CouchDB 3.0 • Hacker News Board

Re: CouchDB 3.0

#21

Post by newfeatureok » Wed, Feb 26, 2020, 7:35 PM UTC

tbrock wrote: post #12

At this point why would you use CouchDB over something like MongoDB? Seriously asking... Over the past 5 years MongoDB has gotten a great storage engine, transactions, distributed transactions, multi master replication, first class change streams and is very very solid as a foundational piece of infrastructure you can rely on while CouchDB has languished. I can’t imagine reaching for it in my tool belt when I need a…

The main reason most people use CouchDB is because of the HTTP API and offline support with Couchbase Mobile and PouchDB. Doesn't CouchDB have most of those things already from 2.3?

Re: CouchDB 3.0

#22

Post by liamdiprose » Wed, Feb 26, 2020, 7:35 PM UTC

tbrock wrote: post #12

At this point why would you use CouchDB over something like MongoDB? Seriously asking... Over the past 5 years MongoDB has gotten a great storage engine, transactions, distributed transactions, multi master replication, first class change streams and is very very solid as a foundational piece of infrastructure you can rely on while CouchDB has languished. I can’t imagine reaching for it in my tool belt when I need a…

The link explains that CouchDB can have replicas on mobile phones and websites, meaning clients don't always have to be connected to the internet.

> The Couch Replication Protocol lets your data flow seamlessly between server clusters to mobile phones and web browsers, enabling a compelling offline-first user-experience

Re: CouchDB 3.0

#23

Post by splatcollision » Wed, Feb 26, 2020, 7:36 PM UTC

CouchDB is awesome, full stop.

While it's missing some popularity from MongoDB and having wide adoption of things like mongoose in lots of open source CMS-type projects, it wins for the (i believe) unique take on map / reduce and writing custom javascript view functions that run on every document, letting you really customize the way you can query slice and access parts of your data...

Example: I'm building a document analysis app that does topic + keyword frequency vectorization of a corpus of documents, only a few thousand for now.

I end up with a bunch of documents that have "text": "here is my document text..." and "vector": [ array of floating point values ...].

What I can do with couchdb is store that 20d vector and emit integers of it as a query key:

    var intVectors = doc.vector.map(function(val){
      return Math.floor(val)
    })
    emit(intVectors, 1);

Then I can match an input document's vector (calculated the same as corpus documents), calculate a 'range' of those vectors, pass it as start and end keys, and super quickly get a result from the database of 'here are documents that have vectors similar to your input'...

Super fun, quick and flexible to work with!

Re: CouchDB 3.0

#24

Post by speedgoose » Wed, Feb 26, 2020, 7:37 PM UTC

tbrock wrote: post #12

At this point why would you use CouchDB over something like MongoDB? Seriously asking... Over the past 5 years MongoDB has gotten a great storage engine, transactions, distributed transactions, multi master replication, first class change streams and is very very solid as a foundational piece of infrastructure you can rely on while CouchDB has languished. I can’t imagine reaching for it in my tool belt when I need a…

Does MongoDB have multi master replication or the classic election of one master from a pool of candidates ?

CouchDB has another pattern, each master is really a master and you can have live replication but also offline replication. You can connect two clusters every new moon and they will synchronize. For sure the clients may have to deal with potential conflicts but in practice it's very neat and that's what makes couchdb worth it if you need this feature.

Re: CouchDB 3.0

#25

Post by splatcollision » Wed, Feb 26, 2020, 7:37 PM UTC

newfeatureok wrote: post #16

Graphguy wrote: post #4

Earlier quoted context omitted.

Cloudant on IBM Cloud is CouchDB API/replication compatible and offers support for Apache CouchDB (1). Also, OpenWhisk integrates nicely with CouchDB/Cloudant and can even be a backing persistence for it (2) (1) https://www.ibm.com/cloud/blog/announcements/announcing-supp... (2) https://github.com/apache/openwhisk/blob/master/tools/db/REA...

Cloudant is awesome, but it's way too expensive IMHO.

Partitioned dbs are supposed to allow you to query more cheaply, haven't implemented those yet.

Re: CouchDB 3.0

#26

Post by Lio » Wed, Feb 26, 2020, 7:38 PM UTC

newfeatureok wrote: post #6

pachico wrote: post #3

Earlier quoted context omitted.

I didn't understand. You mean it's unique to work over http(s)?

I mean your CouchDB instance itself is represented by a host and port and your application's data could be stored there and a native HTTP-based API to access said data . This is contrasted to most where you would need a driver and it's accessible only in the "back-end".

To take that further, I really like the idea of running something locally like PouchDB and then letting it sync with a remote CouchDB using the replication protocol.

Re: CouchDB 3.0

#27

Post by mauflows » Wed, Feb 26, 2020, 7:38 PM UTC

newfeatureok wrote: post #21

tbrock wrote: post #12

At this point why would you use CouchDB over something like MongoDB? Seriously asking... Over the past 5 years MongoDB has gotten a great storage engine, transactions, distributed transactions, multi master replication, first class change streams and is very very solid as a foundational piece of infrastructure you can rely on while CouchDB has languished. I can’t imagine reaching for it in my tool belt when I need a…

The main reason most people use CouchDB is because of the HTTP API and offline support with Couchbase Mobile and PouchDB. Doesn't CouchDB have most of those things already from 2.3?

I don't think couchbase has couchdb in mind for the mobile client anymore

Re: CouchDB 3.0

#28

Post by almery » Wed, Feb 26, 2020, 7:39 PM UTC

tbrock wrote: post #12

At this point why would you use CouchDB over something like MongoDB? Seriously asking... Over the past 5 years MongoDB has gotten a great storage engine, transactions, distributed transactions, multi master replication, first class change streams and is very very solid as a foundational piece of infrastructure you can rely on while CouchDB has languished. I can’t imagine reaching for it in my tool belt when I need a…

CouchDB is licensed under Apache License 2.0, while MongoDB uses SSPL, which was rejected by the OSI https://www.zdnet.com/article/mongodb-open-source-server-sid...

Re: CouchDB 3.0

#29

Post by splatcollision » Wed, Feb 26, 2020, 7:39 PM UTC

hajile wrote: post #18

Reducing max document size from 4GB down to 8MB seems hyper-restrictive. For those interested, looks like the guts of CouchDB are going to be swapped out for FoundationDB. https://blog.couchdb.org/2020/02/26/the-road-to-couchdb-3-0-...

If you're trying to store single GB documents in couch, you're doing it wrong... Unless those are binaries you can usually fragment data logically across many documents, then write custom views to aggregate however you need to.

Updates on huge docs would be painful!

Re: CouchDB 3.0

#30

Post by Graphguy » Wed, Feb 26, 2020, 7:39 PM UTC

speedgoose wrote: post #24

tbrock wrote: post #12

At this point why would you use CouchDB over something like MongoDB? Seriously asking... Over the past 5 years MongoDB has gotten a great storage engine, transactions, distributed transactions, multi master replication, first class change streams and is very very solid as a foundational piece of infrastructure you can rely on while CouchDB has languished. I can’t imagine reaching for it in my tool belt when I need a…

Does MongoDB have multi master replication or the classic election of one master from a pool of candidates ? CouchDB has another pattern, each master is really a master and you can have live replication but also offline replication. You can connect two clusters every new moon and they will synchronize. For sure the clients may have to deal with potential conflicts but in practice it's very neat and that's what makes…

MongoDB does replica sets and sharding. As far as I know it doesn't support multi-master architectures or data syncing. Even the Atlas-style global data distribution is sharding, right?