setting\n * instead.\n */\nfunction clampToU32(value: number | undefined): number | undefined {\n  return value === undefined || Number.isNaN(value) ? undefined : clamp(Math.trunc(value), 0, 0xFF_FF_FF_FF);\n}\n\n/**\n * Raised when no usable native addon can be loaded. Every measurement fails identically until the\n * addon is built, so callers measuring many files (the CLI scan) treat it as fatal for the whole\n * run instead of recording one \"skipped\" entry per file.\n */\nexport class NativeAddonError extends Error {}\n\nlet cachedBinding: NativeBinding | undefined;\nlet cachedFailure: NativeAddonError | undefined;\n\n/** Replaces the N-API addon, e.g. with the WebAssembly build on runtimes without N-API. */\nexport function setNativeBinding(binding: NativeBinding): void {\n  cachedBinding = binding;\n}\n\nfunction loadBinding(): NativeBinding {\n  if (cachedBinding) {\n    return cachedBinding;\n  }\n  // The failure is memoized too: resolution (including platformTriplet's diagnostic-report call\n  // on Linux) would otherwise repeat for every measured file of an already-failing run.\n  if (cachedFailure) {\n    throw cachedFailure;\n  }\n  // Resolved relative to this file, so both src/ (tests) and dist/ (build) find the addon.\n  // node:module is loaded lazily so that runtimes without it can bundle this module.\n  const requireNative = process.getBuiltinModule('node:module').createRequire(import.meta.url);\n  const specifiers = [\n    // A prebuilt platform package, when published for this platform.\n    `code-gauge-${platformTriplet()}`,\n    // A locally built addon (`bun run build-native`).\n    '../native/code-gauge.node',\n  ];\n  const failures: string[] = [];\n  for (const specifier of specifiers) {\n    let binding: NativeBinding;\n    try {\n      binding = requireNative(specifier) as NativeBinding;\n    } catch (error) {\n      failures.push(`  ${specifier}: ${getErrorMessage(error).split('\\n')[0]}`);\n      continue;\n    }\n    const version = binding.payloadVersion?.();\n    if (version !== expectedPayloadVersion) {\n      failures.push(\n        `  ${specifier}: payload version ${version ?? 'unknown'} does not match the expected ` +\n          `${expectedPayloadVersion}; rebuild the addon with \\`bun run build-native\\``\n      );\n      continue;\n    }\n    cachedBinding = binding;\n    return binding;\n  }\n  cachedFailure = new NativeAddonError(\n    `The code-gauge native addon is not available for ${platformTriplet()}. Build it with ` +\n      '`node scripts/buildNative.mjs` in the code-gauge package directory (requires a Rust ' +\n      `toolchain).\\n${failures.join('\\n')}`\n  );\n  throw cachedFailure;\n}\n\n/**\n * The platform-package suffix in the napi-rs naming convention: Linux targets are qualified by\n * libc ABI (`linux-x64-gnu` / `linux-x64-musl`) because a glibc-linked addon cannot load on\n * Alpine/musl, and Windows by toolchain ABI (`win32-x64-msvc`), matching what napi-rs tooling\n * generates. Must match scripts/installNative.mjs and the build-native workflow's target list.\n */\nfunction platformTriplet(): string {\n  const base = `${process.platform}-${process.arch}`;\n  if (process.platform === 'win32') {\n    return `${base}-msvc`;\n  }\n  if (process.platform !== 'linux') {\n    return base;\n  }\n  // Musl builds of Node report no glibc runtime version; see napi-rs's isMusl detection.\n  const report = process.report?.getReport() as { header?: { glibcVersionRuntime?: string } } | undefined;\n  return report?.header?.glibcVersionRuntime ? `${base}-gnu` : `${base}-musl`;\n}\n"],"mappings":"sDAmEA,SAAgB,EACd,EACA,EACA,EACA,EACA,EAAuB,GACD,CACtB,OAAO,KAAK,MACV,EAAY,CAAC,CAAC,kBACZ,GAAG,EAAyB,EAAM,EAAU,EAAmB,EAAa,CAAoB,CAClG,CACF,CACF,CAMA,SAAgB,EACd,EACA,EACA,EACA,EACA,EAAuB,GACQ,CAC/B,OAAO,EAAY,CAAC,CACjB,uBACC,GAAG,EAAyB,EAAM,EAAU,EAAmB,EAAa,CAAoB,CAClG,CAAC,CACA,KAAM,GAAS,KAAK,MAAM,CAAI,CAAyB,CAC5D,CAGA,SAAS,EACP,EACA,EACA,EACA,EACA,EACwB,CACxB,MAAO,CACL,EAAa,CAAI,EACjB,EACA,EACA,EAAW,GAAa,SAAS,EACjC,EAAW,GAAa,YAAY,EACpC,EAAW,GAAa,oBAAoB,EAC5C,CACF,CACF,CAGA,SAAgB,EACd,EACA,EACA,EAC4B,CAC5B,OAAO,KAAK,MACV,EAAY,CAAC,CAAC,2BAA2B,EAAa,CAAI,EAAG,EAAU,EAAW,CAAS,CAAC,CAC9F,CACF,CAGA,SAAgB,EAAoC,EAAc,EAAgC,CAIhG,OAHkB,KAAK,MACrB,EAAY,CAAC,CAAC,oCAAoC,EAAa,CAAI,EAAG,CAAQ,CAEjE,CAAC,CAAC,IAAK,GAAa,WAAW,KAAK,CAAQ,CAAC,CAC9D,CAOA,SAAS,EAAa,EAAsB,CAC1C,OAAO,EAAK,aAAa,EAAI,EAAO,EAAK,aAAa,CACxD,CAUA,SAAS,EAAW,EAA+C,CACjE,OAAO,IAAU,IAAA,IAAa,OAAO,MAAM,CAAK,EAAI,IAAA,IAAA,EAAYA,EAAAA,MAAAA,CAAM,KAAK,MAAM,CAAK,EAAG,EAAG,UAAa,CAC3G,CAOA,IAAa,EAAb,cAAsC,KAAM,CAAC,EAE7C,IAAI,EACA,EAGJ,SAAgB,EAAiB,