Preserve WebVTT tags in subtitle lines (asticode#97)

* [webvtt] Preserve WebVTT tags

* [webvtt/test] Testcase for webvtt tags

---------

Co-authored-by: Nhan Nguyen <nhan.trong.nguyen@edgeware.tv>

Author: NhanNguyen700

subtitles.go

@@ -236,19 +236,49 @@ type StyleAttributes struct {
 	TTMLWritingMode      *string
 	TTMLZIndex           *int
 	WebVTTAlign          string
-	WebVTTItalics        bool
 	WebVTTLine           string
 	WebVTTLines          int
 	WebVTTPosition       string
 	WebVTTRegionAnchor   string
 	WebVTTScroll         string
 	WebVTTSize           string
 	WebVTTStyles         []string
+	WebVTTTags           []WebVTTTag
 	WebVTTVertical       string
 	WebVTTViewportAnchor string
 	WebVTTWidth          string
 }
 
+type WebVTTTag struct {
+	Name       string
+	Annotation string
+	Classes    []string
+}
+
+func (t WebVTTTag) startTag() string {
+	if t.Name == "" {
+		return ""
+	}
+
+	s := t.Name
+	if len(t.Classes) > 0 {
+		s += "." + strings.Join(t.Classes, ".")
+	}
+
+	if t.Annotation != "" {
+		s += " " + t.Annotation
+	}
+
+	return "<" + s + ">"
+}
+
+func (t WebVTTTag) endTag() string {
+	if t.Name == "" {
+		return ""
+	}
+	return "</" + t.Name + ">"
+}
+
 func (sa *StyleAttributes) propagateSSAAttributes() {}
 
 func (sa *StyleAttributes) propagateSTLAttributes() {

webvtt.go

@@ -2,10 +2,10 @@ package astisub
 
 import (
 	"bufio"
-	"bytes"
 	"errors"
 	"fmt"
 	"io"
+	"log"
 	"regexp"
 	"sort"
 	"strconv"
@@ -33,7 +33,7 @@ var (
 	bytesWebVTTItalicEndTag            = []byte("</i>")
 	bytesWebVTTItalicStartTag          = []byte("<i>")
 	bytesWebVTTTimeBoundariesSeparator = []byte(webvttTimeBoundariesSeparator)
-	webVTTRegexpStartTag               = regexp.MustCompile(`(<v([\.\w]*)(.+?)>)`)
+	webVTTRegexpTag                    = regexp.MustCompile(`(</*\s*([^\.\s]+)(\.[^\s/]*)*\s*([^/]*)\s*/*>)`)
 	webVTTEscaper                      = strings.NewReplacer("&", "&amp;", "<", "&lt;")
 	webVTTUnescaper                    = strings.NewReplacer("&amp;", "&", "&lt;", "<")
 )
@@ -306,45 +306,65 @@ func parseTextWebVTT(i string) (o Line) {
 	// Create tokenizer
 	tr := html.NewTokenizer(strings.NewReader(i))
 
+	webVTTTagStack := make([]WebVTTTag, 0, 16)
+
 	// Loop
-	italic := false
 	for {
 		// Get next tag
 		t := tr.Next()
-
 		// Process error
 		if err := tr.Err(); err != nil {
 			break
 		}
 
 		switch t {
 		case html.EndTagToken:
-			// Parse italic
-			if bytes.Equal(tr.Raw(), bytesWebVTTItalicEndTag) {
-				italic = false
-				continue
+			// Pop the top of stack if we meet end tag
+			if len(webVTTTagStack) > 0 {
+				webVTTTagStack = webVTTTagStack[:len(webVTTTagStack)-1]
 			}
 		case html.StartTagToken:
-			// Parse voice name
-			if matches := webVTTRegexpStartTag.FindStringSubmatch(string(tr.Raw())); len(matches) > 3 {
-				if s := strings.TrimSpace(matches[3]); s != "" {
-					o.VoiceName = s
+			if matches := webVTTRegexpTag.FindStringSubmatch(string(tr.Raw())); len(matches) > 4 {
+				tagName := matches[2]
+
+				var classes []string
+				if matches[3] != "" {
+					classes = strings.Split(strings.Trim(matches[3], "."), ".")
+				}
+
+				annotation := ""
+				if matches[4] != "" {
+					annotation = strings.TrimSpace(matches[4])
+				}
+
+				if tagName == "v" {
+					if o.VoiceName == "" {
+						// Only get voicename of the first <v> appears in the line
+						o.VoiceName = annotation
+					} else {
+						// TODO: do something with other <v> instead of ignoring
+						log.Printf("astisub: found another voice name %q in %q. Ignore", annotation, i)
+					}
+					continue
 				}
-				continue
-			}
 
