path: root/subex/parse.go

package subex

import (
	"main/walk"
	"strconv"
	"strings"
)

type RuneReader interface {
	Next() rune
	Rewind()
}

func accept(l RuneReader, chars string) bool {
	r := l.Next()
	for _, char := range chars {
		if char == r {
			return true
		}
	}
	l.Rewind()
	return false
}

func isNumericRune(r rune) bool {
	return '0' <= r && r <= '9' || r == '.'
}

// Having just parsed a `, read until the next ` and parse the contents into a list of non-string atoms
func parseScalarLiteral(l RuneReader) (walk.Scalar, bool) {
	r := l.Next()
	if isNumericRune(r) {
		var builder strings.Builder
		builder.WriteRune(r)
		for {
			r := l.Next()
			if !isNumericRune(r) {
				l.Rewind()
				break
			}
			builder.WriteRune(r)
		}
		numberString := builder.String()
		number, err := strconv.ParseFloat(numberString, 64)
		if err != nil {
			panic("Invalid number literal")
		}
		return walk.NumberScalar(number), true
	}
	switch r {
		case 'n':
			if accept(l, "u") && accept(l, "l") && accept(l, "l") {
				return walk.NullScalar{}, true
			} else {
				panic("Invalid literal")
			}
		case 't':
			if accept(l, "r") && accept(l, "u") && accept(l, "e") {
				return walk.BoolScalar(true), true
			} else {
				panic("Invalid literal")
			}
		case 'f':
			if accept(l, "a") && accept(l, "l") && accept(l, "s") && accept(l, "e") {
				return walk.BoolScalar(false), true
			} else {
				panic("Invalid literal")
			}
		default:
			panic("Invalid literal")
	}
}

func charIsDigit(c rune) bool {
	return '0' <= c && c <= '9'
}

// Parse a positive integer, reads digits 0-9 and stops at the first non-digit
func parseInt(l RuneReader) (output int) {
	for {
		char := l.Next()
		if charIsDigit(char) {
			output = output * 10 + int(char - '0')
		} else {
			break
		}
	}
	l.Rewind()
	return output
}

// Having just read {, read in and parse the range contents
func parseRepeatRange(l RuneReader) (output []ConvexRange) {
	loop: for {
		var start, end int
		char := l.Next()
		l.Rewind()
		if char == '-' {
			start = -1
		} else {
			start = parseInt(l)
		}
		switch l.Next() {
			case ',':
				output = append(output, ConvexRange{start, start})
				continue loop
			case '-':
				char := l.Next()
				if charIsDigit(char) {
					l.Rewind()
					end = parseInt(l)
				} else {
					l.Rewind()
					end = -1
				}
			case '}':
				output = append(output, ConvexRange{start, start})
				break loop
			default:
				panic("Invalid character in repeat specifier")
		}
		switch l.Next() {
			case ',':
				output = append(output, ConvexRange{start, end})
				continue loop
			case '}':
				output = append(output, ConvexRange{start, end})
				break loop
			default:
				panic("Invalid character in repeat specifier")
		}
	}
	return output
}

// TODO: Consider if it's worth making better use of the go type system to enforce output being all runes or all values
func parseReplacement(l RuneReader, runic bool) (output []OutputContentAST) {
	// TODO escaping
	// TODO add arrays, maps and strings
	loop: for {
		r := l.Next()
		switch r {
			case eof:
				panic("Missing closing `")
			case '`':
				break loop
			case '$':
				slot := l.Next()
				if slot == eof {
					panic("Missing slot character")
				}
				output = append(output, OutputLoadAST{slot: slot})
			default:
				if runic {
					output = append(output, OutputRuneLiteralAST {walk.StringRuneAtom(r)})
				} else {
					l.Rewind()
					scalar, ok := parseScalarLiteral(l)
					if !ok {
						panic("Invalid scalar literal")
					}
					output = append(output, OutputValueLiteralAST {scalar})
				}
		}
	}
	return output
}

