Rxdb: A reactive database where you can subscribe to the result of a query

https://github.com/pubkey/rxdb

by dsun180 • 7 years ago

288 122 7 years ago

github.com

siscia• 7 years ago

What is the time complexity of receiving a notification?

Are there triggers that run on every update on the dataset?

throwaway_bad2• 7 years ago

Having client state just be a replica of server state solves so many problems I don't understand why the concept never caught on. Pouchdb/couchdb are still the only ones doing it afaik.

Instead we have a bajillion layers of CRUD all in slightly different protocols just to do the same read or write to the database.

skrebbel• 7 years ago

How would that work, with, say Twitter? Copy all tweets ever to each browser?

I mean, that's why it didn't catch on right :-) It's hard :-)

throwaway_bad2• 7 years ago

Same way you scale replication for any server, by sharding and only replicating the shards you care about.

The "shard" could just be that users own feed in this case. Then you get offline for free where user adds a tweet and it appears immediately, replicating back to server when he goes back online. The server replica side will need to be a lot more complicated to deal with broadcasting but I don't see why it won't work.

bayesian_horse2• 7 years ago

I've heard somewhere that Twitter maintains a copy of all tweets in a user's timeline for that user.

If I had to make a Twitter clone with CouchDB, I would probably have one timeline document per user, and maybe one per day to limit the syncing bandwidth.

dsun18015• 7 years ago

There is a limit on what data can be replicated with the client. For a simple chat-app you can replicate all messages of a user. But you would never replicate the whole state of wikipedia to make it searchable.

koolba• 7 years ago

When your data is public and immutable, this approach is very pleasant. The client becomes just another caching layer and worst case it's presenting a historical version of the truth. You can even extend this across tabs with things like local storage.

This breaks down quickly once you have data that could become private or mutate rather than append.

rockwotj2• 7 years ago

I think the Firebase Realtime Database and Firestore have a good model for offline and being able to have private and mutable data. It does get complex but the Firebase SDKs do the heavy lifting for you here.

tehbeard• 7 years ago

This what stopped me exploring couchdb further.

The "one db per user" model for private data made using other features like views etc more difficult when you have to upgrade,edit,remove them.

Mutability wasn't really a problem, either present the conflicts to user and pick one or write code to merge if possible.

megous2• 7 years ago

In reality data can't become private again, after being available. You may try to contact all users to delete their copy, but they may not respect that.

mch82• 7 years ago

RxDB can do a realtime replication with any CouchDB compliant endpoint and also with GraphQL endpoints.

PouchDB is mentioned a couple of times, including one “PouchDB compatible” mention. Wondering what unique use cases RxDB supports?

gnur• 7 years ago

Firestore is doing something very similar and it is very easy to use.

dustingetz2• 7 years ago

Data security is a huge issue, Facebook.com has very specific whitelisted access patterns encoded as CRUD endpoints. 1) Users want to make sure their data is used in non-creepy or non-stalky ways; 2) Facebook's business needs to control the access point so they can serve you adds or otherwise monetize. So the API exposes only limited access patterns, the API TOS disallows caching, and they go to great lengths to prevent scrapers.

If immutable fact/datom streams with idealized cache infrastructure becomes a thing (and architecturally i hope it does) it's going to need DRM to be accepted by both users and businesses.

azr79• 7 years ago

Security is a huge deal

dtech• 7 years ago

That's exactly the model that Apollo Client library uses (GraphQL-based data store for react), and teams I spoke that tried it are quite enthusiastic for this reason.

gwbas1c• 7 years ago

Having client state just be a replica of server state solves so many problems I don't understand why the concept never caught on.

As soon as your server state is larger than whatever your client can handle, the whole metaphor breaks down.

ralusek• 7 years ago

I'm assuming that the client state would be a lazy representation of the back end state, that only pulls data as needed. The result being that local and server state must both be treated only as asynchronously accessible.

sergiotapia• 7 years ago

MeteorJS tried it and failed miserably. It doesn't scale.

vlasky• 7 years ago

Wrong. This is popular anti-Meteor FUD spread by people who don't know how to use its features properly or have the engineering/computer science background to design a system to be able to manage computational complexity or scalability.

In 2015, my business implemented a Meteor-based real-time vehicle tracking app utilising Blaze, Iron Router, DDP, Pub/Sub

Our Meteor app runs 24hrs/day and handles hundreds of drivers tracking in every few seconds whilst publishing real-time updates and reports to many connected clients. Yes, this means Pub/Sub and DDP.

