<?php
/*************************************************************
# 076 License

Copyright (c) テクニカル諏訪子

Permission is hereby granted to any person obtaining a copy of the software
Little Beast (the "Software") to use, modify, merge, copy, publish, distribute,
sublicense, and/or sell copies of the Software, subject to the following conditions:

    1. **Origin Attribution**:
       - You must not misrepresent the origin of the Software; you must not claim
         you created the original Software.
       - If the Software is used in a product, you must either:
         a. Provide clear attribution in the product's documentation, user interface,
            or other visible areas, **OR**
         b. Pay the original developers a fee they specify in writing.
    2. **Usage Restriction**:
       - The Software, or any derivative works, dependencies, or libraries
         incorporating it, must not be used for censorship or to suppress freedom of
         speech, expression, or creativity. Prohibited uses include, but are not
         limited to:
         - Censorship of so-called "hate speech", visuals, non-mainstream opinions,
           ideas, or objective reality.
         - Tools or systems designed to restrict access to information or
           artistic works.
    3. **Notice Preservation**:
       - This license and the above copyright notice must remain intact in all copies
         of the source code.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED,
INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A
PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF
CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE
OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 */
namespace Std\Lib\Gitler;

use Std\Lib\Git\Diff;
use Std\Lib\Git\GitException;
use Std\Lib\Git\Repository;
use Std\Lib\Git\Tree;

/**
 * リポジトリの言語の割合（GitHubのLinguistの様に）
 *
 * ファイルの名前・拡張子・#!から言語を決め、言語毎のバイト数を数える。数えるのはプログラミングとマークアップの言語だけで、
 * データ（JSON等）と文章（Markdown等）、外部のコード（vendor/等）、ドキュメント（docs/等）、生成されたファイル、リンクとサブモジュールは数えない。
 * ルートの.gitattributesのlinguist-language、linguist-vendored、linguist-generated、linguist-documentation、linguist-detectableで変えられる。
 * 結果はリポジトリの中のgitler-languages.jsonにツリー毎に保存する。
 */
final class Languages {
  /** 保存する形式（言語の表を変えたら上げる） */
  private const VERSION = 1;
  /** 保存するツリーの数 */
  private const KEEP = 20;
  /** これよりファイルが多いリポジトリは数えない */
  private const MAX_FILES = 100000;
  /** 中身で言語を決めるファイルの最大の大きさ */
  private const MAX_PEEK = 524288;
  /** 表示する言語の数（それより多ければ残りは「その他」） */
  private const SHOW = 8;

  private const PROGRAMMING = 'p';
  private const MARKUP = 'm';
  private const DATA = 'd';
  private const PROSE = 't';