EAA8B,CAC7D,EAAgB,CAClB,CAEA,SAAS,GAA6B,CACpC,GAAI,EACF,OAAO,EAIT,GAAI,EACF,MAAM,EAIR,IAAM,EAAgB,QAAQ,iBAAiB,aAAa,CAAC,CAAC,cAAA,QAAA,KAAA,CAAA,CAAA,cAAA,UAAA,CAAA,CAAA,IAA6B,EACrF,EAAa,CAEjB,cAAc,EAAgB,IAE9B,2BACF,EACM,EAAqB,CAAC,EAC5B,IAAK,IAAM,KAAa,EAAY,CAClC,IAAI,EACJ,GAAI,CACF,EAAU,EAAc,CAAS,CACnC,OAAS,EAAO,CACd,EAAS,KAAK,KAAK,EAAU,KAAA,EAAIC,EAAAA,gBAAAA,CAAgB,CAAK,CAAC,CAAC,MAAM;CAAI,CAAC,CAAC,IAAI,EACxE,QACF,CACA,IAAM,EAAU,EAAQ,iBAAiB,EACzC,GAAI,IAAA,EAAoC,CACtC,EAAS,KACP,KAAK,EAAU,oBAAoB,GAAW,UAAU,gFAE1D,EACA,QACF,CAEA,MADA,GAAgB,EACT,CACT,CAMA,KALA,GAAgB,IAAI,EAClB,oDAAoD,EAAgB,EAAE,qHAEpD,EAAS,KAAK;CAAI,GACtC,EACM,CACR,CAQA,SAAS,GAA0B,CACjC,IAAM,EAAO,GAAG,QAAQ,SAAS,GAAG,QAAQ,OAS5C,OARI,QAAQ,WAAa,QAChB,GAAG,EAAK,OAEb,QAAQ,WAAa,SAIV,QAAQ,QAAQ,UAAU,EAAA,EAC1B,QAAQ,oBAAsB,GAAG,EAAK,MAAQ,GAAG,EAAK,OAJ5D,CAKX"}

package/dist/nativeMetrics.js

@@ -1,4 +1,4 @@
 import{clamp as e,getErrorMessage as t}from"@willbooster/shared-lib";function n(e,t,n,r,a=!1){return JSON.parse(p().measureCodeNative(...i(e,t,n,r,a)))}function r(e,t,n,r,a=!1){return p().measureCodeNativeAsync(...i(e,t,n,r,a)).then(e=>JSON.parse(e))}function i(e,t,n,r,i){return[s(e),t,n,c(r?.minTokens),c(r?.maxGapTokens),c(r?.minSimilarityPercent),i]}function a(e,t,n){return JSON.parse(p().collectCrossFileDataNative(s(e),t,c(n)))}function o(e,t){return JSON.parse(p().collectFunctionTokenSequencesNative(s(e),t)).map(e=>Int32Array.from(e))}function s(e){return e.isWellFormed()?e:e.toWellFormed()}function c(t){return t===void 0||Number.isNaN(t)?void 0:e(Math.trunc(t),0,4294967295)}var l=class extends Error{};let u,d;function f(e){u=e}function p(){if(u)return u;if(d)throw d;let e=process.getBuiltinModule(`node:module`).createRequire(import.meta.url),n=[`code-gauge-${m()}`,`../native/code-gauge.node`],r=[];for(let i of n){let n;try{n=e(i)}catch(e){r.push(`  ${i}: ${t(e).split(`
-`)[0]}`);continue}let a=n.payloadVersion?.();if(a!==8){r.push(`  ${i}: payload version ${a??`unknown`} does not match the expected 8; rebuild the addon with \`bun run build-native\``);continue}return u=n,n}throw d=new l(`The code-gauge native addon is not available for ${m()}. Build it with \`node scripts/buildNative.mjs\` in the code-gauge package directory (requires a Rust toolchain); when installing with npm, also allow install scripts for code-gauge so its postinstall build can run.\n${r.join(`
+`)[0]}`);continue}let a=n.payloadVersion?.();if(a!==8){r.push(`  ${i}: payload version ${a??`unknown`} does not match the expected 8; rebuild the addon with \`bun run build-native\``);continue}return u=n,n}throw d=new l(`The code-gauge native addon is not available for ${m()}. Build it with \`node scripts/buildNative.mjs\` in the code-gauge package directory (requires a Rust toolchain).\n${r.join(`
 `)}`),d}function m(){let e=`${process.platform}-${process.arch}`;return process.platform===`win32`?`${e}-msvc`:process.platform===`linux`?(process.report?.getReport())?.header?.glibcVersionRuntime?`${e}-gnu`:`${e}-musl`:e}export{l as NativeAddonError,a as collectCrossFileDataNative,o as collectFunctionTokenSequencesNative,n as measureCodeNative,r as measureCodeNativeAsync,f as setNativeBinding};
 //# sourceMappingURL=nativeMetrics.js.map

package/dist/nativeMetrics.js.map

@@ -1 +1 @@
-{"version":3,"file":"nativeMetrics.js","names":[],"sources":["../src/nativeMetrics.ts"],"sourcesContent":["import { clamp, getErrorMessage } from '@willbooster/shared-lib';\nimport type { CrossFileDuplicateCandidate, Token, TokenRange } from './duplication.js';\nimport type { CodeMetrics, DuplicationOptions, FunctionMetrics } from './types.js';\n\n/**\n * Halstead counts measured natively; the derived float metrics (volume, effort, ...) are\n * computed in TypeScript because they involve transcendental functions (log2) whose last-bit\n * results can differ between V8 and Rust's libm, and results must not depend on the Rust side's\n * libm build.\n */\nexport interface NativeHalsteadCounts {\n  distinctOperators: number;\n  distinctOperands: number;\n  totalOperators: number;\n  totalOperands: number;\n}\n\nexport interface NativeFunctionMetricsPayload extends Omit<FunctionMetrics, 'halstead'> {\n  halsteadCounts: NativeHalsteadCounts;\n}\n\nexport interface NativeMetricsPayload extends Omit<CodeMetrics, 'halstead' | 'functions' | 'syntaxTree'> {\n  functions: NativeFunctionMetricsPayload[];\n  halsteadCounts: NativeHalsteadCounts;\n  syntaxTree?: string;\n  crossFileData?: NativeCrossFileDataPayload;\n}\n\n/** One file's cross-file clone-detection contribution as serialized by the native addon. */\nexport interface