TheAnachronism/docspell
1
0
Fork 0
mirror of https://github.com/TheAnachronism/docspell.git synced 2025-06-22 10:28:27 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add websockets and notify frontend when an item is processed

Browse Source
This commit is contained in:
eikek
2021-11-06 21:32:07 +01:00
parent f38d520a1d
commit 3e58d97f72
17 changed files with 243 additions and 114 deletions
Download Patch File Download Diff File Expand all files Collapse all files

24
modules/pubsub/naive/src/main/scala/docspell/pubsub/naive/NaivePubSub.scala
View File

@ -26,30 +26,30 @@ import org.http4s.client.dsl.Http4sClientDsl
import org.http4s.dsl.Http4sDsl
import org.http4s.{HttpRoutes, Uri}
/** A pubsub implementation that can be used across machines, but uses a rather
* inefficient protocol. The reason is to make it work with the current setup, i.e. not
* requiring to add another complex piece of software to the mix, like Kafka or RabbitMQ.
/** A pubsub implementation that can be used across machines, using a rather inefficient
* but simple protocol. It can therefore work with the current setup, i.e. not requiring
* to add another complex piece of software to the mix, like Kafka or RabbitMQ.
*
* However, the api should allow to be used on top of such a tool. This implementation
* can be used in a personal setting, where there are only a few nodes.
*
* How it works:
* How it works: Each node has a set of local subscribers and a http endpoint. If it
* publishes a message, it notifies all local subscribers and sends out a json message to
* all endpoints that are registered for this topic. If it receives a messagen through
* its endpoint, it notifies all local subscribers.
*
* It is build on the `Topic` class from fs2.concurrent. A map of a topic name to such a
* It is build on the `Topic` class from fs2.concurrent. A map of the name to such a
* `Topic` instance is maintained. To work across machines, the database is used as a
* synchronization point. Each subscriber must provide a http api and so its "callback"
* URL is added into the database to the list of remote subscribers.
* synchronization point. Each node must provide a http api and so its "callback" URL is
* added into the database associated to a topic name.
*
* When publishing a message, the message can be published to the internal fs2 topic.
* Then all URLs to this topic name are looked up in the database and the message is
* POSTed to each URL as JSON. The endpoint of each machine takes this message and
* publishes it to its own internal fs2.concurrent.Topic instance.
*
* Obviously, there are drawbacks: it is slow, because the messages go through http and
* connections must be opened/closed etc and the database is hit as well. Then it doesn't
* scale to lots of machines and messages. The upside is, that it works with the current
* setup and it should be good enough for personal use, where there are only a small
* amount of machines and messages.
* Obviously, this doesn't scale well to lots of machines and messages. It should be good
* enough for personal use, where there are only a small amount of machines and messages.
*
* The main use case for docspell is to communicate between the rest-server and job
* executor. It is for internal communication and all topics are known at compile time.