This is easily being handled by a single Node.js process on a commodity Linux server consuming a fraction of a single core’s available CPU power during peak periods, using only several hundred megabytes of RAM.

How was this achieved?

We chose to use Meteor with MySQL instead of MongoDB. When using the Meteor MySQL package, reactivity is triggered by the MySQL binary log instead of the MongoDB oplog. The MySQL package provides finer-grained control over reactivity by allowing you to provide your own custom trigger functions.

Accordingly, we put a lot of thought into our MySQL schema design and coded our custom trigger functions to be selective as possible to prevent SQL queries from being needlessly executed and wasting CPU, IO and network bandwidth by publishing redundant updates to the client.

In terms of scalability in general, are we limited to a single Node.js process? Absolutely not - we use Nginx to terminate the connection from the client and spread the load across multiple Node.js processes. Similarly, MySQL master-slave replication allows us to spread the load across a cluster of servers.

For those using MongoDB, a Meteor package named RedisOplog provides improved scalability with the assistance of Redis's pub/sub functionality.

liuliu• 7 years ago

In short, server data is more normalized than the data client needs. As you get closer to view layer, your data gets denormalized further and further. Client-server interaction sits somewhere in the middle to both minimize the bytes-over-wire as well as the round-trips to the backend to get up-to-date.

Take a look at GraphQL, its central promise is to let client choose what's the optimal data it needs (that often denormalized through nested GraphQL queries), and send it in one batch.

It is not to say there shouldn't be a simple replica. It is just if we want it to be a simple replica, we should have a server-side mirrored some-what-denormalized representation rather than just the raw server-data models.

venantius• 7 years ago

I'm not sure why you would use this over something more battle-tested like RethinkDB which was also built with this use-case in mind.

silasdavis• 7 years ago

Because rethinkdb development has ceased? Last release was in 2017.

Is there still maintenance?

rustmusic• 7 years ago

Why not maintain it rather than writing something new? Open source is frustrating

jjuel• 7 years ago

The community has taken it over as far as I know. Looking at the github there appears to be some active development still. https://github.com/rethinkdb/rethinkdb

assface• 7 years ago

"Is RethinkDB dead?"

https://github.com/rethinkdb/rethinkdb/issues/6747#issuecomm...

TLDR: They are hoping to announce a revival in the near future.

dsun18015• 7 years ago

RethinkDB is not offline first. It runs queries on the server-side, not on the client. There is horizonDB which can replicate with RethinkDB but the project is dead with no commits since the rethinkdb company gave up.

habitue• 7 years ago

So I was one of the developers of RethinkDB's Horizon, and I'm really glad to see someone take the database change stream to RxJS concept to its logical conclusion.

This project looks really awesome, and very much has the shape we wished Horizon could have turned into

dizaime2• 7 years ago

Meteor do this since inception with MongoDB on server and MiniMongo in client.

dsun18015• 7 years ago

Yes, mongodb and meteor and minimongo do something similar but with much more restrictions on what you can do.

RxDB does exactly one thing which is being a client side database. You are not tied to a specific ecosystem or backend database.

dizaime2• 7 years ago

Yeah, with RxDB, we are tied with RxJS, PouchDB and CouchDB.

Anyway, database is hard, opensource is great. We love them both, RxDB or Meteor.

henriks• 7 years ago

Isn't "subscribe to all state-changes like the result of a query" something you'd implement efficiently using a forward chaining inference engine, like something based on Rete?

dsun18015• 7 years ago

It is implemented with something called QueryChangeDetection here. https://rxdb.info/query-change-detection.html

This is based on meteors oplog-observe-driver https://github.com/meteor/docs/blob/version-NEXT/long-form/o...

So basically when a change-event comes, RxDB does not run the query against the database again, but instead uses the old results together with the event to calculate the new results.

digitalneoplasm• 7 years ago

I don't know too much about database systems, but if we were in an inference (e.g., production) system, then that would make a lot of sense! I build logical inference systems and have implemented this kind of functionality in forward, backward, and bi-directional logical inference.

GrayShade2• 7 years ago

EDIT: never mind, I should have read the article.

dsun18015• 7 years ago

I'm sorry but I think you comment is spam. The product in your article has only one similarity with RxDB which is being a database.

GrayShade2• 7 years ago

You're right, sorry. I didn't click on the link and thought it was something else. I can't delete my comment, unfortunately.

jinjin2• 7 years ago

This is something Realm does really well, both locally on the device and also with live subscriptions to data in the cloud.