NativeCrossFileDataPayload {\n  candidates: CrossFileDuplicateCandidate[];\n  tokens: Token[];\n  containerStatements: TokenRange[][];\n  nearMissBlocks: TokenRange[];\n  /** 1-based lines that are neither blank nor comment-only, sorted ascending. */\n  codeLineNumbers: number[];\n}\n\ntype NativeMeasureArguments = [\n  code: string,\n  language: string,\n  includeSyntaxTree: boolean,\n  minTokens: number | undefined,\n  maxGapTokens: number | undefined,\n  minSimilarityPercent: number | undefined,\n  includeCrossFileData: boolean,\n];\n\nexport interface NativeBinding {\n  measureCodeNative(...args: NativeMeasureArguments): string;\n  measureCodeNativeAsync(...args: NativeMeasureArguments): Promise<string>;\n  collectCrossFileDataNative(code: string, language: string, minTokens?: number): string;\n  collectFunctionTokenSequencesNative(code: string, language: string): string;\n  payloadVersion?(): number;\n}\n\n/**\n * Must equal `payload_version` in native/src/lib.rs. A previously built addon survives a\n * `git pull` untouched, so without this handshake it would silently return payloads missing\n * newer fields, or lack newer binding functions, instead of failing with a clear rebuild message.\n */\nexport const expectedPayloadVersion = 8;\n\n/**\n * Measures one file via the native addon, returning the raw payload for assembly in metrics.ts;\n * with `includeCrossFileData`, the payload also carries the file's cross-file contribution.\n */\nexport function measureCodeNative(\n  code: string,\n  language: string,\n  includeSyntaxTree: boolean,\n  duplication?: DuplicationOptions,\n  includeCrossFileData = false\n): NativeMetricsPayload {\n  return JSON.parse(\n    loadBinding().measureCodeNative(\n      ...toNativeMeasureArguments(code, language, includeSyntaxTree, duplication, includeCrossFileData)\n    )\n  ) as NativeMetricsPayload;\n}\n\n/**\n * measureCodeNative on the addon's worker threads, so several files can be measured in parallel.\n * A missing addon still throws synchronously, like measureCodeNative.\n */\nexport function measureCodeNativeAsync(\n  code: string,\n  language: string,\n  includeSyntaxTree: boolean,\n  duplication?: DuplicationOptions,\n  includeCrossFileData = false\n): Promise<NativeMetricsPayload> {\n  return loadBinding()\n    .measureCodeNativeAsync(\n      ...toNativeMeasureArguments(code, language, includeSyntaxTree, duplication, includeCrossFileData)\n    )\n    .then((json) => JSON.parse(json) as NativeMetricsPayload);\n}\n\n/** The binding arguments shared by the sync and async measurements, so both measure alike. */\nfunction toNativeMeasureArguments(\n  code: string,\n  language: string,\n  includeSyntaxTree: boolean,\n  duplication: DuplicationOptions | undefined,\n  includeCrossFileData: boolean\n): NativeMeasureArguments {\n  return [\n    toWellFormed(code),\n    language,\n    includeSyntaxTree,\n    clampToU32(duplication?.minTokens),\n    clampToU32(duplication?.maxGapTokens),\n    clampToU32(duplication?.minSimilarityPercent),\n    includeCrossFileData,\n  ];\n}\n\n/** Collects one file's cross-file clone-detection contribution via the native addon. */\nexport function collectCrossFileDataNative(\n  code: string,\n  language: string,\n  minTokens?: number\n): NativeCrossFileDataPayload {\n  return JSON.parse(\n    loadBinding().collectCrossFileDataNative(toWellFormed(code), language, clampToU32(minTokens))\n  ) as NativeCrossFileDataPayload;\n}\n\n/** Collects normalized token hash sequences of every function via the native addon. */\nexport function collectFunctionTokenSequencesNative(code: string, language: string): Int32Array[] {\n  const sequences = JSON.parse(\n    loadBinding().collectFunctionTokenSequencesNative(toWellFormed(code), language)\n  ) as number[][];\n  return sequences.map((sequence) => Int32Array.from(sequence));\n}\n\n/**\n * Lone surrogates cannot cross the N-API boundary losslessly, so ill-formed strings (invalid\n * UTF-16 occasionally present in real-world files) are measured with U+FFFD replacements — the\n * same code units V8's own UTF-8 conversion would substitute.\n */\nfunction toWellFormed(code: string): string {\n  return code.isWellFormed() ? code : code.toWellFormed();\n}\n\n/**\n * The duplication settings cross the boundary as u32, whose JavaScript conversion wraps modulo\n * 2^32 (2 ** 32 would become 0 and match everything). The public API accepts any safe integer, so\n * out-of-range values clamp to [0, u32::MAX] — no source can hold 2^32 tokens, so a clamped\n * threshold behaves identically to the requested one. NaN (e.g. `Number(unsetEnvVariable)`)\n * survives clamping arithmetic and would also convert to 0, so it is treated as an absent setting\n * instead.