  /** 名前 => [種類, 色]（色はLinguistと同じ、MaronとSlangは独自） */
  private const LANGS = [
    'Ada' => ['p', '#02f88c'], 'Assembly' => ['p', '#6E4C13'], 'AWK' => ['p', '#c30e9b'], 'Batchfile' => ['p', '#C1F12E'],
    'C' => ['p', '#555555'], 'C#' => ['p', '#178600'], 'C++' => ['p', '#f34b7d'], 'Clojure' => ['p', '#db5855'],
    'CMake' => ['p', '#DA3434'], 'Common Lisp' => ['p', '#3fb68b'], 'Coq' => ['p', '#d0b68c'], 'Crystal' => ['p', '#000100'],
    'CSS' => ['m', '#663399'], 'CUDA' => ['p', '#3A4E3A'], 'Cython' => ['p', '#fedf5b'], 'D' => ['p', '#ba595e'],
    'Dart' => ['p', '#00B4AB'], 'Dockerfile' => ['p', '#384d54'], 'Elixir' => ['p', '#6e4a7e'], 'Elm' => ['p', '#60B5CC'],
    'Emacs Lisp' => ['p', '#c065db'], 'Erlang' => ['p', '#B83998'], 'F#' => ['p', '#b845fc'], 'Forth' => ['p', '#341708'],
    'Fortran' => ['p', '#4d41b1'], 'GDScript' => ['p', '#355570'], 'Gleam' => ['p', '#ffaff3'], 'GLSL' => ['p', '#5686a5'],
    'Go' => ['p', '#00ADD8'], 'Groovy' => ['p', '#4298b8'], 'Handlebars' => ['m', '#f7931e'], 'Hare' => ['p', '#9d7424'],
    'Haskell' => ['p', '#5e5086'], 'Haxe' => ['p', '#df7900'], 'HCL' => ['p', '#844FBA'], 'HLSL' => ['p', '#aace60'],
    'HTML' => ['m', '#e34c26'], 'Java' => ['p', '#b07219'], 'JavaScript' => ['p', '#f1e05a'], 'Julia' => ['p', '#a270ba'],
    'Jupyter Notebook' => ['m', '#DA5B0B'], 'Just' => ['p', '#384d54'], 'Kotlin' => ['p', '#A97BFF'], 'Less' => ['m', '#1d365d'],
    'Lex' => ['p', '#DBCA00'], 'Lua' => ['p', '#000080'], 'Makefile' => ['p', '#427819'], 'Maron' => ['m', '#ea79d8'],
    'MATLAB' => ['p', '#e16737'], 'Meson' => ['p', '#007800'], 'Metal' => ['p', '#8f14e9'], 'Nim' => ['p', '#ffc200'],
    'Nix' => ['p', '#7e7eff'], 'Objective-C' => ['p', '#438eff'], 'Objective-C++' => ['p', '#6866fb'], 'OCaml' => ['p', '#ef7a08'],
    'Odin' => ['p', '#60AFFE'], 'Pascal' => ['p', '#E3F171'], 'Perl' => ['p', '#0298c3'], 'PHP' => ['p', '#4F5D95'],
    'PowerShell' => ['p', '#012456'], 'Python' => ['p', '#3572A5'], 'QML' => ['p', '#44a51c'], 'R' => ['p', '#198CE7'],
    'Racket' => ['p', '#3c5caa'], 'Raku' => ['p', '#0000fb'], 'Roff' => ['m', '#ecdebe'], 'Ruby' => ['p', '#701516'],
    'Rust' => ['p', '#dea584'], 'Sass' => ['m', '#a53b70'], 'Scala' => ['p', '#c22d40'], 'Scheme' => ['p', '#1e4aec'],
    'SCSS' => ['m', '#c6538c'], 'Shell' => ['p', '#89e051'], 'Slang' => ['p', '#1fbec9'], 'Solidity' => ['p', '#AA6746'],
    'Starlark' => ['p', '#76d275'], 'Svelte' => ['m', '#ff3e00'], 'Swift' => ['p', '#F05138'], 'SystemVerilog' => ['p', '#DAE1C2'],
    'Tcl' => ['p', '#e4cc98'], 'TeX' => ['m', '#3D6117'], 'TSX' => ['p', '#3178c6'], 'Twig' => ['m', '#c1d026'],
    'TypeScript' => ['p', '#3178c6'], 'V' => ['p', '#4f87c4'], 'Verilog' => ['p', '#b2b7f8'], 'VHDL' => ['p', '#adb2cb'],
    'Vim Script' => ['p', '#199f4b'], 'Visual Basic .NET' => ['p', '#945db7'], 'Vue' => ['m', '#41b883'], 'WGSL' => ['p', '#1a5e9a'],
    'Yacc' => ['p', '#4B6C4B'], 'Zig' => ['p', '#ec915c'],
    // 数えない（linguist-detectableで数えられる）
    'JSON' => ['d', '#292929'], 'Markdown' => ['t', '#083fa1'], 'SQL' => ['d', '#e38c00'], 'TOML' => ['d', '#9c4221'],
    'XML' => ['d', '#0060ac'], 'YAML' => ['d', '#cb171e'],
  ];

