lib/protoparser/csvimport: properly parse the last empty column in CSV line

Do not ignore the last empty column in CSV line.
While at it, properly parse CSV columns in single quotes, e.g. `'foo,bar',baz` is parsed as two columns - `foo,bar` and `baz`

Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/4048

See also https://github.com/VictoriaMetrics/VictoriaMetrics/pull/4298

docs/CHANGELOG.md

@@ -20,6 +20,7 @@ The following tip changes can be tested by building VictoriaMetrics components f
 
 * BUGFIX: reduce the probability of sudden increase in the number of small parts on systems with small number of CPU cores.
 * BUGFIX: reduce the possibility of increased CPU usage when data with timestamps older than one hour is ingested into VictoriaMetrics. This reduces spikes for the graph `sum(rate(vm_slow_per_day_index_inserts_total))`. See [this pull request](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/4258).
+* BUGFIX: do not ignore trailing empty field in CSV lines when [importing data in CSV format](https://docs.victoriametrics.com/#how-to-import-csv-data). See [this issue](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/4048).
 * BUGFIX: [MetricsQL](https://docs.victoriametrics.com/MetricsQL.html): fix a panic when the duration in the query contains uppercase `M` suffix. Such a suffix isn't allowed to use in durations, since it clashes with `a million` suffix, e.g. it isn't clear whether `rate(metric[5M])` means rate over 5 minutes, 5 months or 5 million seconds. See [this](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3589) and [this](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/4120) issues.
 * BUGFIX: [VictoriaMetrics cluster](https://docs.victoriametrics.com/Cluster-VictoriaMetrics.html): prevent from possible panic when the number of vmstorage nodes increases when [automatic vmstorage discovery](https://docs.victoriametrics.com/Cluster-VictoriaMetrics.html#automatic-vmstorage-discovery) is enabled.
 * BUGFIX: properly limit the number of [OpenTSDB HTTP](https://docs.victoriametrics.com/#sending-opentsdb-data-via-http-apiput-requests) concurrent requests specified via `-maxConcurrentInserts` command-line flag. See [this issue](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/4204). Thanks to @zouxiang1993 for [the fix](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/4208).

lib/protoparser/csvimport/column_descriptor.go

@@ -28,6 +28,10 @@ type ColumnDescriptor struct {
 	MetricName string
 }
 
+func (cd *ColumnDescriptor) isEmpty() bool {
+	return cd.ParseTimestamp == nil && cd.TagName == "" && cd.MetricName == ""
+}
+
 const maxColumnsPerRow = 64 * 1024
 
 // ParseColumnDescriptors parses column descriptors from s.

lib/protoparser/csvimport/parser.go

@@ -88,6 +88,10 @@ func parseRows(sc *scanner, dst []Row, tags []Tag, metrics []metric, cds []Colum
 			}
 			cd := &cds[col]
 			col++
+			if cd.isEmpty() || sc.Column == "" {
+				// Ignore empty column.
+				continue
+			}
 			if parseTimestamp := cd.ParseTimestamp; parseTimestamp != nil {
 				timestamp, err := parseTimestamp(sc.Column)
 				if err != nil {
@@ -105,9 +109,8 @@ func parseRows(sc *scanner, dst []Row, tags []Tag, metrics []metric, cds []Colum
 				continue
 			}
 			metricName := cd.MetricName
-			if metricName == "" || sc.Column == "" {
+			if metricName == "" {
-				// The given field is ignored.
+				logger.Panicf("BUG: unexpected empty MetricName")
-				continue
 			}
 			value, err := fastfloat.Parse(sc.Column)
 			if err != nil {
@@ -127,7 +130,7 @@ func parseRows(sc *scanner, dst []Row, tags []Tag, metrics []metric, cds []Colum
 			continue
 		}
 		if len(metrics) == 0 {
-			logger.Panicf("BUG: expecting at least a single metric in columnDescriptors=%#v", cds)
+			continue
 		}
 		r.Metric = metrics[0].Name
 		r.Tags = tags[tagsLen:]

lib/protoparser/csvimport/parser_test.go

@@ -30,6 +30,8 @@ func TestRowsUnmarshalFailure(t *testing.T) {
 
 	// Missing columns
 	f("3:metric:aaa", "123,456")
+	f("1:metric:foo,2:label:bar", "123")
+	f("1:label:foo,2:metric:bar", "aaa")
 
