Wrote MarshalText() and UnmarshalText() to implement TextMarshaler and TextUnmarshaler

regex/compile.go

@@ -31,6 +31,22 @@ func (re Reg) String() string {
 	return re.str
 }
 
+// MarshalText implements [encoding.TextMarshaler]. The output is equivalent to that of [Reg.String].
+// Any flags passed as arguments (including calling [Reg.Longest]) are lost.
+func (re *Reg) MarshalText() ([]byte, error) {
+	return []byte(re.String()), nil
+}
+
+// UnmarshalText implements [encoding.TextUnmarshaler]. It calls [Reg.Compile] on the given byte-slice. If it returns successfully,
+// then the result of the compilation is stored in re. The result of [Reg.Compile] is returned.
+func (re *Reg) UnmarshalText(text []byte) error {
+	newReg, err := Compile(string(text))
+	if err == nil {
+		*re = newReg
+	}
+	return err
+}
+
 func (re *Reg) Longest() {
 	re.preferLongest = true
 }

regex/doc.go

@@ -105,23 +105,7 @@ The key differences are mentioned below.
 
 1. Greediness:
 
-This engine does not support non-greedy operators. All operators are always greedy in nature, and will try
-to match as much as they can, while still allowing for a successful match. For example, given the regex:
-
-	y*y
-
-The engine will match as many 'y's as it can, while still allowing the trailing 'y' to be matched.
-
-Another, more subtle example is the following regex:
-
-	x|xx
-
-While the stdlib implementation (and most other engines) will prefer matching the first item of the alternation,
-this engine will go for the longest possible match, regardless of the order of the alternation. Although this
-strays from the convention, it results in a nice rule-of-thumb - the engine is ALWAYS greedy.
-
-The stdlib implementation has a function [regexp.Regexp.Longest] which makes future searches prefer the longest match.
-That is the default (and unchangable) behavior in this engine.
+This engine currently does not support non-greedy operators.
 
 2. Byte-slices and runes:
 
@@ -166,7 +150,7 @@ The following features from [regexp] are (currently) NOT supported:
  1. Named capturing groups
  2. Non-greedy operators
  3. Unicode character classes
- 4. Embedded flags (flags are passed as arguments to [Compile])
+ 4. Embedded flags (flags are instead passed as arguments to [Compile])
  5. Literal text with \Q ... \E
 
 The following features are not available in [regexp], but are supported in my engine:

regex/example_test.go

@@ -154,3 +154,11 @@ func ExampleReg_Longest() {
 	// Output: x
 	// xx
 }
+
+func ExampleReg_ReplaceAll() {
+	regexStr := `(\d)(\w)`
+	inputStr := "5d9t"
+	regexComp := regex.MustCompile(regexStr)
+	fmt.Println(regexComp.ReplaceAll(inputStr, `$2$1`))
+	// Output: d5t9
+}

regex/matching.go

@@ -350,7 +350,7 @@ func (re Reg) Expand(dst string, template string, src string, match Match) strin
 				i++
 			} else {
 				numStr := ""
-				for unicode.IsDigit(templateRuneSlc[i]) {
+				for i < len(templateRuneSlc) && unicode.IsDigit(templateRuneSlc[i]) {
 					numStr += string(templateRuneSlc[i])
 					i++
 				}
@@ -395,3 +395,23 @@ func (re Reg) LiteralPrefix() (prefix string, complete bool) {
 	}
 	return prefix, complete
 }
+
+// ReplaceAll replaces all matches of the expression in src, with the text in repl. In repl, variables are interpreted
+// as they are in [Reg.Expand]. The resulting string is returned.
+func (re Reg) ReplaceAll(src string, repl string) string {
+	matches := re.FindAllSubmatch(src)
+	i := 0
+	currentMatch := 0
+	dst := ""
+	for i < len(src) {
+		if currentMatch <= len(matches) && matches[currentMatch][0].IsValid() && i == matches[currentMatch][0].StartIdx {
+			dst += re.Expand("", repl, src, matches[currentMatch])
+			i = matches[currentMatch][0].EndIdx
+			currentMatch++
+		} else {
+			dst += string(src[i])
+			i++
+		}
+	}
+	return dst
+}