  /** 拡張子（小文字） => 言語 */
  private const EXTENSIONS = [
    'adb' => 'Ada', 'ads' => 'Ada', 'asm' => 'Assembly', 'nasm' => 'Assembly', 's' => 'Assembly', 'awk' => 'AWK',
    'bat' => 'Batchfile', 'cmd' => 'Batchfile', 'c' => 'C', 'cs' => 'C#', 'csx' => 'C#',
    'cpp' => 'C++', 'cc' => 'C++', 'cxx' => 'C++', 'c++' => 'C++', 'hpp' => 'C++', 'hh' => 'C++', 'hxx' => 'C++', 'h++' => 'C++',
    'ipp' => 'C++', 'tpp' => 'C++', 'inl' => 'C++', 'ixx' => 'C++', 'cppm' => 'C++',
    'clj' => 'Clojure', 'cljs' => 'Clojure', 'cljc' => 'Clojure', 'cmake' => 'CMake', 'lisp' => 'Common Lisp', 'lsp' => 'Common Lisp',
    'cr' => 'Crystal', 'css' => 'CSS', 'cu' => 'CUDA', 'cuh' => 'CUDA', 'pyx' => 'Cython', 'pxd' => 'Cython', 'dart' => 'Dart',
    'dockerfile' => 'Dockerfile', 'ex' => 'Elixir', 'exs' => 'Elixir', 'elm' => 'Elm', 'el' => 'Emacs Lisp', 'erl' => 'Erlang', 'hrl' => 'Erlang',
    'fsi' => 'F#', 'fsx' => 'F#', 'fth' => 'Forth', '4th' => 'Forth', 'forth' => 'Forth',
    'f' => 'Fortran', 'for' => 'Fortran', 'f77' => 'Fortran', 'f90' => 'Fortran', 'f95' => 'Fortran', 'f03' => 'Fortran', 'f08' => 'Fortran',
    'gd' => 'GDScript', 'gleam' => 'Gleam', 'glsl' => 'GLSL', 'vert' => 'GLSL', 'frag' => 'GLSL', 'geom' => 'GLSL', 'comp' => 'GLSL',
    'tesc' => 'GLSL', 'tese' => 'GLSL', 'rgen' => 'GLSL', 'rchit' => 'GLSL', 'rmiss' => 'GLSL', 'rahit' => 'GLSL', 'mesh' => 'GLSL', 'task' => 'GLSL',
    'go' => 'Go', 'groovy' => 'Groovy', 'gradle' => 'Groovy', 'hbs' => 'Handlebars', 'handlebars' => 'Handlebars', 'ha' => 'Hare',
    'hs' => 'Haskell', 'lhs' => 'Haskell', 'hx' => 'Haxe', 'tf' => 'HCL', 'hcl' => 'HCL', 'hlsl' => 'HLSL', 'hlsli' => 'HLSL', 'fx' => 'HLSL', 'fxh' => 'HLSL',
    'html' => 'HTML', 'htm' => 'HTML', 'xhtml' => 'HTML', 'java' => 'Java', 'js' => 'JavaScript', 'mjs' => 'JavaScript', 'cjs' => 'JavaScript',
    'jsx' => 'JavaScript', 'jl' => 'Julia', 'ipynb' => 'Jupyter Notebook', 'just' => 'Just', 'kt' => 'Kotlin', 'kts' => 'Kotlin', 'less' => 'Less',
    'l' => 'Lex', 'lex' => 'Lex', 'lua' => 'Lua', 'mk' => 'Makefile', 'mak' => 'Makefile', 'maron' => 'Maron', 'metal' => 'Metal',
    'nim' => 'Nim', 'nims' => 'Nim', 'nimble' => 'Nim', 'nix' => 'Nix', 'mm' => 'Objective-C++', 'ml' => 'OCaml', 'mli' => 'OCaml',
    'odin' => 'Odin', 'pas' => 'Pascal', 'pp' => 'Pascal', 'dpr' => 'Pascal', 'lpr' => 'Pascal', 'pl' => 'Perl', 'pm' => 'Perl',
    'php' => 'PHP', 'phtml' => 'PHP', 'php3' => 'PHP', 'php4' => 'PHP', 'php5' => 'PHP', 'phps' => 'PHP',
    'ps1' => 'PowerShell', 'psm1' => 'PowerShell', 'psd1' => 'PowerShell', 'py' => 'Python', 'pyw' => 'Python', 'pyi' => 'Python',
    'qml' => 'QML', 'r' => 'R', 'rkt' => 'Racket', 'raku' => 'Raku', 'rakumod' => 'Raku', 'p6' => 'Raku',
    'roff' => 'Roff', 'man' => 'Roff', 'mdoc' => 'Roff', '1' => 'Roff', '2' => 'Roff', '3' => 'Roff', '4' => 'Roff', '5' => 'Roff',
    '6' => 'Roff', '7' => 'Roff', '8' => 'Roff', '9' => 'Roff',
    'rb' => 'Ruby', 'rake' => 'Ruby', 'gemspec' => 'Ruby', 'rs' => 'Rust', 'sass' => 'Sass', 'scala' => 'Scala', 'scm' => 'Scheme', 'ss' => 'Scheme',
    'scss' => 'SCSS', 'sh' => 'Shell', 'bash' => 'Shell', 'zsh' => 'Shell', 'ksh' => 'Shell', 'slang' => 'Slang', 'sol' => 'Solidity',
    'bzl' => 'Starlark', 'star' => 'Starlark', 'svelte' => 'Svelte', 'swift' => 'Swift', 'sv' => 'SystemVerilog', 'svh' => 'SystemVerilog',
    'tcl' => 'Tcl', 'tex' => 'TeX', 'sty' => 'TeX', 'ltx' => 'TeX', 'tsx' => 'TSX', 'twig' => 'Twig', 'ts' => 'TypeScript', 'mts' => 'TypeScript',
    'cts' => 'TypeScript', 'vhd' => 'VHDL', 'vhdl' => 'VHDL', 'vim' => 'Vim Script', 'vb' => 'Visual Basic .NET', 'vue' => 'Vue',
    'wgsl' => 'WGSL', 'y' => 'Yacc', 'yy' => 'Yacc', 'zig' => 'Zig', 'zon' => 'Zig',
    'json' => 'JSON', 'md' => 'Markdown', 'markdown' => 'Markdown', 'sql' => 'SQL', 'toml' => 'TOML', 'xml' => 'XML', 'yml' => 'YAML', 'yaml' => 'YAML',
  ];