-			// Parse italic
-			if bytes.Equal(tr.Raw(), bytesWebVTTItalicStartTag) {
-				italic = true
-				continue
+				// Push the tag to stack
+				webVTTTagStack = append(webVTTTagStack, WebVTTTag{
+					Name:       tagName,
+					Classes:    classes,
+					Annotation: annotation,
+				})
 			}
+
 		case html.TextToken:
 			if s := strings.TrimSpace(string(tr.Raw())); s != "" {
 				// Get style attribute
 				var sa *StyleAttributes
-				if italic {
+				if len(webVTTTagStack) > 0 {
+					tags := make([]WebVTTTag, len(webVTTTagStack))
+					copy(tags, webVTTTagStack)
 					sa = &StyleAttributes{
-						WebVTTItalics: italic,
+						WebVTTTags: tags,
 					}
 					sa.propagateWebVTTAttributes()
 				}
@@ -545,19 +565,21 @@ func (li LineItem) webVTTBytes() (c []byte) {
 		color = cssColor(*li.InlineStyle.TTMLColor)
 	}
 
-	// Get italics
-	i := li.InlineStyle != nil && li.InlineStyle.WebVTTItalics
-
 	// Append
 	if color != "" {
 		c = append(c, []byte("<c."+color+">")...)
 	}
-	if i {
-		c = append(c, []byte("<i>")...)
+	if li.InlineStyle != nil {
+		for _, tag := range li.InlineStyle.WebVTTTags {
+			c = append(c, []byte(tag.startTag())...)
+		}
 	}
 	c = append(c, []byte(escapeWebVTT(li.Text))...)
-	if i {
-		c = append(c, []byte("</i>")...)
+	if li.InlineStyle != nil {
+		noTags := len(li.InlineStyle.WebVTTTags)
+		for i := noTags - 1; i >= 0; i-- {
+			c = append(c, []byte(li.InlineStyle.WebVTTTags[i].endTag())...)
+		}
 	}
 	if color != "" {
 		c = append(c, []byte("</c>")...)

webvtt_internal_test.go

@@ -96,39 +96,39 @@ func TestCueVoiceSpanRegex(t *testing.T) {
 	}{
 		{
 			give: `<v 中文> this is the content</v>`,
-			want: ` 中文`,
+			want: `中文`,
 		},
 		{
 			give: `<v 中文> this is the content`,
-			want: ` 中文`,
+			want: `中文`,
 		},
 		{
 			give: `<v.abc 中文> this is the content</v>`,
-			want: ` 中文`,
+			want: `中文`,
 		},
 		{
 			give: `<v.jp 言語の> this is the content`,
-			want: ` 言語の`,
+			want: `言語の`,
 		},
 		{
 			give: `<v.ko 언어> this is the content`,
-			want: ` 언어`,
+			want: `언어`,
 		},
 		{
 			give: `<v foo bar> this is the content`,
-			want: ` foo bar`,
+			want: `foo bar`,
 		},
 		{
 			give: `<v هذا عربي> this is the content`,
-			want: ` هذا عربي`,
+			want: `هذا عربي`,
 		},
 	}
 
 	for _, tt := range tests {
 		t.Run(tt.want, func(t *testing.T) {
-			results := webVTTRegexpStartTag.FindStringSubmatch(tt.give)
-			assert.True(t, len(results) == 4)
-			assert.Equal(t, tt.want, results[3])
+			results := webVTTRegexpTag.FindStringSubmatch(tt.give)
+			assert.True(t, len(results) == 5)
+			assert.Equal(t, tt.want, results[4])
 		})
 	}
 }

webvtt_test.go

@@ -162,3 +162,53 @@ Sentence with an & in the middle
 Sentence with an < in the middle
 `, b.String())
 }
+
+func TestWebVTTTags(t *testing.T) {
+	testData := `WEBVTT
+
+	00:01:00.000 --> 00:02:00.000
+	<u><i>Italic with underline text</i></u> some extra
+
+	00:02:00.000 --> 00:03:00.000
+	<lang en>English here</lang> <c.yellow.bg_blue>Yellow text on blue background</c>
+
+	00:03:00.000 --> 00:04:00.000
+	<v Joe><c.red><i>Joe's words are red in italic</i></c>
+
+	00:04:00.000 --> 00:05:00.000
+	<customed_tag.class1.class2>Text here</customed_tag>
+
+	00:05:00.000 --> 00:06:00.000
+	<v Joe>Joe says something</v> <v Bob>Bob says something</v>`
+
+	s, err := astisub.ReadFromWebVTT(strings.NewReader(testData))
+	require.NoError(t, err)
+
+	require.Len(t, s.Items, 5)
+
+	b := &bytes.Buffer{}
+	err = s.WriteToWebVTT(b)
+	require.NoError(t, err)
+	require.Equal(t, `WEBVTT
+
+1
+00:01:00.000 --> 00:02:00.000
+<u><i>Italic with underline text</i></u> some extra
+
+2
+00:02:00.000 --> 00:03:00.000
+<lang en>English here</lang> <c.yellow.bg_blue>Yellow text on blue background</c>
+
+3
+00:03:00.000 --> 00:04:00.000
+<v Joe><c.red><i>Joe's words are red in italic</i></c>
+
+4
+00:04:00.000 --> 00:05:00.000
+<customed_tag.class1.class2>Text here</customed_tag>
+
+5
+00:05:00.000 --> 00:06:00.000
+<v Joe>Joe says something Bob says something
+`, b.String())
+}