We used this for our mobile apps and the experience was pretty awesome. The ability to live observe both individual objects in the dB and results of queries makes building reactive UI’s a very pleasant experience.

bot1• 7 years ago

So like firestore?

egeozcan• 7 years ago

I was complaining about lack of such a database just today: https://news.ycombinator.com/item?id=21352235

I'm not doing field projects anymore but knowing this exists would have helped me tremendously in the last few years.

Thanks for sharing!

stigi• 7 years ago

Another DB that plays in that field is Watermelon DB (https://github.com/Nozbe/WatermelonDB) which supports react-native based on SQLite and the web based on the LokiJS in memory DB.

MehdiHK• 7 years ago

If you want a change-stream out of your regular MySQL or PostgreSQL database systems, check out Debezium: https://debezium.io/

It basically registers itself as a fake replica server so that it can get updates from the master (like binlog in case of MySQL) and then it forwards those updates to Kafka. The possibilities are endless what you can do with those updates as Kafka consumers.

dsun18015• 7 years ago

There is a big difference between the changestream of many SQL and noSQL databases, and what RxDB does. Having a stream of changes is useful but not the whole solution. RxDB is capable of using single document changes of a stream and recalculate the new results of an existing query. This saves you not only much IO performance but makes developing much easier. See https://rxdb.info/query-change-detection.html

tga2• 7 years ago

Sure, that's handy as an extra feature to get better performance, but I see that even in RxDB it's currently in beta and turned off by default.

I suppose this could be the killer feature to warrant using a new database -- because otherwise, from a user point of view, the result is the same.

tga2• 7 years ago

If you want to subscribe to changes in a Postgresql database, whether produced by your app or not, you can also use Hasura or PostGraphile:

https://docs.hasura.io/1.0/graphql/manual/subscriptions/inde...

https://www.graphile.org/postgraphile/realtime/

dsun18015• 7 years ago

There is a big difference between the changestream of many SQL and noSQL databases, and what RxDB does. See https://news.ycombinator.com/item?id=21353971

kiwicopple10• 7 years ago

I tried out Postgraphile Realtime and it's pretty cool.

For anyone who wants something similar that's not GraphQL, then I'm in the early stages of developing a Phoenix (Elixir) implementation which broadcasts changes over websockets: https://github.com/supabase/realtime

See comments here: https://news.ycombinator.com/item?id=21354039

BenjieGillam• 7 years ago

Neat; happy to have inspired you!

ivanjaros• 7 years ago

its all just a for loop anyway...you can do the same if you put tiny wrapper on db-local server to avoid network and you'll get the same result.

Kuinox2• 7 years ago

IMO JavaScript is not compatible with "realtime".

ceejayoz2• 7 years ago

“Real-time” in JS tends to refer to approaches like websockets.

Kuinox2• 7 years ago

So it's not really realtime... No ?

I noticed an instant downvote. It's not the first time i see this behavior on rxdb comments.

dsun18015• 7 years ago

Depends on your definition of realtime. Realtime with RxDB is not like "Real-time Computing" but like realtime synchronisation how it is described by firebase https://firebase.google.com/docs/database

kiwicopple10• 7 years ago

I have been developing something that provides this functionality for PostgreSQL: https://github.com/supabase/realtime

It's still in very early stages (although I am using it in production for my company)

It's very similar to Debezium (mentioned in another comment), but it's built with Phoenix (elixir), so great for listening via websockets.

Basically the Phoenix server listens to PostgreSQL's replication functionality and converts the byte stream into JSON which it then broadcasts over websockets. This is great since you can scale the Phoenix servers without any additional load on your DB. Also it doesn't require wal2json (a Postgres extension). The beauty of listening to the replication functionality is that you can make changes to your database from anywhere - your api, directly in the DB, via a console etc - and you will still receive the changes via Phoenix.

I still have to document a lot of how it works and how to use it, but if anyone is interested then I will make it a priority over the weekend

dsun18015• 7 years ago

kiwicopple10• 7 years ago

Yeah perhaps "same functionality" is an ambitious statement. I just saw debezium mentioned below and thought i'd throw this out in case someone reading this and it fits their needs!

tjchear• 7 years ago

I would love to use this just for the offline and query change detection capabilities alone - but the latter is currently beta and disabled by default. Is there a reason for that?

markandrewj3• 7 years ago

RethinkDB might be another database to look at if this is functionality which you are interested in. My only gripe with doing this sort of thing is that if you offer a public API, it becomes separate to what you are using for the application internally. There are also some GraphQL databases like Prisma that offer this sort of functionality.

kiwicopple10• 7 years ago

I actually used Firebase at the start but this implementation is so that we can migrate 100% to Postgres. I agree though, Rethink is awesome

rch• 7 years ago

Did you evaluate gRPC or gRPC-Web when considering your websockets implementation?

