| package parser |
| |
| import ( |
| "fmt" |
| "strings" |
| |
| "github.com/KorAP/Koral-Mapper/ast" |
| "github.com/alecthomas/participle/v2" |
| "github.com/alecthomas/participle/v2/lexer" |
| ) |
| |
| // GrammarParser parses a simple grammar into AST nodes |
| type GrammarParser struct { |
| defaultFoundry string |
| defaultLayer string |
| tokenParser *participle.Parser[TokenGrammar] |
| mappingParser *participle.Parser[MappingGrammar] |
| } |
| |
| // TokenGrammar represents a single token expression |
| type TokenGrammar struct { |
| Token *TokenExpr `parser:"@@"` |
| } |
| |
| // MappingGrammar represents a mapping rule |
| type MappingGrammar struct { |
| Mapping *MappingRule `parser:"@@"` |
| } |
| |
| // MappingRule represents a mapping between two token expressions |
| type MappingRule struct { |
| Upper *TokenExpr `parser:"@@"` |
| Lower *TokenExpr `parser:"'<>' @@"` |
| } |
| |
| // TokenExpr represents a token expression in square brackets |
| type TokenExpr struct { |
| Expr *Expr `parser:"'[' @@ ']'"` |
| } |
| |
| // Expr represents a sequence of terms and operators |
| type Expr struct { |
| First *Term `parser:"@@"` |
| Rest []*Op `parser:"@@*"` |
| } |
| |
| type Op struct { |
| Operator string `parser:"@('&' | '|')"` |
| Term *Term `parser:"@@"` |
| } |
| |
| // Term represents either a simple term or a parenthesized expression |
| type Term struct { |
| Simple *SimpleTerm `parser:"@@"` |
| Paren *ParenExpr `parser:"| @@"` |
| } |
| |
| type ParenExpr struct { |
| Expr *Expr `parser:"'(' @@ ')'"` |
| } |
| |
| // SimpleTerm represents any valid term form |
| type SimpleTerm struct { |
| WithFoundryLayer *FoundryLayerTerm `parser:"@@"` |
| WithFoundryWildcard *FoundryWildcardTerm `parser:"| @@"` |
| WithFoundryKey *FoundryKeyTerm `parser:"| @@"` |
| WithLayer *LayerTerm `parser:"| @@"` |
| SimpleKey *KeyTerm `parser:"| @@"` |
| } |
| |
| // FoundryLayerTerm represents foundry/layer=key:value |
| type FoundryLayerTerm struct { |
| Foundry string `parser:"@Ident '/'"` |
| Layer string `parser:"@Ident '='"` |
| Key string `parser:"@Ident"` |
| Value string `parser:"(':' @Ident)?"` |
| } |
| |
| // FoundryWildcardTerm represents foundry/*=key (wildcard layer) |
| type FoundryWildcardTerm struct { |
| Foundry string `parser:"@Ident '/' '*' '='"` |
| Key string `parser:"@Ident"` |
| } |
| |
| // FoundryKeyTerm represents foundry/key |
| type FoundryKeyTerm struct { |
| Foundry string `parser:"@Ident '/'"` |
| Key string `parser:"@Ident"` |
| } |
| |
| // LayerTerm represents layer=key:value (only when no foundry is present) |
| type LayerTerm struct { |
| Layer string `parser:"@Ident '='"` |
| Key string `parser:"@Ident"` |
| Value string `parser:"(':' @Ident)?"` |
| } |
| |
| // KeyTerm represents key:value or key=value |
| type KeyTerm struct { |
| Key string `parser:"@Ident"` |
| Value string `parser:"((':' | '=') @Ident)?"` |
| } |
| |
| // EscapedPunct represents an escaped punctuation character like \( |
| type EscapedPunct struct { |
| Prefix string `parser:"@Ident"` |
| Punct string `parser:"@Punct"` |
| } |
| |
| // NewGrammarParser creates a new grammar parser with optional default foundry and layer |
| func NewGrammarParser(defaultFoundry, defaultLayer string) (*GrammarParser, error) { |
| lex := lexer.MustSimple([]lexer.SimpleRule{ |
| {Name: "Ident", Pattern: `(?:[a-zA-Z$,.]|\\.)(?:[a-zA-Z0-9_$,.]|\\.)*`}, |
| {Name: "Punct", Pattern: `[\[\]()&\|=:/\*]|<>`}, |
| {Name: "Whitespace", Pattern: `\s+`}, |
| }) |
| |
| tokenParser, err := participle.Build[TokenGrammar]( |
| participle.Lexer(lex), |
| participle.UseLookahead(2), |
| participle.Elide("Whitespace"), |
| ) |
| if err != nil { |
| return nil, fmt.Errorf("failed to build token parser: %w", err) |
| } |
| |
| mappingParser, err := participle.Build[MappingGrammar]( |