  /** 拡張子で決まらない物（中身で決める、決まらなければこの言語） */
  private const AMBIGUOUS = ['h' => 'C', 'm' => 'Objective-C', 'v' => 'V', 'fs' => 'F#', 'd' => 'D'];

  /** ファイル名 => 言語 */
  private const FILENAMES = [
    'Makefile' => 'Makefile', 'makefile' => 'Makefile', 'GNUmakefile' => 'Makefile', 'BSDmakefile' => 'Makefile', 'Kbuild' => 'Makefile',
    'Makefile.am' => 'Makefile', 'CMakeLists.txt' => 'CMake', 'meson.build' => 'Meson', 'meson_options.txt' => 'Meson', 'meson.options' => 'Meson',
    'Dockerfile' => 'Dockerfile', 'Containerfile' => 'Dockerfile', 'Rakefile' => 'Ruby', 'Gemfile' => 'Ruby', 'Vagrantfile' => 'Ruby',
    'Podfile' => 'Ruby', 'Justfile' => 'Just', 'justfile' => 'Just', '.justfile' => 'Just', 'BUILD' => 'Starlark', 'BUILD.bazel' => 'Starlark',
    'WORKSPACE' => 'Starlark', 'WORKSPACE.bazel' => 'Starlark', 'MODULE.bazel' => 'Starlark', '.bashrc' => 'Shell', '.bash_profile' => 'Shell',
    '.bash_logout' => 'Shell', '.profile' => 'Shell', '.zshrc' => 'Shell', '.zprofile' => 'Shell', '.kshrc' => 'Shell', '.xinitrc' => 'Shell',
    '.xsession' => 'Shell', 'PKGBUILD' => 'Shell', 'APKBUILD' => 'Shell', '.vimrc' => 'Vim Script', '_vimrc' => 'Vim Script',
    '.gvimrc' => 'Vim Script', '.emacs' => 'Emacs Lisp', 'SConstruct' => 'Python', 'SConscript' => 'Python', 'Jenkinsfile' => 'Groovy',
  ];

