package generator import ( "bytes" "fmt" "io/ioutil" "log" "strings" yaml "gopkg.in/yaml.v2" ) const ( multilinePrefix = "(?m)" orPipe = "|" ) // GenHeuristics generates language identification heuristics in Go. // It is of generator.File type. func GenHeuristics(fileToParse, _, outPath, tmplPath, tmplName, commit string) error { heuristicsYaml, err := parseYaml(fileToParse) if err != nil { return err } langPatterns, err := loadHeuristics(heuristicsYaml) if err != nil { return err } buf := &bytes.Buffer{} err = executeTemplate(buf, tmplName, tmplPath, commit, nil, langPatterns) if err != nil { return err } return formatedWrite(outPath, buf.Bytes()) } // loadHeuristics transforms parsed YAML to map[".ext"]->IR for code generation. func loadHeuristics(yaml *Heuristics) (map[string][]*LanguagePattern, error) { var patterns = make(map[string][]*LanguagePattern) for _, disambiguation := range yaml.Disambiguations { var rules []*LanguagePattern for _, rule := range disambiguation.Rules { langPattern := loadRule(yaml.NamedPatterns, rule) if langPattern != nil { rules = append(rules, langPattern) } } // unroll to a single map for _, ext := range disambiguation.Extensions { if _, ok := patterns[ext]; ok { return nil, fmt.Errorf("cannot add extension '%s', it already exists for %+v", ext, patterns[ext]) } patterns[ext] = rules } } return patterns, nil } // loadRule transforms single rule from parsed YAML to IR for code generation. // For OrPattern case, it always combines multiple patterns into a single one. func loadRule(namedPatterns map[string]StringArray, rule *Rule) *LanguagePattern { var result *LanguagePattern if len(rule.And) != 0 { // AndPattern var subPatterns []*LanguagePattern for _, r := range rule.And { subp := loadRule(namedPatterns, r) subPatterns = append(subPatterns, subp) } result = &LanguagePattern{"And", rule.Languages, "", subPatterns} } else if len(rule.Pattern) != 0 { // OrPattern conjunction := strings.Join(rule.Pattern, orPipe) pattern := convertToValidRegexp(conjunction) result = &LanguagePattern{"Or", rule.Languages, pattern, nil} } else if rule.NegativePattern != "" { // NotPattern pattern := convertToValidRegexp(rule.NegativePattern) result = &LanguagePattern{"Not", rule.Languages, pattern, nil} } else if rule.NamedPattern != "" { // Named OrPattern conjunction := strings.Join(namedPatterns[rule.NamedPattern], orPipe) pattern := convertToValidRegexp(conjunction) result = &LanguagePattern{"Or", rule.Languages, pattern, nil} } else { // AlwaysPattern result = &LanguagePattern{"Always", rule.Languages, "", nil} } if isUnsupportedRegexpSyntax(result.Pattern) { log.Printf("skipping rule: language:'%q', rule:'%q'\n", rule.Languages, result.Pattern) return nil } return result } // LanguagePattern is an IR of parsed Rule suitable for code generations. // Strings are used as this is to be be consumed by text/template. type LanguagePattern struct { Op string Langs []string Pattern string Rules []*LanguagePattern } type Heuristics struct { Disambiguations []*Disambiguation NamedPatterns map[string]StringArray `yaml:"named_patterns"` } type Disambiguation struct { Extensions []string `yaml:"extensions,flow"` Rules []*Rule `yaml:"rules"` } type Rule struct { Patterns `yaml:",inline"` Languages StringArray `yaml:"language"` And []*Rule } type Patterns struct { Pattern StringArray `yaml:"pattern,omitempty"` NamedPattern string `yaml:"named_pattern,omitempty"` NegativePattern string `yaml:"negative_pattern,omitempty"` } // StringArray is workaround for parsing named_pattern, // wich is sometimes arry and sometimes not. // See https://github.com/go-yaml/yaml/issues/100 type StringArray []string // UnmarshalYAML allowes to parse element always as a []string func (sa *StringArray) UnmarshalYAML(unmarshal func(interface{}) error) error { var multi []string if err := unmarshal(&multi); err != nil { var single string if err := unmarshal(&single); err != nil { return err } *sa = []string{single} } else { *sa = multi } return nil } func parseYaml(file string) (*Heuristics, error) { data, err := ioutil.ReadFile(file) if err != nil { return nil, err } h := &Heuristics{} if err := yaml.Unmarshal(data, &h); err != nil { return nil, err } return h, nil } // isUnsupportedRegexpSyntax filters regexp syntax that is not supported by RE2. // In particular, we stumbled up on usage of next cases: // - named & numbered capturing group/after text matching // - backreference // For referece on supported syntax see https://github.com/google/re2/wiki/Syntax func isUnsupportedRegexpSyntax(reg string) bool { return strings.Contains(reg, `(?<`) || strings.Contains(reg, `\1`) || // See https://github.com/github/linguist/pull/4243#discussion_r246105067 (strings.HasPrefix(reg, multilinePrefix+`/`) && strings.HasSuffix(reg, `/`)) } // convertToValidRegexp converts Ruby regexp syntaxt to RE2 equivalent. // Does not work with Ruby regexp literals. func convertToValidRegexp(rubyRegexp string) string { return multilinePrefix + rubyRegexp }