Files
utun2/tools/chatgui/src/code_language.cpp
T

101 lines
7.6 KiB
C++

#include "code_language.h"
#include "debug_ui.h"
#include <QRegularExpression>
#include <array>
namespace {
enum Language { C, Cpp, Java, Kotlin, Rust, Php, JavaScript, Css, Html, LanguageCount };
constexpr const char* names[] = {"C", "C++", "Java", "Kotlin", "Rust", "PHP/PHP", "JavaScript", "CSS", "HTML"};
struct Rule {
Language language;
int weight;
QRegularExpression pattern;
};
// Каждый признак учитывается один раз: повторение общего слова не повышает уверенность.
const Rule rules[] = {
{C, 3, QRegularExpression(R"(^\s*#\s*include[ \t]+)", QRegularExpression::MultilineOption)},
{C, 2, QRegularExpression(R"(^\s*#\s*(?:define|ifdef|ifndef)\s+\w+)", QRegularExpression::MultilineOption)},
{C, 2, QRegularExpression(R"(\bunsigned\s+(?:(?:long|int|char)\s+)?\w+\s*[;=])")},
{C, 4, QRegularExpression(R"(\b(?:void|int|char|double|float|size_t)\s+\**\s*\w+\s*\([^;{}]*\)\s*\{)")},
{C, 2, QRegularExpression(R"(\b(?:typedef\s+(?:struct|enum)|malloc\s*\(|calloc\s*\(|sizeof\s*\())")},
{C, 2, QRegularExpression(R"(\b(?:const\s+)?(?:char|void|int)\s*\*+\s*\w+)")},
{Cpp, 8, QRegularExpression(R"(\bstd\s*::|\b(?:static_cast|dynamic_cast|reinterpret_cast|const_cast)\s*<)")},
{Cpp, 7, QRegularExpression(R"(^\s*#\s*include\s*<(?:iostream|vector|string|memory|algorithm|map|utility|array|set|unordered_map)>)",
QRegularExpression::MultilineOption)},
{Cpp, 6, QRegularExpression(R"(\btemplate\s*<|\bnamespace\s+[\w:]+\s*\{)")},
{Cpp, 4, QRegularExpression(R"(\b(?:class|struct)\s+\w+(?:\s*:\s*[^\n{]+)?\s*\{)")},
{Cpp, 3, QRegularExpression(R"(\b(?:public|private|protected)\s*:|\b\w+::~?\w+\s*\()")},
{Java, 3, QRegularExpression(R"(^\s*(?:package|import)\s+(?:static\s+)?[\w.*]+\s*;)", QRegularExpression::MultilineOption)},
{Java, 4, QRegularExpression(R"(\b(?:public|protected|private)\s+(?:(?:static|final|abstract)\s+)*(?:class|interface|enum)\s+\w+)")},
{Java, 7, QRegularExpression(R"(\b(?:public|protected|private)\s+(?:(?:static|final|synchronized)\s+)*)"
R"((?:void|boolean|int|long|double|float|char|byte|short|[A-Z]\w*)\s+\w+\s*\()")},
{Java, 4, QRegularExpression(R"(\b(?:implements|extends)\s+[A-Z]\w*|\bSystem\.(?:out|err)\.)")},
{Kotlin, 6, QRegularExpression(R"(\bfun\s+(?:<[^>\n]+>\s*)?[\w.]+\s*\()")},
{Kotlin, 3, QRegularExpression(R"(^\s*(?:(?:private|public|internal|override|const)\s+)*(?:val|var)\s+\w+\s*[:=])",
QRegularExpression::MultilineOption)},
{Kotlin, 5, QRegularExpression(R"(\b(?:data\s+class|sealed\s+class|companion\s+object)\b|\bwhen\s*\([^\n]*\)\s*\{)")},
{Rust, 7, QRegularExpression(R"(\bfn\s+\w+\s*(?:<[^>\n]+>)?\s*\()")},
{Rust, 7, QRegularExpression(R"(^\s*(?:pub(?:\([^)]*\))?\s+)?(?:use\s+[\w:]+\s*(?:;|::|\{)|impl(?:\s*<[^>\n]+>)?\s+\w+))",
QRegularExpression::MultilineOption)},
{Rust, 3, QRegularExpression(R"(\blet\s+mut\s+|\b(?:println|vec|assert_eq|format)!\s*[(\[]|#\[derive\s*\()")},
{Php, 4, QRegularExpression(R"(\$[a-zA-Z_]\w*\s*(?:->|\[|=))")},
{Php, 4, QRegularExpression(R"(\bfunction\s+\w+\s*\([^\n)]*\$|\b(?:public|private|protected)\s+(?:static\s+)?function\b)")},
{Php, 3, QRegularExpression(R"(\b(?:echo|require_once|include_once)\b|\barray\s*\(|\$this\b)")},
{JavaScript, 4, QRegularExpression(R"(\b(?:const|var)\s+[a-zA-Z_$][\w$]*\s*=)")},
{JavaScript, 2, QRegularExpression(R"(\blet\s+[a-zA-Z_$][\w$]*\s*=)")},
{JavaScript, 3, QRegularExpression(R"(=>|\bfunction\s*[\w$]*\s*\()")},