 	// Invalid value
 	f("1:metric:foo", "12foobar")
@@ -189,6 +191,35 @@ func TestRowsUnmarshalSuccess(t *testing.T) {
 			Value:  2,
 		},
 	})
+	// last metric with empty value
+	// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/4048
+	f("1:metric:foo,2:metric:bar", `123,`, []Row{
+		{
+			Metric: "foo",
+			Value:  123,
+		},
+	})
+	// all the metrics with empty values
+	f(`1:metric:foo,2:metric:bar,3:label:xx`, `,,abc`, nil)
+	// labels with empty value
+	f("1:metric:foo,2:label:bar,3:label:baz,4:label:xxx", "123,x,,", []Row{
+		{
+			Metric: "foo",
+			Tags: []Tag{
+				{
+					Key:   "bar",
+					Value: "x",
+				},
+			},
+			Value: 123,
+		},
+	})
+	f("1:metric:foo,2:label:bar,3:label:baz,4:label:xxx", "123,,,", []Row{
+		{
+			Metric: "foo",
+			Value:  123,
+		},
+	})
 	// see https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3540
 	f("1:label:mytest,2:time:rfc3339,3:metric:M10,4:metric:M20,5:metric:M30,6:metric:M40,7:metric:M50,8:metric:M60",
 		`test,2022-12-25T16:57:12+01:00,10,20,30,,,60,70,80`, []Row{

lib/protoparser/csvimport/scanner.go

@@ -17,6 +17,9 @@ type scanner struct {
 	// It is cleared on NextLine call.
 	Error error
 
+	// isLastColumn is set to true when the last column at the given Line is processed
+	isLastColumn bool
+
 	s string
 }
 
@@ -25,6 +28,7 @@ func (sc *scanner) Init(s string) {
 	sc.Line = ""
 	sc.Column = ""
 	sc.Error = nil
+	sc.isLastColumn = false
 	sc.s = s
 }
 
@@ -35,23 +39,26 @@ func (sc *scanner) Init(s string) {
 // false is returned if no more lines left in sc.s
 func (sc *scanner) NextLine() bool {
 	s := sc.s
+	sc.Line = ""
 	sc.Error = nil
+	sc.isLastColumn = false
 	for len(s) > 0 {
 		n := strings.IndexByte(s, '\n')
 		var line string
 		if n >= 0 {
-			line = trimTrailingSpace(s[:n])
+			line = trimTrailingCR(s[:n])
 			s = s[n+1:]
 		} else {
-			line = trimTrailingSpace(s)
+			line = trimTrailingCR(s)
 			s = ""
 		}
-		sc.Line = line
-		sc.s = s
 		if len(line) > 0 {
+			sc.Line = line
+			sc.s = s
 			return true
 		}
 	}
+	sc.s = ""
 	return false
 }
 
@@ -60,16 +67,28 @@ func (sc *scanner) NextLine() bool {
 // false is returned if no more columns left in sc.Line or if any error occurs.
 // sc.Error is set to error in the case of error.
 func (sc *scanner) NextColumn() bool {
+	if sc.isLastColumn || sc.Error != nil {
+		return false
+	}
 	s := sc.Line
-	if len(s) == 0 {
+	if strings.HasPrefix(s, `"`) || strings.HasPrefix(s, "'") {
-		return false
+		field, tail, err := readQuotedField(s)
-	}
+		if err != nil {
-	if sc.Error != nil {
+			sc.Error = err
-		return false
+			return false
-	}
+		}
-	if s[0] == '"' {
+		sc.Column = field
-		sc.Column, sc.Line, sc.Error = readQuotedField(s)
+		if len(tail) == 0 {
-		return sc.Error == nil
+			sc.isLastColumn = true
+		} else {
+			if tail[0] != ',' {
+				sc.Error = fmt.Errorf("missing comma after quoted field in %q", s)
+				return false
+			}
+			tail = tail[1:]
+		}
+		sc.Line = tail
+		return true
 	}
 	n := strings.IndexByte(s, ',')
 	if n >= 0 {
@@ -78,11 +97,12 @@ func (sc *scanner) NextColumn() bool {
 	} else {
 		sc.Column = s
 		sc.Line = ""
+		sc.isLastColumn = true
 	}
 	return true
 }
 