\n */\nfunction clampToU32(value: number | undefined): number | undefined {\n  return value === undefined || Number.isNaN(value) ? undefined : clamp(Math.trunc(value), 0, 0xFF_FF_FF_FF);\n}\n\n/**\n * Raised when no usable native addon can be loaded. Every measurement fails identically until the\n * addon is built, so callers measuring many files (the CLI scan) treat it as fatal for the whole\n * run instead of recording one \"skipped\" entry per file.\n */\nexport class NativeAddonError extends Error {}\n\nlet cachedBinding: NativeBinding | undefined;\nlet cachedFailure: NativeAddonError | undefined;\n\n/** Replaces the N-API addon, e.g. with the WebAssembly build on runtimes without N-API. */\nexport function setNativeBinding(binding: NativeBinding): void {\n  cachedBinding = binding;\n}\n\nfunction loadBinding(): NativeBinding {\n  if (cachedBinding) {\n    return cachedBinding;\n  }\n  // The failure is memoized too: resolution (including platformTriplet's diagnostic-report call\n  // on Linux) would otherwise repeat for every measured file of an already-failing run.\n  if (cachedFailure) {\n    throw cachedFailure;\n  }\n  // Resolved relative to this file, so both src/ (tests) and dist/ (build) find the addon.\n  // node:module is loaded lazily so that runtimes without it can bundle this module.\n  const requireNative = process.getBuiltinModule('node:module').createRequire(import.meta.url);\n  const specifiers = [\n    // A prebuilt platform package, when published for this platform.\n    `code-gauge-${platformTriplet()}`,\n    // A locally built addon (`bun run build-native`).\n    '../native/code-gauge.node',\n  ];\n  const failures: string[] = [];\n  for (const specifier of specifiers) {\n    let binding: NativeBinding;\n    try {\n      binding = requireNative(specifier) as NativeBinding;\n    } catch (error) {\n      failures.push(`  ${specifier}: ${getErrorMessage(error).split('\\n')[0]}`);\n      continue;\n    }\n    const version = binding.payloadVersion?.();\n    if (version !== expectedPayloadVersion) {\n      failures.push(\n        `  ${specifier}: payload version ${version ?? 'unknown'} does not match the expected ` +\n          `${expectedPayloadVersion}; rebuild the addon with \\`bun run build-native\\``\n      );\n      continue;\n    }\n    cachedBinding = binding;\n    return binding;\n  }\n  cachedFailure = new NativeAddonError(\n    `The code-gauge native addon is not available for ${platformTriplet()}. Build it with ` +\n      '`node scripts/buildNative.mjs` in the code-gauge package directory (requires a Rust ' +\n      'toolchain); when installing with npm, also allow install scripts for code-gauge so its ' +\n      `postinstall build can run.\\n${failures.join('\\n')}`\n  );\n  throw cachedFailure;\n}\n\n/**\n * The platform-package suffix in the napi-rs naming convention: Linux targets are qualified by\n * libc ABI (`linux-x64-gnu` / `linux-x64-musl`) because a glibc-linked addon cannot load on\n * Alpine/musl, and Windows by toolchain ABI (`win32-x64-msvc`), matching what napi-rs tooling\n * generates. Must match scripts/installNative.mjs and the build-native workflow's target list.\n */\nfunction platformTriplet(): string {\n  const base = `${process.platform}-${process.arch}`;\n  if (process.platform === 'win32') {\n    return `${base}-msvc`;\n  }\n  if (process.platform !== 'linux') {\n    return base;\n  }\n  // Musl builds of Node report no glibc runtime version; see napi-rs's isMusl detection.\n  const report = process.report?.getReport() as { header?: { glibcVersionRuntime?: string } } | undefined;\n  return report?.header?.glibcVersionRuntime ? `${base}-gnu` : `${base}-musl`;\n}\n"],"mappings":"qEAmEA,SAAgB,EACd,EACA,EACA,EACA,EACA,EAAuB,GACD,CACtB,OAAO,KAAK,MACV,EAAY,CAAC,CAAC,kBACZ,GAAG,EAAyB,EAAM,EAAU,EAAmB,EAAa,CAAoB,CAClG,CACF,CACF,CAMA,SAAgB,EACd,EACA,EACA,EACA,EACA,EAAuB,GACQ,CAC/B,OAAO,EAAY,CAAC,CACjB,uBACC,GAAG,EAAyB,EAAM,EAAU,EAAmB,EAAa,CAAoB,CAClG,CAAC,CACA,KAAM,GAAS,KAAK,MAAM,CAAI,CAAyB,CAC5D,CAGA,SAAS,EACP,EACA,EACA,EACA,EACA,EACwB,CACxB,MAAO,CACL,EAAa,CAAI,EACjB,EACA,EACA,EAAW,GAAa,SAAS,EACjC,EAAW,GAAa,YAAY,EACpC,EAAW,GAAa,oBAAoB,EAC5C,CACF,CACF,CAGA,SAAgB,EACd,EACA,EACA,EAC4B,CAC5B,OAAO,KAAK,MACV,EAAY,CAAC,CAAC,2BAA2B,EAAa,CAAI,EAAG,EAAU,EAAW,CAAS,CAAC,CAC9F,CACF,CAGA,SAAgB,EAAoC,EAAc,EAAgC,CAIhG,OAHkB,KAAK,MACrB,EAAY,CAAC,CAAC,oCAAoC,EAAa,CAAI,EAAG,CAAQ,CAEjE,CAAC,CAAC,IAAK,GAAa,WAAW,KAAK,CAAQ,CAAC,CAC9D,CAOA,SAAS,EAAa,EAAsB,CAC1C,OAAO,EAAK,aAAa,EAAI,EAAO,EAAK,aAAa,CACxD,