// Parse the contents of a range subex [] into a map
// func parseRangeSubex(l RuneReader) map[walk.AtomOLD]walk.AtomOLD {
// 	// TODO escaping
// 	parts := make(map[walk.AtomOLD]walk.AtomOLD)
// 	var froms []walk.AtomOLD
// 	var hasTo bool
// 	for {
// 		fromsStart := l.Next()
// 		if fromsStart == ']' {
// 			hasTo = false
// 			break
// 		} else if fromsStart == '=' {
// 			hasTo = true
// 			break
// 		} else if fromsStart == '`' {
// 			literals := parseNonStringLiteral(l)
// 			froms = append(froms, literals...)
// 			continue
// 		} else if fromsStart == '"' {
// 			froms = append(froms, walk.NewAtomStringTerminal())
// 			continue
// 		}
// 		if accept(l, "-") {
// 			fromsEnd := l.Next()
// 			if fromsEnd == ']' || fromsEnd == '=' {
// 				l.Rewind()
// 				fromsEnd = fromsStart
// 			}
// 			for i := fromsStart; i <= fromsEnd; i += 1 {
// 				froms = append(froms, walk.NewAtomStringRune(i))
// 			}
// 		} else {
// 			froms = append(froms, walk.NewAtomStringRune(fromsStart))
// 		}
// 	}
// 	if len(froms) == 0 {
// 		panic("Missing from part of range expression")
// 	}

// 	var tos []walk.AtomOLD
// 	if hasTo {
// 		for {
// 			tosStart := l.Next()
// 			if tosStart == ']' {
// 				break
// 			} else if tosStart == '`' {
// 				literals := parseNonStringLiteral(l)
// 				tos = append(tos, literals...)
// 				continue
// 			} else if tosStart == '"' {
// 				tos = append(tos, walk.NewAtomStringTerminal())
// 				continue
// 			}
// 			if accept(l, "-") {
// 				tosEnd := l.Next()
// 				if tosEnd == ']' {
// 					l.Rewind()
// 					tosEnd = tosStart
// 				}
// 				for i := tosStart; i <= tosEnd; i += 1 {
// 					tos = append(tos, walk.NewAtomStringRune(i))
// 				}
// 			} else {
// 				tos = append(tos, walk.NewAtomStringRune(tosStart))
// 			}
// 		}
// 	} else {
// 		tos = froms
// 	}
// 	if len(tos) == 0 {
// 		panic("Missing to part of range expression")
// 	}
	
// 	for i, from := range froms {
// 		parts[from] = tos[i % len(tos)]
// 	}
// 	return parts
// }

