Merge pull request #278 from goccy/feature/fix-parse-multiline-value

Fix indentState logic for multi-line value
goccy · Jan 11, 2022 · 5fdf77f · 5fdf77f
2 parents ac4bc56 + f6e3a46
commit 5fdf77f
Show file tree

Hide file tree

Showing 2 changed files with 70 additions and 34 deletions.
diff --git a/parser/parser_test.go b/parser/parser_test.go
@@ -547,25 +547,42 @@ b: c
 a: |+
   value
 b: c
+`,
+		},
+		{
+			`
+- key1: val
+  key2:
+    (
+      foo
+      +
+      bar
+    )
+`,
+			`
+- key1: val
+  key2: ( foo + bar )
 `,
 		},
 	}
 
 	for _, test := range tests {
-		tokens := lexer.Tokenize(test.source)
-		f, err := parser.Parse(tokens, 0)
-		if err != nil {
-			t.Fatalf("%+v", err)
-		}
-		var v Visitor
-		for _, doc := range f.Docs {
-			ast.Walk(&v, doc.Body)
-		}
-		expect := fmt.Sprintf("\n%+v\n", f)
-		if test.expect != expect {
-			tokens.Dump()
-			t.Fatalf("unexpected output: [%s] != [%s]", test.expect, expect)
-		}
+		t.Run(test.source, func(t *testing.T) {
+			tokens := lexer.Tokenize(test.source)
+			f, err := parser.Parse(tokens, 0)
+			if err != nil {
+				t.Fatalf("%+v", err)
+			}
+			var v Visitor
+			for _, doc := range f.Docs {
+				ast.Walk(&v, doc.Body)
+			}
+			expect := fmt.Sprintf("\n%+v\n", f)
+			if test.expect != expect {
+				tokens.Dump()
+				t.Fatalf("unexpected output: [%s] != [%s]", test.expect, expect)
+			}
+		})
 	}
 }
 

diff --git a/scanner/scanner.go b/scanner/scanner.go
@@ -128,27 +128,16 @@ func (s *Scanner) newLineCount(src []rune) int {
 	return cnt
 }
 
-func (s *Scanner) updateIndent(ctx *Context, c rune) {
-	if s.isFirstCharAtLine && s.isNewLineChar(c) && ctx.isDocument() {
-		return
-	}
-	if s.isFirstCharAtLine && c == ' ' {
-		s.indentNum++
-		return
-	}
-	if !s.isFirstCharAtLine {
-		s.indentState = IndentStateKeep
-		return
-	}
-
+func (s *Scanner) updateIndentState(ctx *Context) {
+	indentNumBasedIndentState := s.indentState
 	if s.prevIndentNum < s.indentNum {
 		s.indentLevel = s.prevIndentLevel + 1
-		s.indentState = IndentStateUp
+		indentNumBasedIndentState = IndentStateUp
 	} else if s.prevIndentNum == s.indentNum {
 		s.indentLevel = s.prevIndentLevel
-		s.indentState = IndentStateEqual
+		indentNumBasedIndentState = IndentStateEqual
 	} else {
-		s.indentState = IndentStateDown
+		indentNumBasedIndentState = IndentStateDown
 		if s.prevIndentLevel > 0 {
 			s.indentLevel = s.prevIndentLevel - 1
 		}
@@ -157,18 +146,48 @@ func (s *Scanner) updateIndent(ctx *Context, c rune) {
 	if s.prevIndentColumn > 0 {
 		if s.prevIndentColumn < s.column {
 			s.indentState = IndentStateUp
-		} else if s.prevIndentColumn == s.column {
-			s.indentState = IndentStateEqual
-		} else {
+		} else if s.prevIndentColumn != s.column || indentNumBasedIndentState != IndentStateEqual {
+			// The following case ( current position is 'd' ), some variables becomes like here
+			// - prevIndentColumn: 1 of 'a'
+			// - indentNumBasedIndentState: IndentStateDown because d's indentNum(1) is less than c's indentNum(3).
+			// Therefore, s.prevIndentColumn(1) == s.column(1) is true, but we want to treat this as IndentStateDown.
+			// So, we look also current indentState value by the above prevIndentNum based logic, and determins finally indentState.
+			// ---
+			// a:
+			//   b
+			//   c
+			// d: e
+			// ^
 			s.indentState = IndentStateDown
+		} else {
+			s.indentState = IndentStateEqual
 		}
+	} else {
+		s.indentState = indentNumBasedIndentState
 	}
+}
+
+func (s *Scanner) updateIndent(ctx *Context, c rune) {
+	if s.isFirstCharAtLine && s.isNewLineChar(c) && ctx.isDocument() {
+		return
+	}
+	if s.isFirstCharAtLine && c == ' ' {
+		s.indentNum++
+		return
+	}
+	if !s.isFirstCharAtLine {
+		s.indentState = IndentStateKeep
+		return
+	}
+	s.updateIndentState(ctx)
 	s.isFirstCharAtLine = false
 	if s.isNeededKeepPreviousIndentNum(ctx, c) {
 		return
 	}
+	if s.indentState != IndentStateUp {
+		s.prevIndentColumn = 0
+	}
 	s.prevIndentNum = s.indentNum
-	s.prevIndentColumn = 0
 	s.prevIndentLevel = s.indentLevel
 }