CAUA,SAAS,EAAW,EAA+C,CACjE,OAAO,IAAU,IAAA,IAAa,OAAO,MAAM,CAAK,EAAI,IAAA,GAAY,EAAM,KAAK,MAAM,CAAK,EAAG,EAAG,UAAa,CAC3G,CAOA,IAAa,EAAb,cAAsC,KAAM,CAAC,EAE7C,IAAI,EACA,EAGJ,SAAgB,EAAiB,EAA8B,CAC7D,EAAgB,CAClB,CAEA,SAAS,GAA6B,CACpC,GAAI,EACF,OAAO,EAIT,GAAI,EACF,MAAM,EAIR,IAAM,EAAgB,QAAQ,iBAAiB,aAAa,CAAC,CAAC,cAAc,YAAY,GAAG,EACrF,EAAa,CAEjB,cAAc,EAAgB,IAE9B,2BACF,EACM,EAAqB,CAAC,EAC5B,IAAK,IAAM,KAAa,EAAY,CAClC,IAAI,EACJ,GAAI,CACF,EAAU,EAAc,CAAS,CACnC,OAAS,EAAO,CACd,EAAS,KAAK,KAAK,EAAU,IAAI,EAAgB,CAAK,CAAC,CAAC,MAAM;CAAI,CAAC,CAAC,IAAI,EACxE,QACF,CACA,IAAM,EAAU,EAAQ,iBAAiB,EACzC,GAAI,IAAA,EAAoC,CACtC,EAAS,KACP,KAAK,EAAU,oBAAoB,GAAW,UAAU,gFAE1D,EACA,QACF,CAEA,MADA,GAAgB,EACT,CACT,CAOA,KANA,GAAgB,IAAI,EAClB,oDAAoD,EAAgB,EAAE,2NAGrC,EAAS,KAAK;CAAI,GACrD,EACM,CACR,CAQA,SAAS,GAA0B,CACjC,IAAM,EAAO,GAAG,QAAQ,SAAS,GAAG,QAAQ,OAS5C,OARI,QAAQ,WAAa,QAChB,GAAG,EAAK,OAEb,QAAQ,WAAa,SAIV,QAAQ,QAAQ,UAAU,EAAA,EAC1B,QAAQ,oBAAsB,GAAG,EAAK,MAAQ,GAAG,EAAK,OAJ5D,CAKX"}
+{"version":3,"file":"nativeMetrics.js","names":[],"sources":["../src/nativeMetrics.ts"],"sourcesContent":["import { clamp, getErrorMessage } from '@willbooster/shared-lib';\nimport type { CrossFileDuplicateCandidate, Token, TokenRange } from './duplication.js';\nimport type { CodeMetrics, DuplicationOptions, FunctionMetrics } from './types.js';\n\n/**\n * Halstead counts measured natively; the derived float metrics (volume, effort, ...) are\n * computed in TypeScript because they involve transcendental functions (log2) whose last-bit\n * results can differ between V8 and Rust's libm, and results must not depend on the Rust side's\n * libm build.\n */\nexport interface NativeHalsteadCounts {\n  distinctOperators: number;\n  distinctOperands: number;\n  totalOperators: number;\n  totalOperands: number;\n}\n\nexport interface NativeFunctionMetricsPayload extends Omit<FunctionMetrics, 'halstead'> {\n  halsteadCounts: NativeHalsteadCounts;\n}\n\nexport interface NativeMetricsPayload extends Omit<CodeMetrics, 'halstead' | 'functions' | 'syntaxTree'> {\n  functions: NativeFunctionMetricsPayload[];\n  halsteadCounts: NativeHalsteadCounts;\n  syntaxTree?: string;\n  crossFileData?: NativeCrossFileDataPayload;\n}\n\n/** One file's cross-file clone-detection contribution as serialized by the native addon. */\nexport interface NativeCrossFileDataPayload {\n  candidates: CrossFileDuplicateCandidate[];\n  tokens: Token[];\n  containerStatements: TokenRange[][];\n  nearMissBlocks: TokenRange[];\n  /** 1-based lines that are neither blank nor comment-only, sorted ascending. */\n  codeLineNumbers: number[];\n}\n\ntype NativeMeasureArguments = [\n  code: string,\n  language: string,\n  includeSyntaxTree: boolean,\n  minTokens: number | undefined,\n  maxGapTokens: number | undefined,\n  minSimilarityPercent: number | undefined,\n  includeCrossFileData: boolean,\n];\n\nexport interface NativeBinding {\n  measureCodeNative(...args: NativeMeasureArguments): string;\n  measureCodeNativeAsync(...args: NativeMeasureArguments): Promise<string>;\n  collectCrossFileDataNative(code: string, language: string, minTokens?: number): string;\n  collectFunctionTokenSequencesNative(code: string, language: string): string;\n  payloadVersion?(): number;\n}\n\n/**\n * Must equal `payload_version` in native/src/lib.rs. A previously built addon survives a\n * `git pull` untouched, so without this handshake it would silently return payloads missing\n * newer fields, or lack newer binding functions, instead of failing with a clear rebuild message.\n */\nexport const expectedPayloadVersion = 8;\n\n/**\n * Measures one file via the native addon, returning the raw payload for assembly in metrics.ts;\n * with `includeCrossFileData`, the payload also carries the file's cross-file contribution.\n */\nexport function measureCodeNative(\n  code: string,\n  language: string,\n  includeSyntaxTree: boolean,\n  duplication?: DuplicationOptions,\n  includeCrossFileData = false\n): NativeMetricsPayload {\n  return JSON.parse(\n    loadBinding().measureCodeNative(\n      ...toNativeMeasureArguments(code, language, includeSyntaxTree, duplication, includeCrossFileData)\n    )\n  ) as NativeMetricsPayload;\n}\n\n/**\n * measureCodeNative on the addon's worker threads, so several files can be measured in parallel.\n * A missing addon still throws synchronously, like measureCodeNative.