[
https://issues.apache.org/jira/browse/TIKA-4743?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18085219#comment-18085219
]
ASF GitHub Bot commented on TIKA-4743:
--------------------------------------
Copilot commented on code in PR #2860:
URL: https://github.com/apache/tika/pull/2860#discussion_r3334247883
##########
docs/publish-docs.sh:
##########
@@ -83,6 +83,11 @@ cp target/site/404.html "${DOCS_DIR}/"
# Lunr index lives next to _/ (one level above docs/), since HTML uses
../../search-index.js.
# Remove the stale copy from its old publish/docs/ location left by earlier
runs.
rm -f "${DOCS_DIR}/search-index.js"
-cp target/site/search-index.js "${PUBLISH_DIR}/"
+# Rewrite URLs in the search index from /tika/X.Y.Z/... (Antora's component-
+# prefixed publish path) to /docs/X.Y.Z/... (the deployed layout). The HTML
+# pages and sitemap.xml above are similarly flattened; without this rewrite,
+# clicking a search result lands on https://tika.apache.org/tika/... which
+# 404s. See TIKA-4743.
+sed 's|"url":"/tika/|"url":"/docs/|g' target/site/search-index.js >
"${PUBLISH_DIR}/search-index.js"
Review Comment:
Because of shell redirection semantics, this command will truncate/create
`${PUBLISH_DIR}/search-index.js` before `sed` runs. If
`target/site/search-index.js` is missing or `sed` errors, the published search
index can be left empty even though the script exits non-zero. Consider writing
to a temporary file first (and explicitly failing with a clear message if the
input file is missing) so a failed rewrite can’t corrupt an existing published
index.
> tika-4.0.0-alpha1 - documentation site
> --------------------------------------
>
> Key: TIKA-4743
> URL: https://issues.apache.org/jira/browse/TIKA-4743
> Project: Tika
> Issue Type: Bug
> Affects Versions: 4.0.0
> Reporter: Adrian Bird
> Priority: Minor
>
> A couple of comments about the [Apache Tika Documentation
> site|https://tika.apache.org/docs/4.0.0-SNAPSHOT/index.html].
> * I was expecting the menu items on the left to open and show lower level
> pages where they have a triangle, but they don't. This does make it tricky to
> navigate because you have to do it via the links on the pages themselves.
> * Search - trying to use the search feature pops up a small box with the
> text 'Loading search index...' (in white on the light gray background means
> it isn't really visible) and nothing happens i.e. no search results are
> displayed.
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