  /** #!のコマンド（後ろの数字を除く） => 言語 */
  private const INTERPRETERS = [
    'sh' => 'Shell', 'bash' => 'Shell', 'zsh' => 'Shell', 'ksh' => 'Shell', 'mksh' => 'Shell', 'oksh' => 'Shell', 'pdksh' => 'Shell',
    'dash' => 'Shell', 'ash' => 'Shell', 'rc' => 'Shell', 'fish' => 'Shell', 'python' => 'Python', 'pypy' => 'Python', 'ruby' => 'Ruby',
    'jruby' => 'Ruby', 'perl' => 'Perl', 'raku' => 'Raku', 'php' => 'PHP', 'node' => 'JavaScript', 'nodejs' => 'JavaScript',
    'bun' => 'JavaScript', 'deno' => 'TypeScript', 'lua' => 'Lua', 'luajit' => 'Lua', 'tclsh' => 'Tcl', 'wish' => 'Tcl', 'awk' => 'AWK',
    'gawk' => 'AWK', 'nawk' => 'AWK', 'mawk' => 'AWK', 'rscript' => 'R', 'make' => 'Makefile', 'gmake' => 'Makefile', 'bmake' => 'Makefile',
    'julia' => 'Julia', 'elixir' => 'Elixir', 'escript' => 'Erlang', 'guile' => 'Scheme', 'csi' => 'Scheme', 'racket' => 'Racket',
    'sbcl' => 'Common Lisp', 'clisp' => 'Common Lisp', 'ecl' => 'Common Lisp', 'ocaml' => 'OCaml', 'runghc' => 'Haskell',
    'runhaskell' => 'Haskell', 'crystal' => 'Crystal', 'pwsh' => 'PowerShell', 'dart' => 'Dart', 'swift' => 'Swift', 'groovy' => 'Groovy',
    'scala' => 'Scala', 'emacs' => 'Emacs Lisp',
  ];

  /** linguist-languageの別名（小文字） */
  private const ALIASES = [
    'cpp' => 'C++', 'csharp' => 'C#', 'js' => 'JavaScript', 'ts' => 'TypeScript', 'sh' => 'Shell', 'bash' => 'Shell', 'golang' => 'Go',
    'py' => 'Python', 'rb' => 'Ruby', 'objc' => 'Objective-C', 'objcpp' => 'Objective-C++', 'asm' => 'Assembly', 'make' => 'Makefile',
    'vim' => 'Vim Script', 'viml' => 'Vim Script', 'elisp' => 'Emacs Lisp', 'md' => 'Markdown', 'yml' => 'YAML', 'vbnet' => 'Visual Basic .NET',
    'terraform' => 'HCL', 'latex' => 'TeX', 'groff' => 'Roff', 'nroff' => 'Roff', 'manpage' => 'Roff',
  ];

  /** 外部のコード（Linguistのvendor.ymlの主な物） */
  private const VENDORED = '~(^|/)(vendor|vendors|node_modules|bower_components|jspm_packages|third[_-]?party|3rd[_-]?party|externals?|extern|deps|'
    .'Pods|Carthage|Godeps|\.git|\.svn|\.hg|\.vscode|\.idea|\.yarn|\.bundle)/'
    .'|(^|/)(gradlew(\.bat)?|mvnw(\.cmd)?|config\.(guess|sub|status)|ltmain\.sh|install-sh|missing|depcomp|aclocal\.m4|Makefile\.in)$'
    .'|\.min\.(js|css)$|-min\.js$|(^|/)(jquery|bootstrap)[^/]*\.(js|css)$~i';
  /** ドキュメント */
  private const DOCUMENTATION = '~^(docs?|documentation|man|examples?|samples?|demos?)/|(^|/)(javadoc|groovydoc)/'
    .'|(^|/)(README|CHANGELOG|CHANGES|LICEN[CS]E|COPYING|CONTRIBUTING|INSTALL|NEWS|AUTHORS|THANKS|HISTORY)(\.[^/]*)?$~i';
  /** 生成されたファイル */
  private const GENERATED = '~(\.pb\.(go|cc|h)|_pb2\.py|\.designer\.cs|\.g\.dart|\.freezed\.dart)$'
    .'|(^|/)(package-lock\.json|yarn\.lock|pnpm-lock\.yaml|Cargo\.lock|composer\.lock|go\.sum)$~';

  /** 言語の色（知らない言語は灰色） */
  public static function color(string $name): string {
    return self::LANGS[$name][1] ?? '#cccccc';
  }