-func trimTrailingSpace(s string) string {
+func trimTrailingCR(s string) string {
 	if len(s) > 0 && s[len(s)-1] == '\r' {
 		return s[:len(s)-1]
 	}
@@ -90,38 +110,32 @@ func trimTrailingSpace(s string) string {
 }
 
 func readQuotedField(s string) (string, string, error) {
-	sOrig := s
+	quote := s[0]
-	if len(s) == 0 || s[0] != '"' {
+	offset := 1
-		return "", sOrig, fmt.Errorf("missing opening quote for %q", sOrig)
+	n := strings.IndexByte(s[offset:], quote)
+	if n < 0 {
+		return "", s, fmt.Errorf("missing closing quote for %q", s)
 	}
-	s = s[1:]
+	offset += n + 1
-	hasEscapedQuote := false
+	if offset >= len(s) || s[offset] != quote {
+		// Fast path - the quoted string doesn't contain escaped quotes
+		return s[1 : offset-1], s[offset:], nil
+	}
+	// Slow path - the quoted string contains escaped quote
+	buf := make([]byte, 0, len(s)-2)
+	buf = append(buf, s[1:offset]...)
 	for {
-		n := strings.IndexByte(s, '"')
+		offset++
+		n := strings.IndexByte(s[offset:], quote)
 		if n < 0 {
-			return "", sOrig, fmt.Errorf("missing closing quote for %q", sOrig)
+			return "", s, fmt.Errorf("missing closing quote for %q", s)
 		}
-		s = s[n+1:]
+		buf = append(buf, s[offset:offset+n]...)
-		if len(s) == 0 {
+		offset += n + 1
-			// The end of string found
+		if offset < len(s) && s[offset] == quote {
-			return unquote(sOrig[1:len(sOrig)-1], hasEscapedQuote), "", nil
+			buf = append(buf, quote)
-		}
-		if s[0] == '"' {
-			// Take into account escaped quote
-			s = s[1:]
-			hasEscapedQuote = true
 			continue
 		}
-		if s[0] != ',' {
+		return string(buf), s[offset:], nil
-			return "", sOrig, fmt.Errorf("missing comma after quoted field in %q", sOrig)
-		}
-		return unquote(sOrig[1:len(sOrig)-len(s)-1], hasEscapedQuote), s[1:], nil
 	}
 }
-
-func unquote(s string, hasEscapedQuote bool) string {
-	if !hasEscapedQuote {
-		return s
-	}
-	return strings.ReplaceAll(s, `""`, `"`)
-}

lib/protoparser/csvimport/scanner_test.go

@@ -1,63 +1,117 @@
 package csvimport
 
 import (
+	"reflect"
 	"testing"
 )
 
