#4306 - Extract tags from SubRip subtitles, add support for alignment tags based on SSA v4+

library/core/src/main/java/com/google/android/exoplayer2/text/subrip/SubripDecoder.java

@@ -15,7 +15,9 @@
  */
 package com.google.android.exoplayer2.text.subrip;
 
+import android.support.annotation.StringDef;
 import android.text.Html;
+import android.text.Layout;
 import android.text.Spanned;
 import android.text.TextUtils;
 import android.util.Log;
@@ -23,7 +25,11 @@ import com.google.android.exoplayer2.text.Cue;
 import com.google.android.exoplayer2.text.SimpleSubtitleDecoder;
 import com.google.android.exoplayer2.util.LongArray;
 import com.google.android.exoplayer2.util.ParsableByteArray;
+
+import java.lang.annotation.Retention;
+import java.lang.annotation.RetentionPolicy;
 import java.util.ArrayList;
+import java.util.List;
 import java.util.regex.Matcher;
 import java.util.regex.Pattern;
 
@@ -38,6 +44,33 @@ public final class SubripDecoder extends SimpleSubtitleDecoder {
   private static final Pattern SUBRIP_TIMING_LINE =
       Pattern.compile("\\s*(" + SUBRIP_TIMECODE + ")\\s*-->\\s*(" + SUBRIP_TIMECODE + ")?\\s*");
 
+  private static final Pattern SUBRIP_TAG_PATTERN = Pattern.compile("\\{\\\\.*?\\}");
+  private static final String SUBRIP_ALIGNMENT_TAG = "\\{\\\\an[1-9]\\}";
+
+  private static final float DEFAULT_START_FRACTION = 0.08f;
+  private static final float DEFAULT_END_FRACTION = 1 - DEFAULT_START_FRACTION;
+  private static final float DEFAULT_MID_FRACTION = 0.5f;
+
+  @Retention(RetentionPolicy.SOURCE)
+  @StringDef({
+      ALIGN_BOTTOM_LEFT, ALIGN_BOTTOM_MID, ALIGN_BOTTOM_RIGHT,
+      ALIGN_MID_LEFT, ALIGN_MID_MID, ALIGN_MID_RIGHT,
+      ALIGN_TOP_LEFT, ALIGN_TOP_MID, ALIGN_TOP_RIGHT
+  })
+
+  private @interface SubRipTag {}
+
+  // Possible valid alignment tags based on SSA v4+ specs
+  private static final String ALIGN_BOTTOM_LEFT  = "{\\an1}";
+  private static final String ALIGN_BOTTOM_MID   = "{\\an2}";
+  private static final String ALIGN_BOTTOM_RIGHT = "{\\an3}";
+  private static final String ALIGN_MID_LEFT     = "{\\an4}";
+  private static final String ALIGN_MID_MID      = "{\\an5}";
+  private static final String ALIGN_MID_RIGHT    = "{\\an6}";
+  private static final String ALIGN_TOP_LEFT     = "{\\an7}";
+  private static final String ALIGN_TOP_MID      = "{\\an8}";
+  private static final String ALIGN_TOP_RIGHT    = "{\\an9}";
+
   private final StringBuilder textBuilder;
 
   public SubripDecoder() {
@@ -95,8 +128,36 @@ public final class SubripDecoder extends SimpleSubtitleDecoder {
         textBuilder.append(currentLine.trim());
       }
 
-      Spanned text = Html.fromHtml(textBuilder.toString());
-      cues.add(new Cue(text));
+      // Extract tags
+      SubtitleTagResult tagResult = extractTags(textBuilder);
+      Spanned text = Html.fromHtml(tagResult.cue);
+
+      Cue cue = null;
+
+      // Check if tags are present
+      if (tagResult.tags.length > 0) {
+
+        boolean alignTagFound = false;
+
+        // At end of this loop the clue must be created with the applied tags
+        for (String tag : tagResult.tags) {
+
+          // Check if the tag is an alignment tag
+          if (tag.matches(SUBRIP_ALIGNMENT_TAG)) {
+
+            // Based on the specs, in case of the alignment tags only the first appearance counts
+            if (alignTagFound) continue;
+            alignTagFound = true;
+
+            AlignmentResult alignmentResult = getAlignmentValues(tag);
+            cue = new Cue(text, Layout.Alignment.ALIGN_NORMAL, alignmentResult.line, Cue.LINE_TYPE_FRACTION,
+                alignmentResult.lineAnchor, alignmentResult.position, alignmentResult.positionAnchor, Cue.DIMEN_UNSET);
+          }
+        }
+      }
+
+      cues.add(cue == null ? new Cue(text) : cue);
+
       if (haveEndTimecode) {
         cues.add(null);
       }
@@ -108,6 +169,111 @@ public final class SubripDecoder extends SimpleSubtitleDecoder {
     return new SubripSubtitle(cuesArray, cueTimesUsArray);
   }
 
