From c8613c1ba2d162348af72caf499bdc790511e835 Mon Sep 17 00:00:00 2001
From: Rockingcool <aadhavan@twomorecents.org>
Date: Mon, 6 Jan 2025 20:08:24 -0600
Subject: [PATCH] Major restructuring - added new type, changed return types
 for shuntingYard and thompson

I added a new function 'Compile' that calls shuntingYard and thompson. I also added
a new type 'Reg' that this function returns - it represents the starting state and contains
the number of capturing groups in the regex. I also rewrote shuntingYard and thompson
to return errors instead of panicking.
---
 compile.go | 60 ++++++++++++++++++++++++++++++++++++++++++++++--------
 1 file changed, 52 insertions(+), 8 deletions(-)

diff --git a/compile.go b/compile.go
index 9de2b81..9311a21 100644
--- a/compile.go
+++ b/compile.go
@@ -7,6 +7,29 @@ import (
 	"unicode"
 )
 
+// A Reg represents the result of compiling a regular expression. It contains
+// the startState of the NFA representation of the regex, and the number of capturing
+// groups in the regex.
+type Reg struct {
+	start     *State
+	numGroups int
+}
+
+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.
@@ -416,7 +439,7 @@ func shuntingYard(re string) ([]postfixNode, error) {
 
 // Thompson's algorithm. Constructs Finite-State Automaton from given string.
 // Returns start state and number of groups in regex.
-func thompson(re []postfixNode) (*State, int) {
+func thompson(re []postfixNode) (Reg, error) {
 	nfa := make([]*State, 0) // Stack of states
 	numGroups := 0           // Number of capturing groups
 	for _, c := range re {
@@ -465,10 +488,16 @@ func thompson(re []postfixNode) (*State, int) {
 							state.assert = NLB
 						}
 					}
-					tmpRe := shuntingYard(state.lookaroundRegex)
-					var numGroupsLookaround int
-					state.lookaroundNFA, numGroupsLookaround = thompson(tmpRe)
-					state.lookaroundNumCaptureGroups = numGroupsLookaround
+					tmpRe, err := shuntingYard(state.lookaroundRegex)
+					if err != nil {
+						return Reg{}, fmt.Errorf("Error parsing lookaround: %w", err)
+					}
+					reg, err := thompson(tmpRe)
+					if err != nil {
+						return Reg{}, fmt.Errorf("Error compiling lookaround: %w", err)
+					}
+					state.lookaroundNFA = reg.start
+					state.lookaroundNumCaptureGroups = reg.numGroups
 
 				}
 			}
@@ -533,7 +562,7 @@ func thompson(re []postfixNode) (*State, int) {
 		}
 		if c.startReps != 1 || c.endReps != 1 { // Must have a numeric specifier attached to it
 			if c.endReps != -1 && c.endReps < c.startReps {
-				panic("ERROR: Numeric specifier - start greater than end.")
+				return Reg{}, fmt.Errorf("Numeric specifier - start greater than end.")
 			}
 			state := mustPop(&nfa)
 			var stateToAdd *State = nil
@@ -564,11 +593,26 @@ func thompson(re []postfixNode) (*State, int) {
 		}
 	}
 	if len(nfa) != 1 {
-		panic("ERROR: Invalid Regex.")
+		return Reg{}, fmt.Errorf("Invalid Regex.")
 	}
 
 	verifyLastStates(nfa)
 
-	return nfa[0], numGroups
+	return Reg{nfa[0], numGroups}, nil
 
 }
+
+// Compiles the given regular expression into a Reg type, suitable for use with the
+// matching functions. The second return value is non-nil if a compilation error has
+// occured. As such, the error value must be checked before using the Reg returned by this function.
+func Compile(re string) (Reg, error) {
+	nodes, err := shuntingYard(re)
+	if err != nil {
+		return Reg{}, fmt.Errorf("Error parsing regex: %w", err)
+	}
+	reg, err := thompson(nodes)
+	if err != nil {
+		return Reg{}, fmt.Errorf("Error compiling regex: %w", err)
+	}
+	return reg, nil
+}