+func TestReadQuotedFieldSuccess(t *testing.T) {
+	f := func(s, resultExpected, tailExpected string) {
+		t.Helper()
+		result, tail, err := readQuotedField(s)
+		if err != nil {
+			t.Fatalf("unexpected error: %s", err)
+		}
+		if result != resultExpected {
+			t.Fatalf("unexpected result; got %q; want %q", result, resultExpected)
+		}
+		if tail != tailExpected {
+			t.Fatalf("unexpected tail; got %q; want %q", tail, tailExpected)
+		}
+	}
+
+	// double quotes
+	f(`""`, ``, ``)
+	f(`"",`, ``, `,`)
+	f(`"",foobar`, ``, `,foobar`)
+	f(`"","bc"`, ``, `,"bc"`)
+	f(`"a"`, `a`, ``)
+	f(`"a"bc`, `a`, `bc`)
+	f(`"foo`+"`',\n\t\r"+`bar"baz`, "foo`',\n\t\rbar", "baz")
+
+	// single quotes
+	f(`''`, ``, ``)
+	f(`'',`, ``, `,`)
+	f(`'',foobar`, ``, `,foobar`)
+	f(`'','bc'`, ``, `,'bc'`)
+	f(`'a'`, `a`, ``)
+	f(`'a'bc`, `a`, `bc`)
+	f(`'foo"`+"`,\n\t\r"+`bar'baz`, "foo\"`,\n\t\rbar", "baz")
+
+	// escaped double quotes
+	f(`" foo""bar"baz`, ` foo"bar`, `baz`)
+	f(`""""bar"baz`, `"`, `bar"baz`)
+	f(`"a,""b""'c",d,"e"`, `a,"b"'c`, `,d,"e"`)
+
+	// escaped single quotes
+	f(`' foo''bar'baz`, ` foo'bar`, `baz`)
+	f(`''''bar'baz`, `'`, `bar'baz`)
+	f(`'''bar'''baz`, `'bar'`, `baz`)
+	f(`'a,''b''"c',d,'e'`, `a,'b'"c`, `,d,'e'`)
+}
+
+func TestReadQuotedFieldFailure(t *testing.T) {
+	f := func(s string) {
+		t.Helper()
+		field, tail, err := readQuotedField(s)
+		if field != "" {
+			t.Fatalf("unexpected non-empty field returned: %q", field)
+		}
+		if tail != s {
+			t.Fatalf("unexpected tail returned; got %q; want %q", tail, s)
+		}
+		if err == nil {
+			t.Fatalf("expecting non-nil error")
+		}
+	}
+	f(`"`)
+	f(`'`)
+	f(`"foo""`)
+	f(`'foo''`)
+	f(`'foo''`)
+}
+
 func TestScannerSuccess(t *testing.T) {
-	var sc scanner
+	f := func(s string, rowsExpected [][]string) {
-	sc.Init("foo,bar\n\"aa,\"\"bb\",\"\"")
+		t.Helper()
-	if !sc.NextLine() {
+		var sc scanner
-		t.Fatalf("expecting the first line")
+		sc.Init(s)
-	}
+		var rows [][]string
-	if sc.Line != "foo,bar" {
+		for sc.NextLine() {
-		t.Fatalf("unexpected line; got %q; want %q", sc.Line, "foo,bar")
+			var row []string
-	}
+			for sc.NextColumn() {
-	if !sc.NextColumn() {
+				row = append(row, sc.Column)
-		t.Fatalf("expecting the first column")
+			}
-	}
+			rows = append(rows, row)
-	if sc.Column != "foo" {
+		}
-		t.Fatalf("unexpected first column; got %q; want %q", sc.Column, "foo")
+		if sc.Error != nil {
-	}
+			t.Fatalf("unexpected error: %s", sc.Error)
-	if !sc.NextColumn() {
+		}
-		t.Fatalf("expecting the second column")
+		if !reflect.DeepEqual(rows, rowsExpected) {
-	}
+			t.Fatalf("unexpected rows;\ngot\n%q\nwant\n%q", rows, rowsExpected)
-	if sc.Column != "bar" {
+		}
-		t.Fatalf("unexpected second column; got %q; want %q", sc.Column, "bar")
-	}
-	if sc.NextColumn() {
-		t.Fatalf("unexpected next column: %q", sc.Column)
-	}
-	if sc.Error != nil {
-		t.Fatalf("unexpected error: %s", sc.Error)
-	}
-	if !sc.NextLine() {
-		t.Fatalf("expecting the second line")
-	}
-	if sc.Line != "\"aa,\"\"bb\",\"\"" {
-		t.Fatalf("unexpected the second line; got %q; want %q", sc.Line, "\"aa,\"\"bb\",\"\"")
-	}
-	if !sc.NextColumn() {
-		t.Fatalf("expecting the first column on the second line")
-	}
-	if sc.Column != "aa,\"bb" {
-		t.Fatalf("unexpected column on the second line; got %q; want %q", sc.Column, "aa,\"bb")
-	}
-	if !sc.NextColumn() {
-		t.Fatalf("expecting the second column on the second line")
-	}
-	if sc.Column != "" {
-		t.Fatalf("unexpected column on the second line; got %q; want %q", sc.Column, "")
-	}
-	if sc.NextColumn() {
-		t.Fatalf("unexpected next column on the second line: %q", sc.Column)
-	}
-	if sc.Error != nil {
-		t.Fatalf("unexpected error: %s", sc.Error)
-	}
-	if sc.NextLine() {
-		t.Fatalf("unexpected next line: %q", sc.Line)
 	}
+
+	f("", nil)
+	f("\n", nil)
+	f("\r\n\n\r", nil)
+	f("foo,bar\n\"aa,\"\"bb\",\"\"", [][]string{
+		{"foo", "bar"},
+		{`aa,"bb`, ``},
+	})
+	f(`fo"bar,baz'a,"bc""de",'g''e'`, [][]string{
+		{`fo"bar`, `baz'a`, `bc"de`, `g'e`},
+	})
+	f(`,`, [][]string{
+		{``, ``},
+	})
+	f(`foo`, [][]string{
+		{`foo`},
+	})
+	f(`foo,,`+"\r\n"+`,bar,`+"\n", [][]string{
+		{`foo`, ``, ``},
+		{``, `bar`, ``},
+	})
 }
 
 func TestScannerFailure(t *testing.T) {