\n */\nexport function measureCodeNativeAsync(\n  code: string,\n  language: string,\n  includeSyntaxTree: boolean,\n  duplication?: DuplicationOptions,\n  includeCrossFileData = false\n): Promise<NativeMetricsPayload> {\n  return loadBinding()\n    .measureCodeNativeAsync(\n      ...toNativeMeasureArguments(code, language, includeSyntaxTree, duplication, includeCrossFileData)\n    )\n    .then((json) => JSON.parse(json) as NativeMetricsPayload);\n}\n\n/** The binding arguments shared by the sync and async measurements, so both measure alike. */\nfunction toNativeMeasureArguments(\n  code: string,\n  language: string,\n  includeSyntaxTree: boolean,\n  duplication: DuplicationOptions | undefined,\n  includeCrossFileData: boolean\n): NativeMeasureArguments {\n  return [\n    toWellFormed(code),\n    language,\n    includeSyntaxTree,\n    clampToU32(duplication?.minTokens),\n    clampToU32(duplication?.maxGapTokens),\n    clampToU32(duplication?.minSimilarityPercent),\n    includeCrossFileData,\n  ];\n}\n\n/** Collects one file's cross-file clone-detection contribution via the native addon. */\nexport function collectCrossFileDataNative(\n  code: string,\n  language: string,\n  minTokens?: number\n): NativeCrossFileDataPayload {\n  return JSON.parse(\n    loadBinding().collectCrossFileDataNative(toWellFormed(code), language, clampToU32(minTokens))\n  ) as NativeCrossFileDataPayload;\n}\n\n/** Collects normalized token hash sequences of every function via the native addon. */\nexport function collectFunctionTokenSequencesNative(code: string, language: string): Int32Array[] {\n  const sequences = JSON.parse(\n    loadBinding().collectFunctionTokenSequencesNative(toWellFormed(code), language)\n  ) as number[][];\n  return sequences.map((sequence) => Int32Array.from(sequence));\n}\n\n/**\n * Lone surrogates cannot cross the N-API boundary losslessly, so ill-formed strings (invalid\n * UTF-16 occasionally present in real-world files) are measured with U+FFFD replacements — the\n * same code units V8's own UTF-8 conversion would substitute.\n */\nfunction toWellFormed(code: string): string {\n  return code.isWellFormed() ? code : code.toWellFormed();\n}\n\n/**\n * The duplication settings cross the boundary as u32, whose JavaScript conversion wraps modulo\n * 2^32 (2 ** 32 would become 0 and match everything). The public API accepts any safe integer, so\n * out-of-range values clamp to [0, u32::MAX] — no source can hold 2^32 tokens, so a clamped\n * threshold behaves identically to the requested one. NaN (e.g. `Number(unsetEnvVariable)`)\n * survives clamping arithmetic and would also convert to 0, so it is treated as an absent setting\n * instead.\n */\nfunction clampToU32(value: number | undefined): number | undefined {\n  return value === undefined || Number.isNaN(value) ? undefined : clamp(Math.trunc(value), 0, 0xFF_FF_FF_FF);\n}\n\n/**\n * Raised when no usable native addon can be loaded. Every measurement fails identically until the\n * addon is built, so callers measuring many files (the CLI scan) treat it as fatal for the whole\n * run instead of recording one \"skipped\" entry per file.\n */\nexport class NativeAddonError extends Error {}\n\nlet cachedBinding: NativeBinding | undefined;\nlet cachedFailure: NativeAddonError | undefined;\n\n/** Replaces the N-API addon, e.g. with the WebAssembly build on runtimes without N-API. */\nexport function setNativeBinding(binding: NativeBinding): void {\n  cachedBinding = binding;\n}\n\nfunction loadBinding(): NativeBinding {\n  if (cachedBinding) {\n    return cachedBinding;\n  }\n  // The failure is memoized too: resolution (including platformTriplet's diagnostic-report call\n  // on Linux) would otherwise repeat for every measured file of an already-failing run.\n  if (cachedFailure) {\n    throw cachedFailure;\n  }\n  // Resolved relative to this file, so both src/ (tests) and dist/ (build) find the addon.\n  // node:module is loaded lazily so that runtimes without it can bundle this module.\n  const requireNative = process.getBuiltinModule('node:module').createRequire(import.meta.url);\n  const specifiers = [\n    // A prebuilt platform package, when published for this platform.\n    `code-gauge-${platformTriplet()}`,\n    // A locally built addon (`bun run build-native`).\n    '../native/code-gauge.node',\n  ];\n  const failures: string[] = [];\n  for (const specifier of specifiers) {\n    let binding: NativeBinding;\n    try {\n      binding = requireNative(specifier) as NativeBinding;\n    } catch (error) {\n      failures.push(`  ${specifier}: ${getErrorMessage(error).split('\\n')[0]}`);\n      continue;\n    }\n    const version = binding.payloadVersion?.();\n    if (version !== expectedPayloadVersion) {\n      failures.push(\n        `  ${specifier}: payload version ${version ?? 