Closed abitrolly closed 5 years ago
Might be if the package is available in the default repo and EPEL, but I'm not sure.
Another example libssh2
shows 2 results, curl
shows 4 results and so on.
I wasn't able to find the pattern for this.
Looking at the sources, it looks like the issue is coming from here https://github.com/fedora-infra/fedora-packages/blob/master/fedoracommunity/search/index.py#L200
PDC is probably returning the duplicate results.
https://pdc.fedoraproject.org/rest_api/v1/global-components/?name=curl returns one result. https://pdc.fedoraproject.org/rest_api/v1/global-components/?name=curl&name=curl one result. https://pdc.fedoraproject.org/rest_api/v1/global-components/?name=curl&name=enki two results.
My bet is that there is some problem with database indexing. Like index key is not unique or contains colums that change for the same product name.
I cannot reproduce the behavior on the version 4.2.0 tag. Not sure when the release for this is though.
@steaksauce- do you use the same DB as production?
@steaksauce- do you use the same DB as production?
tbh, I don't know (probably not though).
Whatever docker-compose
is giving me in the devel folder
I have been working on https://github.com/fedora-infra/fedora-search to replace the xapian backend of fedora-packages. I think having an independent service and the db stored in postgresql will make it easier to investigate and fix this kind of problems
And if the issue is repeated with the same backend, how do you debug it?
Also, how do you share PostreSQL DB?
Also, are you sure that PostreSQL stemmers are good for full text search?
And if the issue is repeated with the same backend, how do you debug it?
The indexing is making use of django ORM which makes the code easier to understand, instead of the current low level xapian api. The other main reason for using postresql is to deploy this application in our Openshift setup.
Also, how do you share PostreSQL DB?
This is relatively easy we already dump some of our databases nightly see here https://infrastructure.fedoraproject.org/infra/db-dumps/
Also, are you sure that PostreSQL stemmers are good for full text search?
Postresql full text search has been around for a while now, and it seems a viable solution for simple search application like this one.
Postresql full text search has been around for a while now, and it seems a viable solution for simple search application like this one.
I couldn't find any serious comparison yet, but there are references that PostgreSQL supports stemming, so I have to believe you that it is good. :)
Closing this ticket
Just search for
enki
.