  /**
   * 言語毎のバイト数（保存した物が有ればそれ）
   *
   * @param bool $head  既定のブランチ（リポジトリの一覧に出す言語）
   * @return array|null  言語 => バイト数（多い順）。ファイルが多過ぎればnull
   */
  public static function get(Repository $repo, string $treeish, bool $head = false): ?array {
    $tree = $repo->treeOf($treeish);
    if ($tree === null) return null;
    $file = $repo->path.'/gitler-languages.json';
    $c = self::load($file);
    if (isset($c['trees'][$tree])) {
      if ($head && $c['head'] !== $tree) {
        $c['head'] = $tree;
        self::save($repo, $file, $c);
      }
      return $c['trees'][$tree];
    }
    $bytes = self::stats($repo, $tree);
    if ($bytes === null) return null;
    $c['trees'][$tree] = $bytes;
    if ($head) $c['head'] = $tree;
    // 古い物から消す（既定のブランチの物は残す）
    foreach (array_keys($c['trees']) as $k) {
      if (count($c['trees']) <= self::KEEP) break;
      if ($k !== $c['head'] && $k !== $tree) unset($c['trees'][$k]);
    }
    self::save($repo, $file, $c);
    return $bytes;
  }

  /**
   * 既定のブランチの一番多い言語（保存した物だけ見る）
   *
   * @param string $dir  リポジトリのディレクトリ
   * @return array|null  name、color
   */
  public static function primary(string $dir): ?array {
    $c = self::load($dir.'/gitler-languages.json');
    $name = array_key_first($c['trees'][$c['head']] ?? []);
    return $name === null ? null : ['name' => (string)$name, 'color' => self::color((string)$name)];
  }

  /**
   * 表示用（多い順、多過ぎれば残りは「その他」）
   *
   * @return array  name、color、percent（「12.3%」）、width（CSSの幅）
   */
  public static function present(array $bytes): array {
    $total = array_sum($bytes);
    if ($total <= 0) return [];
    arsort($bytes);
    if (count($bytes) > self::SHOW) {
      $rest = array_slice($bytes, self::SHOW - 1, null, true);
      $bytes = array_slice($bytes, 0, self::SHOW - 1, true) + ['その他' => array_sum($rest)];
    }
    $out = [];
    foreach ($bytes as $name => $n) {
      $p = $n * 100 / $total;
      $out[] = ['name' => (string)$name, 'color' => $name === 'その他' ? '#746c75' : self::color((string)$name),
                'percent' => $p < 0.1 ? '<0.1%' : number_format($p, 1).'%', 'width' => sprintf('%.3F', $p)];
    }
    return $out;
  }

  /**
   * ツリーの言語毎のバイト数を数える
   *
   * @return array|null  言語 => バイト数（多い順）。ファイルが多過ぎればnull
   */
  public static function stats(Repository $repo, string $treeish): ?array {
    $root = $repo->treeOf($treeish);
    if ($root === null) return null;
    $rules = self::attributes($repo, $root);
    // 除外を取り消す.gitattributesが無ければ、除外するディレクトリの中は見ない
    $undo = false;
    foreach ($rules as [, $set]) {
      foreach (['linguist-vendored', 'linguist-generated', 'linguist-documentation'] as $k) if (array_key_exists($k, $set) && $set[$k] !== true) $undo = true;
    }
    $bytes = [];
    $files = 0;
    $stack = [['', $root]];
    while (!empty($stack)) {
      [$dir, $id] = array_pop($stack);
      foreach ($repo->tree($id)->entries as [$mode, $name, $eid]) {
        $path = $dir.$name;
        if ($mode === Tree::DIR) {
          if ($undo || !preg_match(self::VENDORED, $path.'/') && !preg_match(self::DOCUMENTATION, $path.'/')) $stack[] = [$path.'/', $eid];
          continue;
        }
        // リンクとサブモジュールは数えない
        if ($mode !== Tree::FILE && $mode !== Tree::EXEC) continue;
        if (++$files > self::MAX_FILES) return null;
        $size = 0;
        $lang = self::file($repo, $path, $mode, $eid, self::match($rules, $path), $size);
        if ($lang !== null) $bytes[$lang] = ($bytes[$lang] ?? 0) + $size;
      }
    }
    arsort($bytes);
    return $bytes;
  }

