regex/main.go

package main

import (
	"bufio"
	"fmt"
	"os"
	"slices"

	"github.com/fatih/color"
)

const CONCAT rune = '~'

func isOperator(c rune) bool {
	if c == '+' || c == '?' || c == '*' || c == '|' || c == CONCAT {
		return true
	}
	return false
}

/* priority returns the priority of the given operator */
func priority(op rune) int {
	precedence := []rune{'|', CONCAT, '+', '*', '?'}
	return slices.Index(precedence, op)
}

/*
The Shunting-Yard algorithm is used to convert the given infix (regeular) expression to postfix.
The primary benefit of this is getting rid of parentheses.
It also inserts explicit concatenation operators to make parsing easier in Thompson's algorithm.
See: https://blog.cernera.me/converting-regular-expressions-to-postfix-notation-with-the-shunting-yard-algorithm/
*/
func shuntingYard(re string) []postfixNode {
	re_postfix := make([]rune, 0)
	re_runes := []rune(re) // Convert the string to a slice of runes to allow iteration through it
	/* 	Add concatenation operators.
	Only add a concatenation operator between two characters if both the following conditions are met:
		1. 	The first character isn't an opening parantheses or alteration operator (or an escape character)
			a. This makes sense, because these operators can't be _concatenated_ with anything else.
		2. The second character isn't a 'closing operator' - one that applies to something before it
			a. Again, these operators can'be concatenated _to_. They can, however, be concatenated _from_.
	*/
	i := 0
	for i < len(re_runes) {
		re_postfix = append(re_postfix, re_runes[i])
		if re_runes[i] == '[' && (i == 0 || re_runes[i-1] != '\\') { // We do not touch things inside brackets, unless they are escaped
			for re_runes[i] != ']' {
				i++ // Skip all characters inside brackets
				// TODO: Check for escaped characters

				// Check ahead for character range
				if i < len(re_runes)-2 && re_runes[i+1] == '-' {
					rangeStart := re_runes[i]
					rangeEnd := re_runes[i+2]
					if int(rangeEnd) < int(rangeStart) {
						panic("Range is out of order.")
					}

					for i := rangeStart; i <= rangeEnd; i++ {
						re_postfix = append(re_postfix, i)
					}

					i += 2 // Skip start and hyphen (end will automatically be skipped on next iteration of loop)
					continue
				}
				re_postfix = append(re_postfix, re_runes[i])
			}
			continue
		}
		if re_runes[i] != '(' && re_runes[i] != '|' && re_runes[i] != '\\' {
			if i < len(re_runes)-1 {
				if re_runes[i+1] != '|' && re_runes[i+1] != '*' && re_runes[i+1] != '+' && re_runes[i+1] != '?' && re_runes[i+1] != ')' {
					re_postfix = append(re_postfix, CONCAT)
				}
			}
		}
		i++
	}

	opStack := make([]rune, 0)         // Operator stack
	outQueue := make([]postfixNode, 0) // Output queue

	// Actual algorithm
	for i := 0; i < len(re_postfix); i++ {
		/* Two cases:
		1. Current character is alphanumeric - send to output queue
		2. Current character is operator - do the following:
			a. If current character has greater priority than top of opStack, push to opStack.
			b. If not, keep popping from opStack (and appending to outQueue) until:
				i. opStack is empty, OR
				ii. current character has greater priority than top of opStack
		3. If current character is '(', push to opStack
		4. If current character is ')', pop from opStack (and append to outQueue) until '(' is found. Discard parantheses.
		5. If current character is '[', find all the characters until ']', then create a postfixNode containing all these contents. Add this node to outQueue.
		*/
		c := re_postfix[i]
		if isAlphaNum(c) {
			outQueue = append(outQueue, newPostfixNode(c))
			continue
		}
		// Escape character
		if c == '\\' { // Escape character - invert special and non-special characters eg. \( is treated as a literal parentheses, \b is treated as word boundary
			if i == len(re_postfix)-1 { // End of string - panic, because backslash is an escape character (something needs to come after it)
				panic("ERROR: Backslash with no escape character.")
			}
			i++
			switch re_postfix[i] {
			case 's': // Whitespace
				outQueue = append(outQueue, newPostfixNode(whitespaceChars...))
			case 'S': // Non-whitespace - I am doing this in a fancy way, generating all dot characters, then removing whitespace characters from it
				outQueue = append(outQueue, newPostfixNode(slices.DeleteFunc(dotChars(), func(r rune) bool {
					return slices.Contains(whitespaceChars, r)
				})...))
			case 'd': // Digits
				outQueue = append(outQueue, newPostfixNode(digitChars...))
			case 'D': // Non-digits - same fancy way as 'S'
				outQueue = append(outQueue, newPostfixNode(slices.DeleteFunc(dotChars(), func(r rune) bool {
					return slices.Contains(digitChars, r)
				})...))
			case 'w': // word character
				outQueue = append(outQueue, newPostfixNode(wordChars...))
			case 'W': // Non-word character - same fancy way as 'S' and 'D'
				outQueue = append(outQueue, newPostfixNode(slices.DeleteFunc(dotChars(), func(r rune) bool {
					return slices.Contains(wordChars, r)
				})...))
			default: // None of the above - append it as a regular character
				outQueue = append(outQueue, newPostfixCharNode(re_postfix[i]))
			}
			continue // Escaped character will automatically be skipped when loop variable increments
		}

