code-gen: make content heuristics regexp engine configurable & generation syntax-aware

internal/code-generator/assets/content.go.tmpl

@@ -1,9 +1,8 @@
 package data
 
 import (
-	"regexp"
-
 	"github.com/go-enry/go-enry/v2/data/rule"
+	"github.com/go-enry/go-enry/v2/regex"
 )
 
 var ContentHeuristics = map[string]*Heuristics{
@@ -27,12 +26,12 @@ var ContentHeuristics = map[string]*Heuristics{
 	{{- else if eq .Op "Or" -}}
 		rule.Or(
 			{{ template "Languages" .Langs -}}
-			regexp.MustCompile({{ .Pattern | stringVal }}),
+			{{ template "mustCompile" . }}
 		),
 	{{- else if eq .Op "Not" -}}
 		rule.Not(
 			{{ template "Languages" .Langs -}}
-			regexp.MustCompile({{ .Pattern | stringVal }}),
+			{{ template "mustCompile" . }}
 		),
 	{{- else if eq .Op "Always" -}}
 		rule.Always(
@@ -49,3 +48,11 @@ var ContentHeuristics = map[string]*Heuristics{
 		rule.MatchingLanguages(""),
 	{{end -}}
 {{end}}
+
+{{define "mustCompile" -}}
+	{{ if .IsRE2  -}}
+		regex.MustCompileMultiline({{ .Pattern | stringVal }}),
+	{{- else -}}
+		regex.MustCompileRuby({{ .Pattern | stringVal }}),
+	{{ end -}}
+{{end}}

internal/code-generator/generator/heuristics.go

@@ -70,25 +70,24 @@ func loadRule(namedPatterns map[string]StringArray, rule *Rule) *LanguagePattern
 			subp := loadRule(namedPatterns, r)
 			subPatterns = append(subPatterns, subp)
 		}
-		result = &LanguagePattern{"And", rule.Languages, "", subPatterns}
+		result = &LanguagePattern{"And", rule.Languages, "", subPatterns, true}
 	} else if len(rule.Pattern) != 0 { // OrPattern
-		conjunction := strings.Join(rule.Pattern, orPipe)
-		pattern := convertToValidRegexp(conjunction)
-		result = &LanguagePattern{"Or", rule.Languages, pattern, nil}
+		pattern := strings.Join(rule.Pattern, orPipe)
+		// TODO(bzz): handle len(Languages)==0 better e.g. by emiting rule.Rule
+		// instead of an ugly `rule.Or( rule.MatchingLanguages(""), ... )`
+		result = &LanguagePattern{"Or", rule.Languages, pattern, nil, isRE2(pattern)}
 	} else if rule.NegativePattern != "" { // NotPattern
-		pattern := convertToValidRegexp(rule.NegativePattern)
-		result = &LanguagePattern{"Not", rule.Languages, pattern, nil}
+		pattern := rule.NegativePattern
+		result = &LanguagePattern{"Not", rule.Languages, pattern, nil, isRE2(pattern)}
 	} else if rule.NamedPattern != "" { // Named OrPattern
-		conjunction := strings.Join(namedPatterns[rule.NamedPattern], orPipe)
-		pattern := convertToValidRegexp(conjunction)
-		result = &LanguagePattern{"Or", rule.Languages, pattern, nil}
+		pattern := strings.Join(namedPatterns[rule.NamedPattern], orPipe)
+		result = &LanguagePattern{"Or", rule.Languages, pattern, nil, isRE2(pattern)}
 	} else { // AlwaysPattern
-		result = &LanguagePattern{"Always", rule.Languages, "", nil}
+		result = &LanguagePattern{"Always", rule.Languages, "", nil, true}
 	}
 
-	if isUnsupportedRegexpSyntax(result.Pattern) {
-		log.Printf("skipping rule: language:'%q', rule:'%q'\n", rule.Languages, result.Pattern)
-		return nil
+	if !isRE2(result.Pattern) {
+		log.Printf("RE2 incompatible rule: language:'%q', rule:'%q'\n", rule.Languages, result.Pattern)
 	}
 	return result
 }
@@ -100,6 +99,7 @@ type LanguagePattern struct {
 	Langs   []string
 	Pattern string
 	Rules   []*LanguagePattern
+	IsRE2   bool
 }
 
 type Heuristics struct {
@@ -125,7 +125,7 @@ type Patterns struct {
 }
 
 // StringArray is workaround for parsing named_pattern,
-// wich is sometimes arry and sometimes not.
+// wich is sometimes an array and sometimes is not.
 // See https://github.com/go-yaml/yaml/issues/100
 type StringArray []string
 
@@ -173,8 +173,6 @@ func isUnsupportedRegexpSyntax(reg string) bool {
 		(strings.HasPrefix(reg, multilinePrefix+`/`) && strings.HasSuffix(reg, `/`))
 }
 
-// convertToValidRegexp converts Ruby regexp syntax to RE2 equivalent.
-// Does not work with Ruby regexp literals.
-func convertToValidRegexp(rubyRegexp string) string {
-	return multilinePrefix + rubyRegexp
+func isRE2(s string) bool {
+	return !isUnsupportedRegexpSyntax(s)
 }

regex/oniguruma.go

@@ -1,3 +1,4 @@
+//go:build oniguruma
 // +build oniguruma
 
 package regex
@@ -8,8 +9,17 @@ import (
 
 type EnryRegexp = *rubex.Regexp
 
-func MustCompile(str string) EnryRegexp {
-	return rubex.MustCompileASCII(str)
+func MustCompile(s string) EnryRegexp {
+	return rubex.MustCompileASCII(s)
+}
+
+// MustCompileMultiline matches in multi-line mode by default with Oniguruma.
+func MustCompileMultiline(s string) EnryRegexp {
+	return MustCompile(s)
+}
+
+func MustCompileRuby(s string) EnryRegexp {
+	return MustCompile(s)
 }
 
 func QuoteMeta(s string) string {

regex/standard.go

@@ -1,3 +1,4 @@
+//go:build !oniguruma
 // +build !oniguruma
 
 package regex
@@ -12,6 +13,20 @@ func MustCompile(str string) EnryRegexp {
 	return regexp.MustCompile(str)
 }
 
+// MustCompileMultiline mimics Ruby defaults for regexp, where ^$ matches begin/end of line.
+// I.e. it converts Ruby regexp syntaxt to RE2 equivalent
+func MustCompileMultiline(s string) EnryRegexp {
+	const multilineModeFlag = "(?m)"
+	return regexp.MustCompile(multilineModeFlag + s)
+}
+
+// MustCompileRuby used for expressions with syntax not supported by RE2.
+func MustCompileRuby(s string) EnryRegexp {
+	// TODO(bzz): find a bettee way?
+	// This will only trigger a panic on .Match() for the clients
+	return nil
+}
+
 func QuoteMeta(s string) string {
 	return regexp.QuoteMeta(s)
 }