+  /**
+   * Extracts the tags from the given {@code cue}
+   * The pattern that is used to extract the tags is specified in SSA v4+ specs and
+   * has the following form: "{\...}".
+   * <p>
+   * "All override codes appear within braces {}"
+   * "All override codes are always preceded by a backslash \"
+   *
+   * @param cue Cue text
+   * @return {@link SubtitleTagResult} that holds new cue and also the extracted tags
+   */
+  private SubtitleTagResult extractTags(StringBuilder cue) {
+    StringBuilder cueCopy = new StringBuilder(cue.toString());
+    List<String> tags = new ArrayList<>();
+
+    int replacedCharacters = 0;
+
+    Matcher matcher = SUBRIP_TAG_PATTERN.matcher(cue.toString());
+    while (matcher.find()) {
+      String tag = matcher.group();
+      tags.add(tag);
+      cueCopy.replace(matcher.start() - replacedCharacters, matcher.end() - replacedCharacters, "");
+      replacedCharacters += tag.length();
+    }
+
+    return new SubtitleTagResult(tags.toArray(new String[tags.size()]), cueCopy.toString());
+  }
+
+  /**
+   * Match the alignment tag and calculate the line, position, position anchor accordingly
+   *
+   * Based on SSA v4+ specs the alignment tag can have the following form: {\an[1-9},
+   * where the number specifies the direction (based on the numpad layout).
+   * Note. older SSA scripts may contain tags like {\a1[1-9]} but these are based on
+   * other direction rules, but multiple sources says that these are deprecated, so no support here either
+   *
+   * @param tag Alignment tag
+   * @return {@link AlignmentResult} that holds the line, position, position anchor values
+   */
+  private AlignmentResult getAlignmentValues(String tag) {
+    // Default values used for positioning the subtitle in case of align tags
+    float line = DEFAULT_END_FRACTION, position = DEFAULT_MID_FRACTION;
+    @Cue.AnchorType int positionAnchor = Cue.ANCHOR_TYPE_MIDDLE;
+    @Cue.AnchorType int lineAnchor = Cue.ANCHOR_TYPE_END;
+
+    switch (tag) {
+      case ALIGN_BOTTOM_LEFT:
+        line = DEFAULT_END_FRACTION;
+        position = DEFAULT_START_FRACTION;
+        positionAnchor = Cue.ANCHOR_TYPE_START;
+        lineAnchor = Cue.ANCHOR_TYPE_END;
+        break;
+      case ALIGN_BOTTOM_MID:
+        line = DEFAULT_END_FRACTION;
+        position = DEFAULT_MID_FRACTION;
+        positionAnchor = Cue.ANCHOR_TYPE_MIDDLE;
+        lineAnchor = Cue.ANCHOR_TYPE_END;
+        break;
+      case ALIGN_BOTTOM_RIGHT:
+        line = DEFAULT_END_FRACTION;
+        position = DEFAULT_END_FRACTION;
+        positionAnchor = Cue.ANCHOR_TYPE_END;
+        lineAnchor = Cue.ANCHOR_TYPE_END;
+        break;
+      case ALIGN_MID_LEFT:
+        line = DEFAULT_MID_FRACTION;
+        position = DEFAULT_START_FRACTION;
+        positionAnchor = Cue.ANCHOR_TYPE_START;
+        lineAnchor = Cue.ANCHOR_TYPE_MIDDLE;
+        break;
+      case ALIGN_MID_MID:
+        line = DEFAULT_MID_FRACTION;
+        position = DEFAULT_MID_FRACTION;
+        positionAnchor = Cue.ANCHOR_TYPE_MIDDLE;
+        lineAnchor = Cue.ANCHOR_TYPE_MIDDLE;
+        break;
+      case ALIGN_MID_RIGHT:
+        line = DEFAULT_MID_FRACTION;
+        position = DEFAULT_END_FRACTION;
+        positionAnchor = Cue.ANCHOR_TYPE_END;
+        lineAnchor = Cue.ANCHOR_TYPE_MIDDLE;
+        break;
+      case ALIGN_TOP_LEFT:
+        line = DEFAULT_START_FRACTION;
+        position = DEFAULT_START_FRACTION;
+        positionAnchor = Cue.ANCHOR_TYPE_START;
+        lineAnchor = Cue.ANCHOR_TYPE_START;
+        break;
+      case ALIGN_TOP_MID:
+        line = DEFAULT_START_FRACTION;
+        position = DEFAULT_MID_FRACTION;
+        positionAnchor = Cue.ANCHOR_TYPE_MIDDLE;
+        lineAnchor = Cue.ANCHOR_TYPE_START;
+        break;
+      case ALIGN_TOP_RIGHT:
+        line = DEFAULT_START_FRACTION;
+        position = DEFAULT_END_FRACTION;
+        positionAnchor = Cue.ANCHOR_TYPE_END;
+        lineAnchor = Cue.ANCHOR_TYPE_START;
+        break;
+    }
+
+    return new AlignmentResult(positionAnchor, position, lineAnchor, line);
+  }
+
   private static long parseTimecode(Matcher matcher, int groupOffset) {
     long timestampMs = Long.parseLong(matcher.group(groupOffset + 1)) * 60 * 60 * 1000;
     timestampMs += Long.parseLong(matcher.group(groupOffset + 2)) * 60 * 1000;
@@ -116,4 +282,35 @@ public final class SubripDecoder extends SimpleSubtitleDecoder {
     return timestampMs * 1000;
   }
 