I'm using gRPC in other languages/frameworks and I'd really like to migrate one of those to Elixir in the near future, but I'm curious to know if anyone else has given it a try.

There's a prominent warning on the repo:

https://github.com/elixir-grpc/grpc

kiwicopple10• 7 years ago

I didn’t try, to be honest i think the implementation of this might be beyond my elixir capabilities. But I’ll check it out - thanks for the link!

pbowyer• 7 years ago

Interesting, thanks for sharing how you're doing it.

Is there any mileage in doing this with triggers? I have a _very_ legacy system which needs caching adding. Rather than dig through the code to invalidate the cache every time a record is updated/deleted in 20+ tables, I am thinking that being able to listen to the SQL executed and invalidate the cache based on the tables involved would be a clean approach.

But not found any way to make that possible - yet.

kiwicopple10• 7 years ago

Yeah I also used triggers at the start. There are 2 things that this implementation achieve over triggers.

1. Triggers have an 8000 byte limit. I ran against these limits pretty quickly

2. You need to attach the trigger each time you create a new table. With this you can set and forget

ronmex• 7 years ago

You can use apache ignite as a caching layer over your db. (Or as a db itself). You can then register a ContinuousQuery to listen for changes. A grpc service can be stood up to stream these changes to other clients, even front end ones (via grpc-web).

marknadal2• 7 years ago

There is also (mine) https://github.com/amark/gun

It is run by Internet Archive, HackerNoon, DTube, Notabug, etc.

Handles about 8M monthly active users!

+ end-to-end encryption, graphQL & graph data, upcoming Svelte support, decentralized, etc.

MIT/Zlib/Apache2 licensed.

yc-kraln• 7 years ago

I read through your whole readme and I still am not able to answer the question: "what kind of project would I use GUN for?"

or, even, "what is gun replacing?"

I mean, it looks interesting and I would like to play around with it, but I don't understand enough to know what kind of stuff I can build with it.

dsun18015• 7 years ago

I tried several times to understand what GUNdb does and how its different or which features it has. I have given up. As far as I can tell it is something between Blockchain, graph-database and sync.

I also tried to understand the codebase which is impossible with obfuscated code like this one: https://github.com/amark/gun/blob/master/lib/store.js#L16

thunderbong• 7 years ago

Gun is essentially similar to Firebase[0]

I've used it in some projects and found it useful. Thanks.

[0] https://firebase.google.com/

maxmcd2• 7 years ago

General question about reacting to database events: It seems like when responding to DB events it would be easy to accidentally create an infinite loop. Is that an issue with this pattern, or is it easy to avoid? Do any of these data subscription tools have safeguards in place to prevent this?

rockostrich• 7 years ago

Isn't this the case with any interaction between 2 systems? Between a client and server, you can have logic in the client that reacts to a response from the server that triggers another request to the server.

maxmcd2• 7 years ago

Very true, maybe I'm overthinking it? In my head: If I update full_name from first_name and last_name whenever a user table row changes, if I do this naively I will update the user on every update and trigger a continuous loop of updates.

I have certainly created infinite loops while using React's componentDidUpdate, maybe it's just important to define triggers on single attributes rather than entire database rows.

ihrimech• 7 years ago

This is where command and query (with subscribe) must be neatly separated. An event A launches an updates of the DB, that launches the query part to react, then stop. There is the risk of having an infinite loop, if the event A is launched by the query part, which is very related to the behavior of your app.

jimmyspice• 7 years ago

Naively it sounds like the halting problem. You would have to verify that no consumer can make a change to the data set you are listening to. You would want to manage yourself when not to respond to an event.

hechang1997• 7 years ago

The ideal case is that the DB can work out the difference in the subscribed query caused by the DB update so that the front end can make changes incrementally and doesn't have to rerender the entire list/table.

hashkb• 7 years ago

Hasura does this for Postgres. What are some areas where I'd use this instead?

dsun18015• 7 years ago

RxDB is offline first. You can still query your data even when the user has no internet. Hasura will not make your app workable without a stable connection to the server.