func parseSubex(l RuneReader, minPower int, runic bool) SubexAST {
	var lhs SubexAST
	r := l.Next()
	switch r {
		case eof:
			return nil
		case '(':
			lhs = parseSubex(l, 0, runic)
			if !accept(l, ")") {
				panic("Missing matching )")
			}
		// TODO
		// case '[':
		// 	rangeParts := parseRangeSubex(l)
		// 	lhs = SubexASTRange {rangeParts}
		case ')', ']', '"', '|', ';', '{', '+', '-', '*', '/', '!', '=', '$':
			l.Rewind()
			return SubexASTEmpty{}
		// case '=':
		// 	replacement := parseReplacement(l)
		// 	lhs = SubexASTOutput{replacement}
		// case '^':
		// 	replacement := parseReplacement(l)
		// 	replacement = append(
		// 		[]OutputContentAST{OutputValueLiteralAST {walk.NewAtomStringTerminal()}},
		// 		replacement...
		// 	)
		// 	replacement = append(
		// 		replacement,
		// 		OutputValueLiteralAST {walk.NewAtomStringTerminal()},
		// 	)
		// 	lhs = SubexASTOutput {replacement}
		case '.':
			if runic {
				lhs = SubexASTCopyAnyRune{}
			} else {
				lhs = SubexASTCopyAnyValue{}
			}
		case '?':
			lhs = SubexASTCopyBool{}
		case '%':
			lhs = SubexASTCopyNumber{}
		case ':':
			if runic {
				lhs = SubexASTCopyRune {':'}
			} else {
				if !accept(l, "[") {
					panic("Missing [ after :")
				}
				lhs = SubexASTEnterArray {parseSubex(l, 0, runic)}
				if !accept(l, "]") {
					panic("Missing matching ]")
				}
			}
		case '`':
			lhs = SubexASTOutput {parseReplacement(l, runic)}
		case '~':
			if runic {
				lhs = SubexASTCopyRune {'~'}
			} else {
				if !accept(l, "\"") {
					panic("Missing \" after ~")
				}
				lhs = SubexASTEnterString {parseSubex(l, 0, true)}
				if !accept(l, "\"") {
					panic("Missing matching \"")
				}
			}
		// TODO
		// case '_':
		// 	lhs = SubexASTCopyStringAtom{}
		// case '#':
		// 	lhs = SubexASTCopyString{}
		// case ',':
		// 	lhs = SubexASTCopyValue{}
		// case '"':
		// 	lhs = SubexASTCopyScalar {walk.NewAtomStringTerminal()}
		// case '~':
		// 	literals := parseNonStringLiteral(l)
		// 	var replacement []OutputContentAST
		// 	for _, literal := range literals {
		// 		replacement = append(replacement, OutputValueLiteralAST {literal})
		// 	}
		// 	lhs = SubexASTOutput {replacement}
		default:
			if runic {
				lhs = SubexASTCopyRune {r}
			} else {
				l.Rewind()
				scalar, ok := parseScalarLiteral(l)
				if !ok {
					panic("Invalid subex")
				}
				lhs = SubexASTCopyScalar {scalar}
			}
	}
	loop: for {
		if minPower <= 20 {
			next := parseSubex(l, 21, runic)
			if next != nil && (next != SubexASTEmpty{}) {
				lhs = SubexASTConcat{lhs, next}
				continue loop
			}
		}
		r := l.Next()
		switch {
			case r == '{' && minPower <= 4:
				lhs = SubexASTRepeat {
					Content: lhs,
					Acceptable: parseRepeatRange(l),
				}
			case r == '+' && minPower <= 4:
				lhs = SubexASTSum {lhs}
			case r == '*' && minPower <= 4:
				lhs = SubexASTProduct {lhs}
			case r == '-' && minPower <= 4:
				lhs = SubexASTNegate {lhs}
			case r == '/' && minPower <= 4:
				lhs = SubexASTReciprocal {lhs}
			case r == '!' && minPower <= 4:
				lhs = SubexASTNot {lhs}
			case r == '=' && minPower <= 4:
				lhs = SubexASTEqual {lhs}
			case r == '$' && minPower <= 4:
				slot := l.Next()
				if slot == eof {
					panic("Missing slot character")
				}
				if slot == '_' {
					lhs = SubexASTDiscard {lhs}
				} else {
					lhs = SubexASTStore{
						Match: lhs,
						Slot: slot,
					}
				}
			case r == '|' && minPower <= 8:
				rhs := parseSubex(l, 9, runic)
				if rhs == nil {
					panic("Missing subex after |")
				}
				lhs = SubexASTOr{lhs, rhs}
			case r == ';' && minPower <= 10:
				rhs := parseSubex(l, 11, runic)
				if rhs == nil {
					panic("Missing subex after ;")
				}
				lhs = SubexASTJoin{
					Content: lhs,
					Delimiter: rhs,
				}
			default:
				l.Rewind()
				break loop
		}
	}
	return lhs
}

func Parse(l RuneReader) SubexAST {
	ast := parseSubex(l, 0, false)
	if ast == nil {
		return SubexASTEmpty{}
	}
	return ast
}