[
https://issues.apache.org/jira/browse/CAMEL-14952?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17152661#comment-17152661
]
Zoran Regvart commented on CAMEL-14952:
---------------------------------------
I'd try to make the configuration as simple as possible. [~Aemie] is it
possible to omit the values for variables, or that would not work. We release
quite often and with the approach you're suggesting we would need to create PRs
against Algolia repository every time we release a new version. I'm unsure
about indexing the {{/categories}}, that seems redundant in addition to
indexing {{/blog}}.
I guess it's somewhat difficult to ascertain which configuration yields better
search index, I think we need to define a set of criteria/tests to check. For
example:
* If a documentation page has multiple versions, search results show what
version a search hit belongs to
* In addition to the title we can see abstract of the page text
* Each search hit belongs to a category (blog, user manual, component
reference, camel-k, camel-quarkus, camel-spring-boot, camel-karaf,
camel-kafka-connector)
* Search hits contain breadcrumbs (Camel Components, Components, AWS, AWS
Elastic Container Service (ECS))
* ...
Let's see about compiling a list so we can determine what configuration is an
improvement.
For the time being, we can start with the suggestions that [~aashnajena] did on
the [PR#423|https://github.com/apache/camel-website/pull/423], and make
subsequent PRs to improve it. How does that sound?
> Better search on the website
> ----------------------------
>
> Key: CAMEL-14952
> URL: https://issues.apache.org/jira/browse/CAMEL-14952
> Project: Camel
> Issue Type: Improvement
> Components: website
> Reporter: Zoran Regvart
> Priority: Major
> Attachments:
> BH4D9OD16A_apache_camel_20200608-20200614_no_result_searches.csv,
> List_Of_Crawled_Pages_by_DocSearch.txt, apache_demo.json, camel.json,
> getbootstrap-searchresult.png, image-2020-06-13-14-39-08-776.png,
> list_of_crawled_pages.txt, sitemap-camel.png
>
>
> We use [Algolia|http://algolia.com/] for the search functionality on the
> website using their [free plan|https://www.algolia.com/for-open-source/] for
> Open Source projects. The index is built by Algolia's crawler using the
> [DocSearch|https://docsearch.algolia.com/].
> When this was done we built our own UI on top of Algolia JavaScript API, as
> one if requirements is that clients use Algolia's JavaScript clients. We did
> not use Algolia UI as at that point it was rather large JavaScript dependency
> to add and it would slow down the loading of the website.
> We also have some [initial
> work|https://github.com/apache/camel-website/pull/74] on creating our own
> Algolia index at build time.
> The current search doesn't seem to index the whole website, some results
> don't appear in the search, looks like most of the content from Antora is not
> indexed: trying to search for {{removeHeader}}, the [FAQ
> entry|https://camel.apache.org/manual/latest/faq/how-to-avoid-sending-some-or-all-message-headers.html]
> is not found. There's also a list of failed searches on the Algolia
> dashboard we can use to benchmark the search.
> What we need is to build the search index over the whole content. Approach
> taken in [#74|https://github.com/apache/camel-website/pull/74] is good start
> for Hugo generated content. We need to expand that to Antora built content as
> well.
> This search index would be built at the website build time and would include
> both Hugo and Algolia content in the same file or possibly in several files
> if we use multi-index search. More on how indexes are built can be seen in
> the [Algolia
> documentation|https://www.algolia.com/doc/guides/sending-and-managing-data/prepare-your-data/].
> We need to figure out what data to send and how to integrate this with
> Antora, for Hugo we have a good idea from
> [#74|https://github.com/apache/camel-website/pull/74], importantly the
> structure needs to be the same. One good source of inspiration on building
> the index for Antora content is in the [Lunr.js
> integration|https://github.com/Mogztter/antora-lunr].
> We need to build the index with the search UI in mind, i.e. the index needs
> to contain the data we wish to present in the UI as well as enough content
> for Algolia to be able to use the content to perform search. So starting with
> a mockup of what we wish to present/utilize in the search UI and deriving the
> data structure for the index from that would be a good start.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
