[
https://issues.apache.org/jira/browse/TIKA-4838?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18107550#comment-18107550
]
ASF GitHub Bot commented on TIKA-4838:
--------------------------------------
dschmidt commented on code in PR #3062:
URL: https://github.com/apache/tika/pull/3062#discussion_r3846531666
##########
tika-parsers/tika-parsers-standard/tika-parsers-standard-modules/tika-parser-audiovideo-module/src/main/java/org/apache/tika/parser/mp4/TikaMp4VideoHandler.java:
##########
@@ -76,6 +76,16 @@ private void extractFromSampleDescriptions(byte[] b) {
break;
}
int end = pos + (int) size;
+ String format = fourCc(b, pos + 4);
+ //protected streams replace the codec fourcc with 'encv' (ISO
common
+ //encryption) or 'drmi' (FairPlay); the original codec is in the
nested frma box
+ if ("encv".equals(format) || "drmi".equals(format)) {
+ String original = findOriginalFormat(b, pos +
VISUAL_ENTRY_SIZE, end, 0);
+ if (original != null) {
+ format = original;
+ }
+ }
+ tikaMetadata.set(Video.FORMAT, format);
Review Comment:
Stale: this reviewed an earlier revision. The frma-recovery path was removed
from the PR; protected streams now just report the entry fourcc, and the
protected-audio behavior that remains is covered by testDrmProtectedM4a.
##########
tika-parsers/tika-parsers-standard/tika-parsers-standard-modules/tika-parser-audiovideo-module/src/main/java/org/apache/tika/parser/mp4/TikaMp4VideoHandler.java:
##########
@@ -105,6 +115,38 @@ private static int findBtrtAverageBitRate(byte[] b, int
pos, int end) {
return 0;
}
+ private static final int MAX_BOX_DEPTH = 10;
+
+ /**
+ * Scans the child boxes of a protected ('encv'/'drmi') sample entry for
the
+ * 'frma' (original format) box nested in the 'sinf' protection scheme
+ * information box, and returns its codec fourcc, or null if there is none.
+ */
+ private static String findOriginalFormat(byte[] b, int pos, int end, int
depth) {
+ if (depth > MAX_BOX_DEPTH) {
+ return null;
+ }
+ while (pos + 8 <= end) {
+ long size = EndianUtils.getUIntBE(b, pos);
+ if (size < 8 || size > end - pos) {
+ return null;
+ }
+ int boxEnd = pos + (int) size;
+ String type = fourCc(b, pos + 4);
+ if ("frma".equals(type) && pos + 12 <= boxEnd) {
+ return fourCc(b, pos + 8);
+ }
+ if ("sinf".equals(type)) {
+ String original = findOriginalFormat(b, pos + 8, boxEnd, depth
+ 1);
+ if (original != null) {
+ return original;
+ }
+ }
+ pos = boxEnd;
+ }
+ return null;
+ }
Review Comment:
Stale: findOriginalFormat was removed along with the frma-recovery feature,
so there is nothing left to extract.
##########
tika-core/src/main/java/org/apache/tika/metadata/Video.java:
##########
@@ -37,4 +37,12 @@ public interface Video {
* reflects the last one.
*/
Property BITRATE = Property.internalInteger("video:bitrate");
+
+ /**
+ * The video track's four-character codec identifier from the MP4/QuickTime
+ * sample description (e.g. "avc1" for H.264, "hvc1"/"hev1" for HEVC). For
+ * protected streams this is the original codec from the 'frma' box.
Distinct
+ * from {@link XMPDM#VIDEO_COMPRESSOR}, the human-readable compressor name.
Review Comment:
The last-wins behavior is already documented in the javadoc and matches the
existing per-stream convention (e.g. Video.BITRATE).
##########
tika-core/src/main/java/org/apache/tika/metadata/Audio.java:
##########
@@ -84,6 +84,15 @@ public interface Audio {
*/
Property BITS_PER_SAMPLE =
Property.internalInteger("audio:bits-per-sample");
+ /**
+ * The audio track's four-character codec identifier from the MP4/QuickTime
+ * sample description (e.g. "mp4a" for AAC, "alac", "ac-3"), or the
protection
+ * scheme ("drms"/"enca") for protected streams, where {@link #HAS_DRM} is
also
+ * set. A per-stream value: with several audio tracks it reflects the last
one.
+ * Distinct from {@link XMPDM#AUDIO_COMPRESSOR}, the human-readable codec
name.
Review Comment:
Good catch, fixed in 9ec25dcca2 (javadoc and the handler comments now say
protected sample entry format).
##########
tika-core/src/main/java/org/apache/tika/metadata/Video.java:
##########
@@ -37,4 +37,13 @@ public interface Video {
* reflects the last one.
*/
Property BITRATE = Property.internalInteger("video:bitrate");
+
+ /**
+ * The video track's four-character codec identifier from the MP4/QuickTime
+ * sample description (e.g. "avc1" for H.264, "hvc1"/"hev1" for HEVC), or
the
+ * protection scheme ("encv"/"drmi") for protected streams. A per-stream
value:
+ * with several video tracks it reflects the last one. Distinct from
Review Comment:
Fixed in 9ec25dcca2, same wording change as on the audio side.
> Expose the MP4 audio and video track codecs (audio:format, video:format)
> ------------------------------------------------------------------------
>
> Key: TIKA-4838
> URL: https://issues.apache.org/jira/browse/TIKA-4838
> Project: Tika
> Issue Type: New Feature
> Reporter: Dominik Schmidt
> Priority: Major
>
> The MP4/QuickTime parser does not expose the per-track codec identifiers.
> TikaMp4SoundHandler and TikaMp4VideoHandler read the sample-entry format
> fourcc but only use it for other purposes (audio: DRM detection; video:
> nothing).
> Add audio:format and video:format, each carrying the four-character codec
> identifier from the track's sample description (audio mp4a/alac/ac-3, video
> avc1/hev1). For protected streams (drms/enca, encv/drmi) the original codec
> is recovered from the nested frma box.
> The existing xmpDM:audioCompressor only fires for audio-typed files and
> carries the container major brand, and xmpDM:videoCompressor carries the
> compressor name (encoder string, e.g. "Lavc… libx264"), so neither gives the
> per-track codec fourcc for video/* files.
> Rounds out the per-track metadata added in TIKA-4779, TIKA-4800 and TIKA-4802.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