'unknown'} does not match the expected ` +\n          `${expectedPayloadVersion}; rebuild the addon with \\`bun run build-native\\``\n      );\n      continue;\n    }\n    cachedBinding = binding;\n    return binding;\n  }\n  cachedFailure = new NativeAddonError(\n    `The code-gauge native addon is not available for ${platformTriplet()}. Build it with ` +\n      '`node scripts/buildNative.mjs` in the code-gauge package directory (requires a Rust ' +\n      `toolchain).\\n${failures.join('\\n')}`\n  );\n  throw cachedFailure;\n}\n\n/**\n * The platform-package suffix in the napi-rs naming convention: Linux targets are qualified by\n * libc ABI (`linux-x64-gnu` / `linux-x64-musl`) because a glibc-linked addon cannot load on\n * Alpine/musl, and Windows by toolchain ABI (`win32-x64-msvc`), matching what napi-rs tooling\n * generates. Must match scripts/installNative.mjs and the build-native workflow's target list.\n */\nfunction platformTriplet(): string {\n  const base = `${process.platform}-${process.arch}`;\n  if (process.platform === 'win32') {\n    return `${base}-msvc`;\n  }\n  if (process.platform !== 'linux') {\n    return base;\n  }\n  // Musl builds of Node report no glibc runtime version; see napi-rs's isMusl detection.\n  const report = process.report?.getReport() as { header?: { glibcVersionRuntime?: string } } | undefined;\n  return report?.header?.glibcVersionRuntime ? `${base}-gnu` : `${base}-musl`;\n}\n"],"mappings":"qEAmEA,SAAgB,EACd,EACA,EACA,EACA,EACA,EAAuB,GACD,CACtB,OAAO,KAAK,MACV,EAAY,CAAC,CAAC,kBACZ,GAAG,EAAyB,EAAM,EAAU,EAAmB,EAAa,CAAoB,CAClG,CACF,CACF,CAMA,SAAgB,EACd,EACA,EACA,EACA,EACA,EAAuB,GACQ,CAC/B,OAAO,EAAY,CAAC,CACjB,uBACC,GAAG,EAAyB,EAAM,EAAU,EAAmB,EAAa,CAAoB,CAClG,CAAC,CACA,KAAM,GAAS,KAAK,MAAM,CAAI,CAAyB,CAC5D,CAGA,SAAS,EACP,EACA,EACA,EACA,EACA,EACwB,CACxB,MAAO,CACL,EAAa,CAAI,EACjB,EACA,EACA,EAAW,GAAa,SAAS,EACjC,EAAW,GAAa,YAAY,EACpC,EAAW,GAAa,oBAAoB,EAC5C,CACF,CACF,CAGA,SAAgB,EACd,EACA,EACA,EAC4B,CAC5B,OAAO,KAAK,MACV,EAAY,CAAC,CAAC,2BAA2B,EAAa,CAAI,EAAG,EAAU,EAAW,CAAS,CAAC,CAC9F,CACF,CAGA,SAAgB,EAAoC,EAAc,EAAgC,CAIhG,OAHkB,KAAK,MACrB,EAAY,CAAC,CAAC,oCAAoC,EAAa,CAAI,EAAG,CAAQ,CAEjE,CAAC,CAAC,IAAK,GAAa,WAAW,KAAK,CAAQ,CAAC,CAC9D,CAOA,SAAS,EAAa,EAAsB,CAC1C,OAAO,EAAK,aAAa,EAAI,EAAO,EAAK,aAAa,CACxD,CAUA,SAAS,EAAW,EAA+C,CACjE,OAAO,IAAU,IAAA,IAAa,OAAO,MAAM,CAAK,EAAI,IAAA,GAAY,EAAM,KAAK,MAAM,CAAK,EAAG,EAAG,UAAa,CAC3G,CAOA,IAAa,EAAb,cAAsC,KAAM,CAAC,EAE7C,IAAI,EACA,EAGJ,SAAgB,EAAiB,EAA8B,CAC7D,EAAgB,CAClB,CAEA,SAAS,GAA6B,CACpC,GAAI,EACF,OAAO,EAIT,GAAI,EACF,MAAM,EAIR,IAAM,EAAgB,QAAQ,iBAAiB,aAAa,CAAC,CAAC,cAAc,YAAY,GAAG,EACrF,EAAa,CAEjB,cAAc,EAAgB,IAE9B,2BACF,EACM,EAAqB,CAAC,EAC5B,IAAK,IAAM,KAAa,EAAY,CAClC,IAAI,EACJ,GAAI,CACF,EAAU,EAAc,CAAS,CACnC,OAAS,EAAO,CACd,EAAS,KAAK,KAAK,EAAU,IAAI,EAAgB,CAAK,CAAC,CAAC,MAAM;CAAI,CAAC,CAAC,IAAI,EACxE,QACF,CACA,IAAM,EAAU,EAAQ,iBAAiB,EACzC,GAAI,IAAA,EAAoC,CACtC,EAAS,KACP,KAAK,EAAU,oBAAoB,GAAW,UAAU,gFAE1D,EACA,QACF,CAEA,MADA,GAAgB,EACT,CACT,CAMA,KALA,GAAgB,IAAI,EAClB,oDAAoD,EAAgB,EAAE,qHAEpD,EAAS,KAAK;CAAI,GACtC,EACM,CACR,CAQA,SAAS,GAA0B,CACjC,IAAM,EAAO,GAAG,QAAQ,SAAS,GAAG,QAAQ,OAS5C,OARI,QAAQ,WAAa,QAChB,GAAG,EAAK,OAEb,QAAQ,WAAa,SAIV,QAAQ,QAAQ,UAAU,EAAA,EAC1B,QAAQ,oBAAsB,GAAG,EAAK,MAAQ,GAAG,EAAK,OAJ5D,CAKX"}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "code-gauge",
-  "version": "6.0.0",
+  "version": "7.0.0",
   "description": "Measure code metrics with tree-sitter.",
   "keywords": [
     "cli",
@@ -41,7 +41,6 @@
     "native/code-gauge.wasm",
     "native/src/",
     "scripts/buildNative.mjs",
-    "scripts/installNative.mjs",
     "THIRD-PARTY-NOTICES.txt"
   ],
   "scripts": {
@@ -51,10 +50,8 @@
     "build-wasm": "node scripts/buildWasm.mjs",
     "cleanup": "bun wb lint --fix --format",
     "format": "bun wb lint --format",
-    "postinstall": "node scripts/installNative.mjs",
     "lint": "bun wb lint",
     "lint-fix": "bun wb lint --fix",
-    "prepare": "(bun lefthook install || true) && (test ! -f .lefthook/normalize-bun-lockfile.sh || bash .lefthook/normalize-bun-lockfile.sh)",
     "test": "bun wb test",
     "test/ci": "bun wb test-on-ci",
     "typecheck": "bun wb typecheck",
@@ -95,12 +92,12 @@
     "registry": "https://registry.npmjs.org/"
   },
   "optionalDependencies": {
-    "code-gauge-linux-x64-gnu": "6.0.0",
-    "code-gauge-linux-arm64-gnu": "6.0.0",
-    "code-gauge-linux-x64-musl": "6.0.0",
-    "code-gauge-linux-arm64-musl": "6.0.0",
-    "code-gauge-darwin-x64": "6.0.0",
-    "code-gauge-darwin-arm64": "6.0.0",
-    "code-gauge-win32-x64-msvc": "6.0.0"
+    "code-gauge-linux-x64-gnu": "7.0.0",
+    "code-gauge-linux-arm64-gnu": "7.0.0",
+    "code-gauge-linux-x64-musl": "7.0.0",
+    "code-gauge-linux-arm64-musl": "7.0.0",