| participle.Lexer(lex), |
| participle.UseLookahead(2), |
| participle.Elide("Whitespace"), |
| ) |
| if err != nil { |
| return nil, fmt.Errorf("failed to build mapping parser: %w", err) |
| } |
| |
| return &GrammarParser{ |
| defaultFoundry: defaultFoundry, |
| defaultLayer: defaultLayer, |
| tokenParser: tokenParser, |
| mappingParser: mappingParser, |
| }, nil |
| } |
| |
| // preprocessInput normalizes the input string by handling operators and parentheses |
| func (p *GrammarParser) preprocessInput(input string) string { |
| // Remove extra spaces around operators to help the parser |
| input = strings.ReplaceAll(input, " & ", "&") |
| input = strings.ReplaceAll(input, " | ", "|") |
| input = strings.ReplaceAll(input, " <> ", "<>") |
| |
| // Add spaces around parentheses that are not escaped |
| result := make([]rune, 0, len(input)*2) |
| runes := []rune(input) |
| for i, r := range runes { |
| if (r == '(' || r == ')') && (i == 0 || runes[i-1] != '\\') { |
| // Check if this parenthesis is inside brackets (part of an identifier) |
| insideBrackets := false |
| bracketDepth := 0 |
| for j := 0; j < i; j++ { |
| if runes[j] == '[' { |
| bracketDepth++ |
| } else if runes[j] == ']' { |
| bracketDepth-- |
| } |
| } |
| insideBrackets = bracketDepth > 0 |
| |
| if !insideBrackets { |
| result = append(result, ' ', r, ' ') |
| } else { |
| result = append(result, r) |
| } |
| } else { |
| result = append(result, r) |
| } |
| } |
| return strings.TrimSpace(string(result)) |
| } |
| |
| // ParseMapping parses a mapping rule string into a MappingResult |
| func (p *GrammarParser) ParseMapping(input string) (*MappingResult, error) { |
| input = p.preprocessInput(input) |
| |
| grammar, err := p.mappingParser.ParseString("", input) |
| if err != nil { |
| return nil, fmt.Errorf("failed to parse grammar: %w", err) |
| } |
| |
| if grammar.Mapping == nil { |
| return nil, fmt.Errorf("expected mapping rule, got token expression") |
| } |
| |
| upper, err := p.parseExpr(grammar.Mapping.Upper.Expr) |
| if err != nil { |
| return nil, err |
| } |
| |
| lower, err := p.parseExpr(grammar.Mapping.Lower.Expr) |
| if err != nil { |
| return nil, err |
| } |
| |
| return &MappingResult{ |
| Upper: &ast.Token{Wrap: upper}, |
| Lower: &ast.Token{Wrap: lower}, |
| }, nil |
| } |
| |
| // MappingResult represents the parsed mapping rule |
| type MappingResult struct { |
| Upper *ast.Token |
| Lower *ast.Token |
| } |
| |
| // parseExpr builds the AST from the parsed Expr |
| func (p *GrammarParser) parseExpr(expr *Expr) (ast.Node, error) { |
| var operands []ast.Node |
| var operators []string |
| |
| // Parse the first term |
| first, err := p.parseTerm(expr.First) |
| if err != nil { |
| return nil, err |
| } |
| operands = append(operands, first) |
| |
| // Parse the rest |
| for _, op := range expr.Rest { |
| node, err := p.parseTerm(op.Term) |
| if err != nil { |
| return nil, err |
| } |
| operands = append(operands, node) |
| operators = append(operators, op.Operator) |
| } |
| |
| // If only one operand, return it |
| if len(operands) == 1 { |
| return operands[0], nil |
| } |
| |
| // Group operands by operator precedence (left-to-right, no precedence between & and |) |
| // We'll group by runs of the same operator |
| var groupOperands []ast.Node |
| var currentOp string |
| var currentGroup []ast.Node |
| for i, op := range operators { |
| if i == 0 { |
| currentOp = op |
| currentGroup = append(currentGroup, operands[i]) |
| } |
| if op == currentOp { |
| currentGroup = append(currentGroup, operands[i+1]) |
| } else { |
| groupOperands = append(groupOperands, &ast.TermGroup{ |
| Operands: append([]ast.Node{}, currentGroup...), |
| Relation: toRelation(currentOp), |
| }) |
| currentOp = op |
