Closing the loop here all the PRs are merged. Thanks Max and everyone for the review!
https://github.com/apache/iceberg/pull/16368 https://github.com/apache/iceberg/pull/16371 https://github.com/apache/iceberg/pull/16398 On Mon, May 18, 2026 at 10:00 AM Kevin Liu <[email protected]> wrote: > Thanks Max for the fix [1], we were able to remove the previous doc > versions from the search index, and it's already live! > > While working on this, we found two follow-up issues: > > 1. Duplicate results -- There's another duplicate issue causing 2 > identical results to appear. I have a fix ready here [2]. > > 2. "Nightly" vs. "Latest" docs -- Both the "Docs" tab on > https://iceberg.apache.org/ and the search index currently point to > "nightly" (the most up-to-date documentation based on the main branch). > This could be confusing if it includes changes that haven't been released > yet. I'd suggest updating both to link to "latest" (the most recent release > version) instead. Thoughts? > > Best, > Kevin Liu > > > [1] https://github.com/apache/iceberg/pull/16368 > [2] https://github.com/apache/iceberg/pull/16371 > > On Sat, May 16, 2026 at 9:49 AM Kevin Liu <[email protected]> wrote: > >> Thanks for looking into this, Max. Yes please create a PR, I'll take a >> look. This is a great quick win for our docs :) >> >> >> On Fri, May 15, 2026 at 7:18 PM Max Konstantinov < >> [email protected]> wrote: >> >>> > mkdocs-exclude-search ( >>> https://github.com/chrieke/mkdocs-exclude-search) seems promising >>> >>> Hey Kevin, >>> >>> I tried the mkdocs-exclude-search approach and it works: >>> >>> mkdocs-exclude-search==0.6.6 >>> >>> - exclude-search: >>> exclude: >>> - 'docs/latest/*' >>> - 'docs/1*/*' >>> exclude_unreferenced: true >>> >>> Config above leaves only nightly present in the index. Should I create a >>> PR? >>> >>> > I've looked into this before, the best solution I found is to migrate >>> over to `mike` (see the tracking issue >>> https://github.com/apache/iceberg/issues/14521) It's a lot of work >>> though and I didn't want to risk breaking the site. >>> >>> Let's decide if we want a quick solution for search as >>> described above and then I can look into this large task with a version >>> selector? >>> >>> >>> On Thu, May 14, 2026 at 9:59 AM Kevin Liu <[email protected]> wrote: >>> >>>> mkdocs-exclude-search (https://github.com/chrieke/mkdocs-exclude-search) >>>> seems promising >>>> >>>> We can try something like this to exclude all the versioned paths: >>>> ``` >>>> plugins: >>>> - search >>>> - exclude-search: >>>> exclude: >>>> - "*/[0-9]*/**" >>>> ``` >>>> >>>> Note that nightly and latest are symlinks. >>>> >>>> >>>> >>>> >>>> On Thu, May 14, 2026 at 9:54 AM Kevin Liu <[email protected]> >>>> wrote: >>>> >>>>> +1 to only indexing the latest/nightly >>>>> mkdoc is using the `search` plugin [1]. And like you mentioned, we can >>>>> use its `exclude ` property to configure exclusion paths. >>>>> >>>>> >>>>> > Is there a way we can add a Version dropdown to the search? >>>>> "Latests, 1.11, etc ...." >>>>> >>>>> I've looked into this before, the best solution I found is to migrate >>>>> over to `mike` (see the tracking issue >>>>> https://github.com/apache/iceberg/issues/14521) It's a lot of work >>>>> though and I didn't want to risk breaking the site. >>>>> >>>>> Best, >>>>> Kevin Liu >>>>> >>>>> >>>>> [1] >>>>> https://github.com/apache/iceberg/blob/87a7e4b13470b238e86872e5b194b80c3a89a80f/site/mkdocs.yml#L51 >>>>> [2] >>>>> https://squidfunk.github.io/mkdocs-material/plugins/search/#meta.search.exclude >>>>> >>>>> >>>>> >>>>> On Thu, May 14, 2026 at 9:33 AM Max Konstantinov < >>>>> [email protected]> wrote: >>>>> >>>>>> That was my first idea too. I did some research and it seems it will >>>>>> be more complicated to implement and might require an overhaul of the >>>>>> site >>>>>> build process or/and site structure. Let me look into that again and get >>>>>> back with an overview of potential changes. >>>>>> >>>>>> On Thu, May 14, 2026 at 9:13 AM Russell Spitzer < >>>>>> [email protected]> wrote: >>>>>> >>>>>>> Is there a way we can add a Version dropdown to the search? >>>>>>> "Latests, 1.11, etc ...." >>>>>>> >>>>>>> On Thu, May 14, 2026 at 10:26 AM Max Konstantinov < >>>>>>> [email protected]> wrote: >>>>>>> >>>>>>>> I looked into options for resolving that and the easiest solution I >>>>>>>> found is to exclude all md files from search for all Icerbrg >>>>>>>> documentation >>>>>>>> versions, except for the just one, nightly probably being the best >>>>>>>> candidate as it is a defult landing version of documentation. That can >>>>>>>> be >>>>>>>> achieved by adding a mkdocs-material's YAML frontmatter property to md >>>>>>>> files during the site build process: >>>>>>>> >>>>>>>> --- >>>>>>>> <https://squidfunk.github.io/mkdocs-material/setup/setting-up-site-search/#__codelineno-6-2>search: >>>>>>>> >>>>>>>> <https://squidfunk.github.io/mkdocs-material/setup/setting-up-site-search/#__codelineno-6-3> >>>>>>>> exclude: true >>>>>>>> <https://squidfunk.github.io/mkdocs-material/setup/setting-up-site-search/#__codelineno-6-4>--- >>>>>>>> >>>>>>>> >>>>>>>> >>>>>>>> On Thu, May 14, 2026 at 1:39 AM Robin Moffatt via dev < >>>>>>>> [email protected]> wrote: >>>>>>>> >>>>>>>>> I agree, the current behaviour is not ideal for the end-user. The >>>>>>>>> results don't indicate the version either so you end up with literal >>>>>>>>> duplicates: >>>>>>>>> [image: CleanShot 2026-05-14 at [email protected]] >>>>>>>>> >>>>>>>>> On Thu, 14 May 2026 at 07:17, Max Konstantinov < >>>>>>>>> [email protected]> wrote: >>>>>>>>> >>>>>>>>>> Hi all, >>>>>>>>>> >>>>>>>>>> >>>>>>>>>> Problem statement >>>>>>>>>> If you open https://iceberg.apache.org/ >>>>>>>>>> <https://urldefense.com/v3/__https://iceberg.apache.org/__;!!Ayb5sqE7!oRqHG1ZbjiQCzTf4geY7acwJ5T348ItdBhCxA93gVPFAC7C6I6nQHtkpxvkEN1zEBSfkq7BdqNaJnYBf6wfaUGD1HAZo$> >>>>>>>>>> and >>>>>>>>>> use a search bar to look for anything, ex: "rewrite", you get >>>>>>>>>> duplicated >>>>>>>>>> results because results(ex: rewrite_table_path) are found multiple >>>>>>>>>> times, >>>>>>>>>> once for each version of Iceberg documentation(1.4 to 1.10), all >>>>>>>>>> these >>>>>>>>>> versions are present on the website under the previous section in the >>>>>>>>>> navigation tab and all of them are indexed. >>>>>>>>>> >>>>>>>>>> Question >>>>>>>>>> Is that a bug and does it need to be addressed? I have a >>>>>>>>>> suggestion on how to approach this. >>>>>>>>>> >>>>>>>>>> Notes >>>>>>>>>> Iceberg's site build uses mkdocs and mkdocs-material which in >>>>>>>>>> turn utilize lunrjs for search. >>>>>>>>>> >>>>>>>>>> >>>>>>>>>> Max. >>>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>>
