Hacker News Reader

Comment by Tomdarkness | Hacker News Reader

Or you could actually use something designed for indexing data and searches, like Elasticsearch or Solr.

Either solution would have no problem indexing all their data, rather than having to limit it to a subset to fit in a in-memory table.

View on HN

+1 for Solr - looking at the search on domcop.com, it seems like a perfect fit for the faceting stuff.

There might be some messiness with the regex search (looking for user-defined patterns of consonants, vowels, etc) as I've never needed to set that up, but the rest would be really clean.

See also this blog post that discusses how RoomKey uses read-only Solr instances that are repopulated daily to speed up searching for latency-insensitive data (think "does this hotel have a pool?" etc):

http://www.colinsteele.org/post/23103789647/against-the-grai...

This was initial thought as well, seems like both would fit the needs specified, multiple column search etc. I've had good experiences with both ES and Solr. They both have pretty healthy user bases and are well documented. Only hang-up is if you use primarily reg-ex style searching; turning on n-grams could work there, but it might be too slow.

Those solutions assume text is tokenizable, and suck at regex. I think they are the wrong tool for anything having to do with domain search. Furthermore, they don't play nice with continuous cron indexing.

solr has real time indexing and batch indexing

I haven't used either elasticsearch or solr, so I don't have firsthand experience on how they compare to Sphinx search (http://sphinxsearch.com) but I really love Sphinx. Also integrates nicely with MySQL if you're already running that.

I think Solr and Sphinx pretty much tick the same boxes in terms of features and performance though, so use what you know and like.

Thanks for the info. I haven't previously come across Elastisearch or Solr. Will definitely check these out.

Does Elasticsearch or Solr work with tabular data, can you search across multiple columns?

Yes you can, and a whole lot more.

I can't speak well for Solr, but with elasticsearch the answer depends on what you exactly mean with "search across multiple columns", but it's probably yes.

Lucene models documents as a collection of fields, each with a textual value.

http://lucene.apache.org/core/4_0_0/core/org/apache/lucene/d...

At search time you can use the default field or specify the fields to be searched:

http://lucene.apache.org/core/2_9_4/queryparsersyntax.html#F...

Yes. Read the example SOLR schema, it's very easy to understand.

>Elasticsearch >Register to watch

Any site that requires me to register an account before I can even look at their product is going to have a bad time.

As far as I know you can download ElasticSearch from here without registering: http://www.elasticsearch.org/download/ ... what better way to check out the product than play around with it?

It is a open source project. They sell support, but elasticsearch itself is licensed under the Apache license. Just click overview if you want to kniw what it is, I'm not sure what you are clicking on that requires registration. Perhaps their training sessions?

Their videos require registration but the docs do not.

Reply on news.ycombinator.com