| currentGroup = []ast.Node{operands[i+1]} |
| } |
| } |
| if len(currentGroup) > 0 { |
| groupOperands = append(groupOperands, &ast.TermGroup{ |
| Operands: append([]ast.Node{}, currentGroup...), |
| Relation: toRelation(currentOp), |
| }) |
| } |
| if len(groupOperands) == 1 { |
| return groupOperands[0], nil |
| } |
| // If mixed operators, nest them left-to-right |
| result := groupOperands[0] |
| for i := 1; i < len(groupOperands); i++ { |
| result = &ast.TermGroup{ |
| Operands: []ast.Node{result, groupOperands[i]}, |
| Relation: toRelation(operators[0]), |
| } |
| } |
| return result, nil |
| } |
| |
| // parseTerm converts a Term into an AST node |
| func (p *GrammarParser) parseTerm(term *Term) (ast.Node, error) { |
| if term.Simple != nil { |
| return p.parseSimpleTerm(term.Simple) |
| } |
| if term.Paren != nil { |
| return p.parseExpr(term.Paren.Expr) |
| } |
| return nil, fmt.Errorf("invalid term: neither simple nor parenthesized") |
| } |
| |
| func toRelation(op string) ast.RelationType { |
| if op == "|" { |
| return ast.OrRelation |
| } |
| return ast.AndRelation |
| } |
| |
| // unescapeString handles unescaping of backslash-escaped characters |
| func unescapeString(s string) string { |
| if s == "" { |
| return s |
| } |
| |
| // Modify string in-place by overwriting characters |
| bytes := []byte(s) |
| j := 0 |
| for i := 0; i < len(bytes); i++ { |
| if bytes[i] == '\\' && i+1 < len(bytes) { |
| // Skip backslash and copy next char |
| bytes[j] = bytes[i+1] |
| i++ |
| } else { |
| // Copy current char |
| bytes[j] = bytes[i] |
| } |
| j++ |
| } |
| return string(bytes[:j]) |
| } |
| |
| // parseSimpleTerm converts a SimpleTerm into an AST Term node |
| func (p *GrammarParser) parseSimpleTerm(term *SimpleTerm) (ast.Node, error) { |
| var foundry, layer, key, value string |
| |
| switch { |
| case term.WithFoundryLayer != nil: |
| foundry = unescapeString(term.WithFoundryLayer.Foundry) |
| layer = unescapeString(term.WithFoundryLayer.Layer) |
| key = unescapeString(term.WithFoundryLayer.Key) |
| value = unescapeString(term.WithFoundryLayer.Value) |
| case term.WithFoundryWildcard != nil: |
| foundry = unescapeString(term.WithFoundryWildcard.Foundry) |
| key = unescapeString(term.WithFoundryWildcard.Key) |
| case term.WithFoundryKey != nil: |
| foundry = unescapeString(term.WithFoundryKey.Foundry) |
| key = unescapeString(term.WithFoundryKey.Key) |
| case term.WithLayer != nil: |
| // Special case: if LayerTerm was parsed but the layer doesn't match the default layer, |
| // treat it as a key=value pattern instead |
| parsedLayer := unescapeString(term.WithLayer.Layer) |
| parsedKey := unescapeString(term.WithLayer.Key) |
| parsedValue := unescapeString(term.WithLayer.Value) |
| |
| if p.defaultLayer != "" && parsedLayer == p.defaultLayer { |
| // This is a genuine layer=key pattern when the layer matches the default |
| layer = parsedLayer |
| key = parsedKey |
| value = parsedValue |
| } else if p.defaultLayer != "" && parsedLayer != p.defaultLayer { |
| // This should be treated as key=value pattern when there's a default layer but it doesn't match |
| key = parsedLayer |
| value = parsedKey |
| } else { |
| // No default layer context, treat as genuine layer=key pattern |
| layer = parsedLayer |
| key = parsedKey |
| value = parsedValue |
| } |
| case term.SimpleKey != nil: |
| key = unescapeString(term.SimpleKey.Key) |
| value = unescapeString(term.SimpleKey.Value) |
| default: |
| return nil, fmt.Errorf("invalid term: no valid form found") |
| } |
| |
| if foundry == "" { |
| foundry = p.defaultFoundry |
| } |
| if layer == "" { |
| layer = p.defaultLayer |
| } |
| |
| return &ast.Term{ |
| Foundry: foundry, |
| Key: key, |
| Layer: layer, |
| Match: ast.MatchEqual, |
| Value: value, |
| }, nil |
| } |