  /**
   * 1つのファイルの言語（数えなければnull）
   *
   * @param array $a  .gitattributesの属性
   */
  private static function file(Repository $repo, string $path, string $mode, string $id, array $a, int &$size): ?string {
    if ($a['linguist-vendored'] ?? (bool)preg_match(self::VENDORED, $path)) return null;
    if ($a['linguist-generated'] ?? (bool)preg_match(self::GENERATED, $path)) return null;
    if ($a['linguist-documentation'] ?? (bool)preg_match(self::DOCUMENTATION, $path)) return null;
    $size = (int)($repo->objects->info($id)[1] ?? 0);
    $base = basename($path);
    $dot = strrpos($base, '.');
    $ext = $dot === false || $dot === 0 ? '' : strtolower(substr($base, $dot + 1));
    $lang = isset($a['linguist-language']) ? self::named((string)$a['linguist-language']) : null;
    $lang ??= self::FILENAMES[$base] ?? null;
    if ($lang === null && preg_match('/^((GNU|BSD)?[Mm]akefile|Dockerfile|Containerfile)\./', $base, $m)) $lang = str_ends_with($m[1], 'akefile') ? 'Makefile' : 'Dockerfile';
    $lang ??= self::EXTENSIONS[$ext] ?? null;
    // 中身を見る：拡張子で決まらない物と、拡張子の無いスクリプト
    if ($lang === null && (isset(self::AMBIGUOUS[$ext]) || $ext === '' || $ext === 'cgi' || $mode === Tree::EXEC)) {
      if ($size > self::MAX_PEEK) {
        $lang = self::AMBIGUOUS[$ext] ?? null;
      } else {
        $data = $repo->blob($id);
        if (Diff::isBinary($data)) return null;
        // Autoconfのconfigure等は生成された物
        if (str_contains(substr($data, 0, 4096), 'Generated by GNU Autoconf')) return null;
        $lang = isset(self::AMBIGUOUS[$ext]) ? self::guess($ext, $data) : self::shebang($data);
      }
    }
    if ($lang === null) return null;
    $type = self::LANGS[$lang][0] ?? self::PROGRAMMING;
    return ($a['linguist-detectable'] ?? ($type === self::PROGRAMMING || $type === self::MARKUP)) ? $lang : null;
  }

  /** linguist-languageの名前（大文字・小文字、空白・「-」・「_」は区別しない） */
  private static function named(string $value): string {
    $key = strtolower($value);
    if (isset(self::ALIASES[$key])) return self::ALIASES[$key];
    $norm = fn(string $s): string => str_replace([' ', '-', '_'], '', strtolower($s));
    foreach (array_keys(self::LANGS) as $name) if ($norm($name) === $norm($value)) return $name;
    return $value;
  }

  /** 拡張子で決まらない物を中身で決める */
  private static function guess(string $ext, string $data): ?string {
    return match ($ext) {
      'h' => preg_match('/^\s*(@interface|@implementation|@protocol|@end|#import)\b/m', $data) ? 'Objective-C'
           : (preg_match('/^\s*(class|namespace|template\s*<|using\s+namespace)\b|\bstd::|^\s*(public|private|protected)\s*:|#include\s*<(iostream|string|vector|map|memory|algorithm|cstd\w+)>/m', $data) ? 'C++' : 'C'),
      'm' => preg_match('/^\s*(@interface|@implementation|@protocol|@end|#import|#include)\b/m', $data) ? 'Objective-C'
           : (preg_match('/^\s*(function\b|%)/m', $data) ? 'MATLAB' : 'Objective-C'),
      'v' => preg_match('/^\s*module\s+\w+[\s\S]*^\s*endmodule\b/m', $data) ? 'Verilog'
           : (preg_match('/^\s*(Theorem|Lemma|Proof|Qed|Require\s+Import|Inductive)\b/m', $data) ? 'Coq' : 'V'),
      'fs' => preg_match('/^\s*#version\s+\d+|\bgl_Frag\w+|^\s*(uniform|in|out)\s+(vec|float|int|mat|sampler)\w*\s+\w+\s*;/m', $data) ? 'GLSL' : 'F#',
      // makeの依存関係のファイル（生成された物）
      'd' => preg_match('/^[^\s:#]+\.o\s*:/m', $data) ? null : 'D',
      default => self::AMBIGUOUS[$ext] ?? null,
    };
  }

  /** #!から言語を決める（「#!/usr/bin/env -S php」等） */
  private static function shebang(string $data): ?string {
    if (!str_starts_with($data, '#!')) return null;
    $words = preg_split('/\s+/', trim(strtok(substr($data, 2), "\n")));
    $prog = basename((string)array_shift($words));
    if ($prog === 'env') {
      $prog = '';
      foreach ($words as $w) {
        if ($w === '' || $w[0] === '-' || str_contains($w, '=')) continue;
        $prog = basename($w);
        break;
      }
    }
    $prog = (string)preg_replace('/[\d.\-]+$/', '', strtolower($prog));
    return self::INTERPRETERS[$prog] ?? null;
  }

