Expose Purge Infos to Clients

[janl]

Summary

CouchDB keeps a list of purge infos to ensure that purges can be applied on a cluster without purged documents being re-introduced by internal replication.

It would be useful to make this list available for replication clients like PouchDB, who then could apply local purges on their own. I know PouchDB doesn’t implement purge just yet, but it’s something that folks will need before long.

Desired Behaviour

GET /db/_purge_infos[?limit=X&offset=Y]
{
  "purge_infos": [
    {...purge info...},
    ...
  ]
}

Possible Solution

Expose the internal purge_infos API to chttpd.

[jiangphcn]

any other contributor is welcome to start with this. If not, I can do this. Just one information, we didn't keep all historical purge_infos and there is limit for purged-info https://docs.couchdb.org/en/stable/api/database/misc.html#db-purged-infos-limit. If the stored purged info exceeds limit and all related parties (such as internal replication, mrview, search index, etc) purged their own stuff, the purged info itself will be deleted permanently. So the returned purge_info might not be complete.

[palash25]

Hi @janl I am learning Erlang and wanted to take a shot at this. Just to be sure we need to use the load_purge_infos/2 function to get the purge info right?

[jiangphcn]

thanks @palash25 You can get purge info from this and underlying API couch_db:fold_purge_infos, and just pay attention to return aggregation of purge infos from all shard db because request was distributed here https://github.com/apache/couchdb/blob/master/src/fabric/src/fabric_doc_purge.erl#L40

[palash25]

thanks for the help @jiangphcn will take a look at it in a few days.

[janl]

Heya @palash25 how’s it going, anything we can help with? :)

[palash25]

Hi @janl haven't had the time to take a crack at this the past week, give me till the end of this week if it isn't urgent.

I was also having some problem with the dev setup so need to fix that first.

[wohali]

Hey @palash25 , if you have access to Docker, here's a development environment for you:

docker pull couchdbdev/debian-buster-erlang-20.3.8.25-1:latest
docker run -it couchdbdev/debian-buster-erlang-20.3.8.25-1:latest /bin/bash

Once inside the container:

cd

will get you to the jenkins user home directory. You can now checkout CouchDB and develop from there.

Hope this helps!

[palash25]

Hi @wohali thanks for this also I actually found this issue thread yesterday which describes the exact problem I am facing with the build now. #455

I don't prefer to develop using docker, I tried the steps you listed in that issue but wasn't able to build couchdb from source still got the same jsapi.h error. Should I move the conversation to #445 ?

I would only like to develop using docker as a last resort.

[wohali]

@palash25 No, #445 is closed. Do not reopen that.

Which operating system are you using? If Linux, which distribution? If it's one of the following:

• Debian stretch/buster
• Ubuntu 16/18
• RedHat 6/7/8
• CentOS 6/7/8
• macOS
• FreeBSD

you can use the script here to install all the dependencies you need in your base operating system:

https://github.com/apache/couchdb-ci/blob/master/bin/install-dependencies.sh

You'll need to clone the entire repository, then run that script. That should get you running.

If it's not one of those, you may be forced to compile SpiderMonkey from scratch, which is not fun.

[palash25]

I am on Ubuntu18.04 @wohali , I ran that script but I still keep getting the jsapi.h error, nevermind I will go with the docker setup