[
https://issues.apache.org/jira/browse/TIKA-4852?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18109437#comment-18109437
]
ASF GitHub Bot commented on TIKA-4852:
--------------------------------------
Copilot commented on code in PR #3093:
URL: https://github.com/apache/tika/pull/3093#discussion_r3886823697
##########
tika-parsers/tika-parsers-standard/tika-parsers-standard-modules/tika-parser-miscoffice-module/src/main/java/org/apache/tika/parser/epub/EpubParser.java:
##########
@@ -442,6 +448,10 @@ private void handleEmbedded(ZipFile zipFile, String
relativePath, HRefMediaPair
embeddedMetadata.set(HttpHeaders.CONTENT_TYPE,
hRefMediaPair.media);
}
embeddedMetadata.set(TikaCoreProperties.RESOURCE_NAME_KEY, fullPath);
+ if (cover) {
+ embeddedMetadata.set(TikaCoreProperties.EMBEDDED_RESOURCE_TYPE,
+
TikaCoreProperties.EmbeddedResourceType.THUMBNAIL.toString());
+ }
Review Comment:
`handleEmbedded` can be invoked when `hRefMediaPair.media` is null (e.g.,
OPF item missing `media-type`; `shouldHandleEmbedded(null)` returns true), but
later this method unconditionally calls `hRefMediaPair.media.contains("font")`,
which will throw a NullPointerException. Guard the `contains` check to keep
embedded extraction robust for incomplete/invalid OPFs.
> EpubParser: mark the cover image as a THUMBNAIL embedded resource
> -----------------------------------------------------------------
>
> Key: TIKA-4852
> URL: https://issues.apache.org/jira/browse/TIKA-4852
> Project: Tika
> Issue Type: Improvement
> Reporter: Dominik Schmidt
> Priority: Major
>
> EpubParser emits every image of the package as an untyped embedded document,
> in manifest order. The OPF names the cover, though: EPUB 3 via
> properties="cover-image" on the manifest item, EPUB 2 via <meta name="cover"
> content="<item id>"/>. Neither is read (OPFParser has no notion of a cover),
> so a client that wants the representative image of an EPUB has to guess.
> Proposal: resolve the cover item from the OPF (properties first, then the
> meta fallback) and emit it with tk:embedded-resource-type=THUMBNAIL, like the
> preview image of the other container formats (ODF, OOXML, iWork, GeoGebra,
> audio cover art as of TIKA-4850). Other images keep their current handling.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
