feat: carry the byte offset and column in SyntaxError

Assisted-by: GLM 5.3 Flash

parser.go

@@ -4,6 +4,7 @@
 package interpres
 
 import (
+	"bytes"
 	"context"
 	"fmt"
 	"strconv"
@@ -652,7 +653,7 @@ func (p *parser) parseKeyComponent() (string, error) {
 		// does not decode names that, before any grammar message can.
 		if !p.eof() && p.peek() >= utf8.RuneSelf {
 			if r, size := utf8.DecodeRune(p.src[p.pos:]); r == utf8.RuneError && size == 1 {
-				return "", p.errf("invalid UTF-8 in key")
+				return "", p.errf("invalid UTF-8 in key at byte offset %d", p.pos)
 			}
 		}
 		if p.pos == start {
@@ -707,7 +708,7 @@ func (p *parser) parseAtom() (any, error) {
 		return nil, p.errf("expected a value")
 	}
 	if hasHighByte(tok) && !utf8.ValidString(tok) {
-		return nil, p.errf("invalid UTF-8 in value")
+		return nil, p.errf("invalid UTF-8 in value at byte offset %d", p.pos)
 	}
 	// A date may be followed by a space and a time, forming one date-time.
 	if isDateToken(tok) && !p.eof() && p.peek() == ' ' {
@@ -885,7 +886,7 @@ func (p *parser) writeContentRune(b *strings.Builder) error {
 	}
 	r, size := utf8.DecodeRune(p.src[p.pos:])
 	if r == utf8.RuneError && size == 1 {
-		return p.errf("invalid UTF-8 in string")
+		return p.errf("invalid UTF-8 in string at byte offset %d", p.pos)
 	}
 	p.pos += size
 	b.WriteRune(r)
@@ -1393,7 +1394,7 @@ func (p *parser) skipComment() (string, error) {
 		default:
 			r, size := utf8.DecodeRune(p.src[p.pos:])
 			if r == utf8.RuneError && size == 1 {
-				return "", p.errf("invalid UTF-8 in comment")
+				return "", p.errf("invalid UTF-8 in comment at byte offset %d", p.pos)
 			}
 			p.pos += size
 		}
@@ -1451,13 +1452,21 @@ func (p *parser) expectLineEnd() error {
 	}
 	r, size := utf8.DecodeRune(p.src[p.pos:])
 	if r == utf8.RuneError && size == 1 {
-		return p.errf("invalid UTF-8 after value")
+		return p.errf("invalid UTF-8 after value at byte offset %d", p.pos)
 	}
 	return p.errf("unexpected %q after value", string(r))
 }
 
+// errf builds the SyntaxError with the position the scan stopped at: the line,
+// the byte offset in the input, and the 1-based column on that line. The
+// offset is the cursor, which on an escape or a delimiter run sits just after
+// the bytes that caused the complaint; SourceLine renders the caret there.
 func (p *parser) errf(format string, args ...any) error {
-	return &SyntaxError{Line: p.line, Msg: fmt.Sprintf(format, args...)}
+	col := p.pos + 1
+	if start := bytes.LastIndexByte(p.src[:p.pos], '\n'); start >= 0 {
+		col = p.pos - start
+	}
+	return &SyntaxError{Line: p.line, Offset: p.pos, Column: col, Msg: fmt.Sprintf(format, args...)}
 }
 
 // pathKey joins key components with a NUL separator so a dotted path can be