refactor(jsoncommentstrip): replace go-jsoncommentstrip with custom JSON comment stripping

utils/jsoncommentstrip/jsoncommentstrip.go

@@ -0,0 +1,128 @@
+// Package jsoncommentstrip provides an io.Reader that strips JavaScript-style
+// comments (// line and /* block */) from JSON input while preserving
+// comment-like sequences inside JSON string values.
+package jsoncommentstrip
+
+import (
+	"bufio"
+	"io"
+)
+
+type state int
+
+const (
+	stateNormal state = iota
+	stateInString
+	stateInStringEscape
+	stateMaybeComment  // saw '/'
+	stateLineComment   // inside // ...
+	stateBlockComment  // inside /* ... */
+	stateMaybeBlockEnd // saw '*' inside block comment
+)
+
+type reader struct {
+	r     *bufio.Reader
+	state state
+}
+
+// NewReader returns an io.Reader that strips JSON comments from the
+// underlying reader. It removes single-line comments (// to end of line)
+// and block comments (/* ... */), while preserving comment-like sequences
+// that appear inside JSON string values.
+func NewReader(r io.Reader) io.Reader {
+	return &reader{
+		r:     bufio.NewReader(r),
+		state: stateNormal,
+	}
+}
+
+func (cr *reader) Read(p []byte) (int, error) {
+	n := 0
+	for n < len(p) {
+		b, err := cr.r.ReadByte()
+		if err != nil {
+			if cr.state == stateMaybeComment {
+				// Emit the pending '/' before returning EOF
+				p[n] = '/'
+				n++
+				cr.state = stateNormal
+			}
+			return n, err
+		}
+
+		switch cr.state {
+		case stateNormal:
+			switch b {
+			case '"':
+				p[n] = b
+				n++
+				cr.state = stateInString
+			case '/':
+				cr.state = stateMaybeComment
+			default:
+				p[n] = b
+				n++
+			}
+
+		case stateInString:
+			p[n] = b
+			n++
+			switch b {
+			case '\\':
+				cr.state = stateInStringEscape
+			case '"':
+				cr.state = stateNormal
+			}
+
+		case stateInStringEscape:
+			p[n] = b
+			n++
+			cr.state = stateInString
+
+		case stateMaybeComment:
+			switch b {
+			case '/':
+				cr.state = stateLineComment
+			case '*':
+				cr.state = stateBlockComment
+			default:
+				// The '/' was not a comment start; emit it and the current byte
+				p[n] = '/'
+				n++
+				if n < len(p) {
+					p[n] = b
+					n++
+				} else {
+					// We need to "unread" the current byte since buffer is full
+					_ = cr.r.UnreadByte()
+				}
+				cr.state = stateNormal
+			}
+
+		case stateLineComment:
+			if b == '\n' || b == '\r' {
+				p[n] = b
+				n++
+				cr.state = stateNormal
+			}
+			// Otherwise, consume and discard
+
+		case stateBlockComment:
+			if b == '*' {
+				cr.state = stateMaybeBlockEnd
+			}
+			// Otherwise, consume and discard
+
+		case stateMaybeBlockEnd:
+			if b == '/' {
+				cr.state = stateNormal
+			} else if b == '*' {
+				// Stay in stateMaybeBlockEnd (consecutive *'s)
+				cr.state = stateMaybeBlockEnd
+			} else {
+				cr.state = stateBlockComment
+			}
+		}
+	}
+	return n, nil
+}