+  /**
+   * Class that holds the tags, new clue after the tag extraction
+   */
+  private static final class SubtitleTagResult {
+    public final String[] tags;
+    public final String cue;
+
+    public SubtitleTagResult(String[] tags, String cue) {
+      this.tags = tags;
+      this.cue = cue;
+    }
+  }
+
+  /**
+   * Class that holds the parsed and mapped alignment values (such as line,
+   * position and anchor type of line)
+   */
+  private static final class AlignmentResult {
+
+    public @Cue.AnchorType int positionAnchor;
+    public @Cue.AnchorType int lineAnchor;
+    public float position, line;
+
+    public AlignmentResult(@Cue.AnchorType int positionAnchor, float position, @Cue.AnchorType int lineAnchor, float line) {
+      this.positionAnchor = positionAnchor;
+      this.position = position;
+      this.line = line;
+      this.lineAnchor = lineAnchor;
+    }
+  }
+
 }

library/core/src/test/assets/subrip/typical_with_tags

+1
+00:00:00,000 --> 00:00:01,234
+This is {\an1} the first subtitle.
+
+2
+00:00:02,345 --> 00:00:03,456
+This is the second subtitle.
+Second {\ an 2} subtitle with second line.
+
+3
+00:00:04,567 --> 00:00:08,901
+This {\an2} is the third {\ tag} subtitle.
+
+4
+00:00:09,567 --> 00:00:12,901
+This { \an2} is the fourth subtitle.
+
+5
+00:00:013,567 --> 00:00:14,901
+This {\an2} is the fifth subtitle with multiple {\xyz} valid {\qwe} tags.

library/core/src/test/java/com/google/android/exoplayer2/text/subrip/SubripDecoderTest.java

@@ -36,6 +36,7 @@ public final class SubripDecoderTest {
   private static final String TYPICAL_MISSING_SEQUENCE = "subrip/typical_missing_sequence";
   private static final String TYPICAL_NEGATIVE_TIMESTAMPS = "subrip/typical_negative_timestamps";
   private static final String TYPICAL_UNEXPECTED_END = "subrip/typical_unexpected_end";
+  private static final String TYPICAL_WITH_TAGS = "subrip/typical_with_tags";
   private static final String NO_END_TIMECODES_FILE = "subrip/no_end_timecodes";
 
   @Test
@@ -154,6 +155,27 @@ public final class SubripDecoderTest {
         .isEqualTo("Or to the end of the media.");
   }
 
+  @Test
+  public void testDecodeCueWithTag() throws IOException{
+    SubripDecoder decoder = new SubripDecoder();
+    byte[] bytes = TestUtil.getByteArray(RuntimeEnvironment.application, TYPICAL_WITH_TAGS);
+    SubripSubtitle subtitle = decoder.decode(bytes, bytes.length, false);
+    assertThat(subtitle.getCues(subtitle.getEventTime(0)).get(0).text.toString())
+        .isEqualTo("This is the first subtitle.");
+    assertThat(subtitle.getCues(subtitle.getEventTime(2)).get(0).text.toString())
+        .isEqualTo("This is the second subtitle.\nSecond subtitle with second line.");
+    assertThat(subtitle.getCues(subtitle.getEventTime(4)).get(0).text.toString())
+        .isEqualTo("This is the third subtitle.");
+
+    // Based on the SSA v4+ specs the curly bracket must be followed by a backslash, so this is
+    // not a valid tag (won't be parsed / replaced)
+    assertThat(subtitle.getCues(subtitle.getEventTime(6)).get(0).text.toString())
+        .isEqualTo("This { \\an2} is the fourth subtitle.");
+
+    assertThat(subtitle.getCues(subtitle.getEventTime(8)).get(0).text.toString())
+        .isEqualTo("This is the fifth subtitle with multiple valid tags.");
+  }
+
   private static void assertTypicalCue1(SubripSubtitle subtitle, int eventIndex) {
     assertThat(subtitle.getEventTime(eventIndex)).isEqualTo(0);
     assertThat(subtitle.getCues(subtitle.getEventTime(eventIndex)).get(0).text.toString())