+    "code-gauge-darwin-x64": "7.0.0",
+    "code-gauge-darwin-arm64": "7.0.0",
+    "code-gauge-win32-x64-msvc": "7.0.0"
   }
 }

package/scripts/installNative.mjs

@@ -1,96 +0,0 @@
-#!/usr/bin/env node
-// Postinstall hook ensuring the native Rust addon is available: a prebuilt platform package or an
-// already-built native/code-gauge.node is kept when it serves the payload version this checkout
-// expects (a stale addon surviving a `git pull` must not be kept — the runtime loader would
-// reject it), otherwise the addon is built from the bundled sources when a Rust toolchain is
-// available. Never fails the install — the whole body is guarded and the process always exits 0
-// (npm runs lifecycle scripts through cmd.exe on Windows, where a `|| true` suffix would not
-// work) — and the runtime loader raises a descriptive error on first use if no addon could be
-// provided.
-
-import { execFileSync } from 'node:child_process';
-import { existsSync, readFileSync, rmSync } from 'node:fs';
-import path from 'node:path';
-import { fileURLToPath } from 'node:url';
-
-const packageRoot = path.join(path.dirname(fileURLToPath(import.meta.url)), '..');
-
-// No explicit process.exit: nothing here keeps the event loop alive, so the process ends with
-// code 0 on its own once the synchronous work returns — and exit() could truncate the warning's
-// asynchronous stderr write (stderr is a pipe under npm, which is asynchronous on Windows).
-try {
-  installNativeAddon();
-} catch (error) {
-  warnBuildFailure(error);
-}
-
-function installNativeAddon() {
-  // The authoritative payload version lives in the bundled Rust source; a parse failure yields
-  // undefined, which no addon reports, so validation then always falls through to a fresh build.
-  const expectedPayloadVersion = /pub fn payload_version\(\) -> u32 \{\s*(\d+)/.exec(
-    readFileSync(path.join(packageRoot, 'native', 'src', 'lib.rs'), 'utf8')
-  )?.[1];
-
-  if (
-    servesExpectedPayload(`code-gauge-${platformTriplet()}`, expectedPayloadVersion) ||
-    servesExpectedPayload(path.join(packageRoot, 'native', 'code-gauge.node'), expectedPayloadVersion)
-  ) {
-    return;
-  }
-
-  try {
-    execFileSync(process.execPath, [path.join(packageRoot, 'scripts', 'buildNative.mjs')], { stdio: 'inherit' });
-  } catch (error) {
-    warnBuildFailure(error);
-    return;
-  }
-  // In an installed package the cargo build tree (hundreds of MB) buys nothing once the addon is
-  // copied out; a repository checkout (recognized by its sources) keeps it as the build cache.
-  // Guarded separately: a cleanup failure (e.g. a held Windows file handle) leaves a usable addon
-  // behind and must not be reported as a build failure.
-  if (!existsSync(path.join(packageRoot, 'src'))) {
-    try {
-      rmSync(path.join(packageRoot, 'native', 'target'), { recursive: true, force: true });
-    } catch {
-      // The leftover build tree only costs disk space.
-    }
-  }
-}
-
-// The platform-package suffix in the napi-rs naming convention: Linux is qualified by libc ABI
-// (a glibc-linked addon cannot load on Alpine/musl) and Windows by toolchain ABI. Must match
-// platformTriplet in src/nativeMetrics.ts.
-function platformTriplet() {
-  const base = `${process.platform}-${process.arch}`;
-  if (process.platform === 'win32') {
-    return `${base}-msvc`;
-  }
-  if (process.platform !== 'linux') {
-    return base;
-  }
-  return process.report?.getReport()?.header?.glibcVersionRuntime ? `${base}-gnu` : `${base}-musl`;
-}
-
-// Probed in a child process: loading the addon here would keep its library mapped for this
-// process's lifetime, and overwriting a mapped DLL is refused on Windows — exactly what the
-// stale-addon rebuild below must be able to do.
-function servesExpectedPayload(specifier, expectedPayloadVersion) {
-  try {
-    const reported = execFileSync(
-      process.execPath,
-      ['-p', `String(require(${JSON.stringify(specifier)}).payloadVersion?.())`],
-      // stderr is discarded: a missing platform package is the normal case, not a diagnostic.
-      { cwd: packageRoot, encoding: 'utf8', stdio: ['ignore', 'pipe', 'ignore'] }
-    ).trim();
-    return expectedPayloadVersion !== undefined && reported === expectedPayloadVersion;
-  } catch {
-    return false;
-  }
-}
-
-function warnBuildFailure(error) {
-  console.warn(
-    `code-gauge: could not build the native addon (${error instanceof Error ? error.message : String(error)}). ` +
-      'Install a Rust toolchain and run `node scripts/buildNative.mjs` in the package directory to enable it.'
-  );
-}