Simplify handling of newline characters at the end of previous items …

…in the parse loop.

PiperOrigin-RevId: 597594817

parser/parser.go

@@ -567,24 +567,23 @@ func (p *parser) parse(isRoot bool) (result []*ast.Node, endPos ast.Position, er
 			return nil, ast.Position{}, err
 		}
 
-		startPos := p.position()
 		if p.nextInputIs('\n') {
 			// p.parse is often invoked with the index pointing at the
 			// newline character after the previous item.
 			// We should still report that this item starts in the next line.
-			startPos.Byte++
-			startPos.Line++
-			startPos.Column = 1
+			p.consume('\n')
 		}
+		startPos := p.position()
 
 		// Read PreComments.
 		comments, blankLines := p.skipWhiteSpaceAndReadComments(true /* multiLine */)
 
 		// Handle blank lines.
-		if blankLines > 1 {
+		if blankLines > 0 {
 			if p.config.infoLevel() {
 				p.config.infof("blankLines: %v", blankLines)
 			}
+			// Here we collapse the leading blank lines into one blank line.
 			comments = append([]string{""}, comments...)
 		}
 
@@ -824,15 +823,20 @@ func (p *parser) readContinuousBlocksOfComments() []string {
 // skipWhiteSpaceAndReadComments has multiple cases:
 //   - (1) reading a block of comments followed by a blank line
 //   - (2) reading a block of comments followed by non-blank content
-//   - (3) reading the inline comments between the current char and the end of the
-//     current line
+//   - (3) reading the inline comments between the current char and the end of
+//     the current line
+//
+// In both cases (1) and (2), there can also be blank lines before the comment
+// starts.
 //
-// Lines of comments and number of blank lines will be returned.
+// Lines of comments and number of blank lines before the comment will be
+// returned. If there is no comment, the returned slice will be empty.
 func (p *parser) skipWhiteSpaceAndReadComments(multiLine bool) ([]string, int) {
 	i := p.index
 	var foundComment, insideComment bool
 	commentBegin := 0
 	var comments []string
+	// Number of blanks lines *before* the comment (if any) starts.
 	blankLines := 0
 	for ; i < p.length; i++ {
 		if p.in[i] == '#' && !insideComment {

parser/parser_test.go

@@ -1126,6 +1126,57 @@ presubmit: {
     prohibited_regexp: "UnsafeFunction"
   }
 }
+`}, {
+		name: "blank lines before comment blocks are collapsed to one",
+		in: `# txtpbfmt: sort_repeated_fields_by_content
+presubmit: {
+  check_contents: {
+
+
+    # Should go after ADD.
+    # And the empty lines above this are collapsed into one blank line.
+    operation: EDIT
+    operation: ADD
+
+
+
+    operation: REMOVE
+  }
+}
+`,
+		out: `# txtpbfmt: sort_repeated_fields_by_content
+presubmit: {
+  check_contents: {
+    operation: ADD
+
+    # Should go after ADD.
+    # And the empty lines above this are collapsed into one blank line.
+    operation: EDIT
+
+    operation: REMOVE
+  }
+}
+`}, {
+		name: "blank lines before comment blocks are collapsed to one",
+		in: `# txtpbfmt: sort_repeated_fields_by_content
+presubmit: {
+  check_contents: {
+    # This comment is a separate node and does not move when the fields are sorted.
+
+    operation: EDIT
+    operation: ADD
+  }
+}
+`,
+		out: `# txtpbfmt: sort_repeated_fields_by_content
+presubmit: {
+  check_contents: {
+    # This comment is a separate node and does not move when the fields are sorted.
+    operation: ADD
+
+    operation: EDIT
+  }
+}
 `}, {
 		name: "sort by subfield values",
 		in: `# txtpbfmt: sort_repeated_fields_by_subfield=operation.name
@@ -1318,6 +1369,33 @@ field: "a"
 field: "b"
 message: { id: "a" }
 message: { id: "b" }
+`}, {
+		name: "detached comment creates a new group for sorting",
+		in: `# txtpbfmt: sort_repeated_fields_by_content
+
+# field c
+field: "c"
+
+# field a
+field: "a"
+
+# new group - the fields below don't get sorted with the fields above
+
+# field b
+field: "b"
+`,
+		out: `# txtpbfmt: sort_repeated_fields_by_content
+
+# field a
+field: "a"
+
+# field c
+field: "c"
+
+# new group - the fields below don't get sorted with the fields above
+
+# field b
+field: "b"
 `}, {
 		name: "trailing comma / semicolon",
 		in: `dict: {