pitchfork is a unicorn fork with some things removed and some things added.
We don't need any of these things so it should be an easy replace.
There is no worker-killer for pitchfork, it is however trivial to implement ourselves.
unicorn is pretty much dead. The author implies as much in the various readme updates
he made. It also doesn't work with rack 3 and I doubt a new release will even be made.
If it will, it highly likely won't be on rubygems.
This removes the need for all the ruby-lsp setup stuff.
ruby-lsp will try to install the debug gem which contains native code, add that to the gemfile so it doesn't cause trouble.
I'm probably going to remove the rubocop/test compose services at some point. With devcontainers you are just able to run the commands directly.
It just works.
Requires a bit of a workaround for the .rubocop.yml env setup currently in use but thats alright.
The companion ruby-lsp-rails doesn't do much right now but I imagine that to become quite powerful in the future.
The author has passed away and some functionality is broken with recent
ruby versions already. We only use a subset of it, so it's easy to replace.
Some places:
* Didn't even use it
* Used it but the method was already cached through memcached
* Just don't need it
Some places I question if it is needed but I'll just leave them there
since I don't want to benchmark
Rails minor updates are basically like major ones, without removing
deprecations but all the potential bugs that come with a major update.
Make a concious decision to update to these new versions.
Run the following commands to import data into OpenSearch:
Post.document_store.create_index!
PostVersion.document_store.create_index!
Post.document_store.import
PostVersion.document_store.import
* Add opensearch client
* Add url param to force use opensearch
* Switch import method over to opensearch
* Index to opensearch as well
* Add option to gradually roll out os
* Index os with separate queue
* Move os post creation to job as well
* Exclusively use the OpenSearch client
* Stop enqueuing OsIndexUpdateJob
* Remove remaining elasticsearch code
Bump faraday since v2 is no longer blocked
There is no counterpart for OpenSearch.
`elasticsearch-model` needs to go too but that is more involved than
just removing something that only facilitates logging in dev.
This was responsible for
* displaying the time elasticsearch queries took per request
* logging the search parameters being sent to elasticsearch per query
The implementation isn't that complex and I could roll something myself but
considering that I haven't sought that information out by now I'll just don't.
Logging queries is trivial since we craft these manually ourselves.
Notes are formatted using dtext only. Even though the existance of this
class suggests that some html elements are allowed, it doesn't actually do anything.
Link relativization also didn't work, same with dtext colors.