  //////// .gitattributes

  /**
   * ルートの.gitattributesのlinguist-*の属性
   *
   * @return array  [正規表現, 属性 => true、false、null（未指定）又は言語]の配列
   */
  private static function attributes(Repository $repo, string $root): array {
    $e = $repo->tree($root)->find('.gitattributes');
    if ($e === null || ($e[0] !== Tree::FILE && $e[0] !== Tree::EXEC)) return [];
    $rules = [];
    foreach (preg_split('/\r?\n/', $repo->blob($e[2])) as $line) {
      $line = trim($line);
      if ($line === '' || $line[0] === '#') continue;
      $words = preg_split('/\s+/', $line);
      $pattern = (string)array_shift($words);
      $set = [];
      foreach ($words as $w) {
        if (!preg_match('/^([-!]?)(linguist-[a-z]+)(?:=(.*))?$/', $w, $m)) continue;
        if ($m[1] === '!') $set[$m[2]] = null;
        else if ($m[2] === 'linguist-language') $set[$m[2]] = $m[1] === '' && ($m[3] ?? '') !== '' ? $m[3] : null;
        else $set[$m[2]] = $m[1] === '-' ? false : !in_array(strtolower($m[3] ?? 'true'), ['false', '0', 'no'], true);
      }
      if (!empty($set)) $rules[] = [self::glob($pattern), $set];
    }
    return $rules;
  }

  /** 後に書いた物が優先 */
  private static function match(array $rules, string $path): array {
    $a = [];
    foreach ($rules as [$re, $set]) if (preg_match($re, $path)) $a = array_merge($a, $set);
    return $a;
  }

  /** .gitattributesのパターンを正規表現にする（「/」が無ければどの深さの名前にも一致する） */
  public static function glob(string $pattern): string {
    // ディレクトリだけのパターンはファイルに一致しない
    if (str_ends_with($pattern, '/')) return '~(?!)~';
    $anchored = str_contains($pattern, '/');
    if (str_starts_with($pattern, '/')) $pattern = substr($pattern, 1);
    $re = '';
    $len = strlen($pattern);
    for ($i = 0; $i < $len; $i++) {
      $ch = $pattern[$i];
      $start = $i === 0 || $pattern[$i - 1] === '/';
      if ($ch === '*' && $start && substr($pattern, $i, 3) === '**/') {
        $re .= '(?:.*/)?';
        $i += 2;
      } else if ($ch === '*' && $start && $i + 2 === $len && substr($pattern, $i, 2) === '**') {
        $re .= '.*';
        $i++;
      } else if ($ch === '*') {
        $re .= '[^/]*';
      } else if ($ch === '?') {
        $re .= '[^/]';
      } else if ($ch === '\\' && $i + 1 < $len) {
        $re .= preg_quote($pattern[++$i], '~');
      } else if ($ch === '[' && ($end = strpos($pattern, ']', $i + 2)) !== false) {
        $class = substr($pattern, $i + 1, $end - $i - 1);
        if ($class[0] === '!') $class = '^'.substr($class, 1);
        $re .= '['.str_replace(['\\', '~'], ['\\\\', '\~'], $class).']';
        $i = $end;
      } else {
        $re .= preg_quote($ch, '~');
      }
    }
    return '~^'.($anchored ? '' : '(?:.*/)?').$re.'$~';
  }

  //////// 保存

  private static function load(string $file): array {
    $c = json_decode((string)@file_get_contents($file), true);
    if (!is_array($c) || ($c['version'] ?? 0) !== self::VERSION || !is_array($c['trees'] ?? null)) return ['version' => self::VERSION, 'head' => '', 'trees' => []];
    $c['head'] = (string)($c['head'] ?? '');
    return $c;
  }

  private static function save(Repository $repo, string $file, array $c): void {
    try {
      $repo->writeFile($file, (string)json_encode($c, JSON_UNESCAPED_UNICODE | JSON_UNESCAPED_SLASHES));
    } catch (GitException) {
      // 保存出来なくても表示は出来る
    }
  }
}