Fork me on GitHub
#cljdoc
<
2019-10-28
>
schmee13:10:50

is cljdoc down? 😮

martinklepsch13:10:22

looks like it! thanks for the ping, investigating 🙂

martinklepsch20:10:27

@borkdude I think we’ve considered it but running an extra process wasn’t worth what we would gain over plain Lucene

martinklepsch20:10:36

or so anyways 🙂

martinklepsch20:10:18

Do you think ElasticSearch would be much easier/get us anything over Lucene as we use it now?

borkdude21:10:09

@martinklepsch ElasticSearch provides a convenient layer over Lucene with a REST API. It depends what you want I guess. If you throw JSON at ES you already can get a long way without much config maybe. But Lucene may work good as well. Just wanted to hear your thought on this.

martinklepsch22:10:47

Yeah I think given cljdoc’s single node “architecture” we didn’t really need that REST API and a Java API might actually be easier to deal with. That said it could totally be that some things would be easier with ElasticSearch. I remember that one thing @holyjak was looking into was adjusting the search so that results are ordered by download count and that isn’t completely straightforward. Not sure if it would be easier with ES though.

holyjak09:10:43

I do not see much gain in using ES. It is still Lucene underneath though you talk to it differently and there are perhaps some conveniences…. But I know little about it. The download count inclusion is a work in progress, the trouble is setting weights correctly so that it has just the right effect. See https://github.com/cljdoc/cljdoc/pull/359