1337shadow• 7 years ago

Nice ! reminds me of the Ryzom project in Python/Django/Postgres which implements a meteorjs-like protocol and proposes the same example with a todo list

https://github.com/yourlabs/ryzom

bilekas• 7 years ago

I personally don't like the idea of that coupling, I prefer to use my own events, and go through websockets for ex..

This is a nice feature for some, but not all situations..

jtmarmon• 7 years ago

You can do this in datomic! I made an end-to-end proof of concept of this using datomic and websockets. The client can open a websocket and subscribe to a query and then the server will react to changes in the database and automatically update the model and therefore the UI (no eventing code required).

Demo vid https://v.usetapes.com/85knuAPruB Longer description in github readme https://github.com/jtmarmon/hackerthreads

dustingetz2• 7 years ago

Also see https://github.com/metasoarous/datsync - note that this project stalled out due to technical challenges, though i hear they're stealthily working on it again and coordinating with the Declarative Dataflows project https://github.com/comnik/declarative-dataflow

dsun18015• 7 years ago

I think you have not understood what RxDB is. It is a client side database. It works also when the client is offline. It does not need a stable internet connection over a websocket.

Your example is similar to RethinkDB and others. A websocket streaming json.

kickopotomus• 7 years ago

A more common solution includes using Datascript[1] as the client side DB and then using something like Posh[2] or Datsync[3] to handle query/pull subscription changes.

[1] https://github.com/tonsky/datascript

[2] https://github.com/mpdairy/posh

[3] https://github.com/metasoarous/datsync

amelius2• 7 years ago

How useful is real-time data really? For example, HN isn't updating in real time and for me that's fine. If real-time updates come at a ridiculous complexity and performance cost, then is it really worth it?

See also Google Wave, which had "real time" as its main novelty, but ultimately failed.

s_y_n_t_a_x• 7 years ago

You know that it depends on the situation.

beaconstudios3• 7 years ago

for my project it's very useful for collaborative features. Multiple people need to be able to edit the same graph structures at the same time, which would be difficult to scale out via a notify-and-poll architecture.

amelius2• 7 years ago

I've found that collaborative editing can be very confusing, and prefer a commit/merge step over simultaneous editing. I think all collaborative editing software should at least offer this kind of interface.

beaconstudios3• 7 years ago

hey, this is great! I've been building my own version of exactly this concept based on an event-sourced in-memory graph for my visual programming environment and it works really well for creating a collaborative editing system. I'll be investigating this project now to see if I can't use it instead of my hand-rolled solution.

lessname• 7 years ago

What I personally like about RxDB is that it also works completely offline. However, I don't really understand encryption: As far as I understood, it encrypts on the client side with the db passwort which is sent to the server.

Another issue is Authorization and Authentication, I could not find a good solution for me for CouchDB. Couchbase seems to have better solutions for this but the premium plan seems really expensive and as far as I understood you need a "server" and a "sync server" which don't have low system requirements, at least for me.

dsun18015• 7 years ago

No you normally do not send the password to the server. You can ask the user to enter the password when the application is openend.

Authentication is much easier when you use the GraphQL replication. There you are much more flexible on which data you return depending on which user is asking for it.

winrid2• 7 years ago

Cool. But how is this better than Meteor for building webapps? Meteor would scale better too with Apollo.

dsun18015• 7 years ago

I do not think that meteor will scale better then your GraphQL server with whatever database you want to have. With meteor you are bound to a specific ecosystem which is often a pain. RxDB does only one thing, it is a client side database. Everything else in your stack is free to choose.

winrid2• 7 years ago

Oh, I didn't know this was just for the client. Missed that.

psoren• 7 years ago

Maybe this is an uninformed comment, but how is this different from the Firestore database in Firebase where you can listen for when a document changes?

mizzao• 7 years ago

Anyone know how this is different from the reactive query layer that Meteor built on top of Mongo?

jaime-ez• 7 years ago

deepstream.io has similar functionality and supports postgres, rethinkdb and other databases. It is a great tool and promotes a very nice paradigm.

TheUndead96• 7 years ago

This project has all of the goals that excited me about RethinkDB, I hope this project is more successful.

mitar• 7 years ago

I have made something similar trying to match more the concept of a materialized view, but just on the client instead of inside the database: https://github.com/tozd/node-reactive-postgres

Rxdb: A reactive database where you can subscribe to the result of a query

Related Stories