{JavaScript, 4, QRegularExpression(R"(\b(?:module\.exports|exports\.\w+|console\.\w+|document\.\w+)\b)")},
{JavaScript, 4, QRegularExpression(R"(^\s*(?:export\s+(?:default|function|const|class)\b|import\s+[^;\n]+\s+from\b))",
QRegularExpression::MultilineOption)},
{Css, 3, QRegularExpression(R"((?:^|[}\n])\s*[.#]?[\w*-][\w\s.#:>+~,*\[\]="'-]*\s*\{)")},
{Css, 4, QRegularExpression(R"((?:^|[;{\n])\s*(?:color|background(?:-color)?|font(?:-size|-family|-weight)?|display|position)\s*:[^;{}\n]+;)")},
{Css, 4, QRegularExpression(R"((?:^|[;{\n])\s*(?:margin(?:-\w+)?|padding(?:-\w+)?|border(?:-\w+)?|width|height|opacity|align-items))"
R"(\s*:[^;{}\n]+;)")},
};
}
QString detectCodeLanguage(QStringView source) {
const QString sample = source.left(16384).toString();
// Явные заголовки контейнеров важнее вложенного CSS/JS/PHP-кода.
static const QRegularExpression php(R"(^\s*<\?php\b)", QRegularExpression::CaseInsensitiveOption);
static const QRegularExpression html(R"(^\s*(?:<!doctype\s+html\b|<html\b))", QRegularExpression::CaseInsensitiveOption);
if (php.match(sample).hasMatch() || html.match(sample).hasMatch()) {
const QString language = QString::fromLatin1(names[php.match(sample).hasMatch() ? Php : Html]);
GUI_DEBUG("code language: container=%s sampled=%d", qPrintable(language), int(sample.size()));
return language;
}
// Убираем строки и комментарии перед поиском конструкций, сохраняя границы строк.
static const QRegularExpression ignored(
R"re(/\*[\s\S]*?(?:\*/|$)|<!--[\s\S]*?(?:-->|$)|//[^\n]*|"""[\s\S]*?(?:"""|$))re"
R"re(|`(?:\\[\s\S]|[^`\\])*(?:`|$)|"(?:\\.|[^"\\\n])*(?:"|$)|'(?:\\.|[^'\\\n])*(?:'|$))re");
QString code = sample;
auto matches = ignored.globalMatch(sample);
while (matches.hasNext()) {
const auto match = matches.next();
for (int i = match.capturedStart(); i < match.capturedEnd(); ++i) if (code[i] != '\n') code[i] = ' ';
}
std::array<int, LanguageCount> scores{};
for (const auto& rule : rules) if (rule.pattern.match(code).hasMatch()) scores[rule.language] += rule.weight;
// Структурные теги отличают HTML с вложенным JS/CSS; одиночный div внутри JSX недостаточен.
static const QRegularExpression structuralTags(R"(</?(?:html|head|body|script|style)\b[^<>]*>)",
QRegularExpression::CaseInsensitiveOption);
static const QRegularExpression tags(R"(<(div|span|p|a|ul|li|table|section|form|button|h[1-6])\b[^<>]*>[\s\S]*?</\1\s*>)",
QRegularExpression::CaseInsensitiveOption);
if (structuralTags.match(code).hasMatch() || (scores[JavaScript] == 0 && tags.match(code).hasMatch())) {
GUI_DEBUG("code language: result=HTML markup tags sampled=%d", int(sample.size()));
return QStringLiteral("HTML");
}
// std:: и Type::method встречаются также в Rust; use/fn/impl отличают его от C++.
if (scores[Rust] >= 7) scores[Cpp] = 0;
// C является подмножеством C++: специфичные конструкции C++ имеют приоритет.
if (scores[Cpp] >= 6) scores[C] = 0;
int best = 0, second = 0;
for (int i = 1; i < LanguageCount; ++i) if (scores[i] > scores[best]) best = i;
for (int i = 0; i < LanguageCount; ++i) if (i != best && scores[i] > second) second = scores[i];
const bool accepted = scores[best] >= 6 && scores[best] - second >= 3;
GUI_DEBUG("code language: result=%s candidate=%s score=%d runner_up=%d sampled=%d total=%lld",
accepted ? names[best] : "unknown", names[best], scores[best], second, int(sample.size()), (long long)source.size());
return accepted ? QString::fromLatin1(names[best]) : QString();
}