		if c == '.' { // Dot metacharacter - represents 'any' character, but I am only adding Unicode 0020-007E
			outQueue = append(outQueue, newPostfixNode(dotChars()...))
			continue
		}
		if isOperator(c) {
			if len(opStack) == 0 {
				opStack = append(opStack, c)
			} else {
				topStack, err := peek(opStack)
				if err != nil {
					panic("ERROR: Operator without operand.")
				}
				if priority(c) > priority(topStack) { // 2a
					opStack = append(opStack, c)
				} else {
					for priority(c) <= priority(topStack) { // 2b
						to_append := mustPop(&opStack)
						outQueue = append(outQueue, newPostfixNode(to_append))
						topStack, _ = peek(opStack)
					}
					opStack = append(opStack, c)
				}
			}
		}
		if c == '[' { // Used for character classes
			i++                      // Step forward so we can look at the character class
			chars := make([]rune, 0) // List of characters -  used only for character classes
			for i < len(re_postfix) {
				if re_postfix[i] == ']' {
					break
				}
				chars = append(chars, re_postfix[i])
				i++
			}
			if i == len(re_postfix) { // We have reached the end of the string, so we didn't encounter a closing brakcet. Panic.
				panic("ERROR: Opening bracket without closing bracket.")
			}
			outQueue = append(outQueue, newPostfixNode(chars...))
			i++ // Step forward to skip closing bracket
			continue
		}
		if c == '(' {
			opStack = append(opStack, c)
		}
		if c == ')' {
			// Keep popping from opStack until we encounter an opening parantheses. Panic if we reach the end of the stack.
			for val, err := peek(opStack); val != '('; val, err = peek(opStack) {
				if err != nil {
					panic("ERROR: Imbalanced parantheses.")
				}
				to_append := mustPop(&opStack)
				outQueue = append(outQueue, newPostfixNode(to_append))
			}
			_ = mustPop(&opStack) // Get rid of opening parantheses
		}
	}

	// Pop all remaining operators (and append to outQueue)
	for len(opStack) > 0 {
		to_append := mustPop(&opStack)
		outQueue = append(outQueue, newPostfixNode(to_append))
	}

	return outQueue
}

// Thompson's algorithm. Constructs Finite-State Automaton from given string.
// Returns start state.
func thompson(re []postfixNode) *State {
	nfa := make([]*State, 0) // Stack of states
	for _, c := range re {
		if c.nodetype == CHARACTER {
			state := State{}
			state.transitions = make(map[int][]*State)
			state.content = rune2Contents(c.contents)
			state.output = make([]*State, 0)
			state.output = append(state.output, &state)
			state.isEmpty = false
			nfa = append(nfa, &state)
		}
		// Must be an operator if it isn't a character
		switch c.nodetype {
		case CONCATENATE:
			s2 := mustPop(&nfa)
			s1 := mustPop(&nfa)
			s1 = concatenate(s1, s2)
			nfa = append(nfa, s1)
		case KLEENE: // Create a 0-state, concat the popped state after it, concat the 0-state after the popped state
			s1 := mustPop(&nfa)
			stateToAdd := kleene(*s1)
			nfa = append(nfa, stateToAdd)
		case PLUS: // a+ is equivalent to aa*
			s1 := mustPop(&nfa)
			s2 := kleene(*s1)
			s1 = concatenate(s1, s2)
			nfa = append(nfa, s1)
		case QUESTION: // ab? is equivalent to a(b|)
			s1 := mustPop(&nfa)
			s2 := &State{}
			s2.transitions = make(map[int][]*State)
			s2.content = newContents(EPSILON)
			s2.output = append(s2.output, s2)
			s2.isEmpty = true
			s3 := alternate(s1, s2)
			nfa = append(nfa, s3)
		case PIPE:
			s1 := mustPop(&nfa)
			s2 := mustPop(&nfa)
			s3 := alternate(s1, s2)
			nfa = append(nfa, s3)
		}
	}
	if len(nfa) != 1 {
		panic("ERROR: Invalid Regex.")
	}

	verifyLastStates(nfa)

	return nfa[0]

}

func main() {
	// Process:
	// 1. Convert regex into postfix notation (Shunting-Yard algorithm)
	// 		a. Add explicit concatenation operators to facilitate this
	// 2. Build NFA from postfix representation (Thompson's algorithm)
	// 3. Run the string against the NFA
	if len(os.Args) != 2 {
		fmt.Println("ERROR: Missing cmdline args")
		os.Exit(22)
	}
	var re string
	re = os.Args[1]
	var test_str string
	// Read test string from stdin
	reader := bufio.NewReader(os.Stdin)
	test_str, err := reader.ReadString('\n')
	if err != nil {
		panic(err)
	}
	fmt.Scanln(&test_str)
	re_postfix := shuntingYard(re)
	// fmt.Println(re_postfix)
	startState := thompson(re_postfix)
	matchIndices := findAllMatches(startState, test_str)
	inColor := false
	if len(matchIndices) > 0 {
		for i, c := range test_str {
			for _, indices := range matchIndices {
				if i >= indices.startIdx && i < indices.endIdx {
					color.New(color.FgRed).Printf("%c", c)
					inColor = true
					break
				}
			}
			if inColor == false {
				fmt.Printf("%c", c)
			}
			inColor = false
		}
	} else {
		fmt.Print(test_str)
	}
}