Shoud fuzzy search be supported?

[suzumiyasmith]

Now, it is obvious that fuzzy search is not supported. e.g. search rethink won't get rethinkdb package as a result. Sometimes, it can be annoying. :(

Any idea or plan about fuzzy searching?

Thanks.

[tfausak]

I would like to see this too. I can't remember the number of times I've searched for a partial string and been surprised by the lack of results. Some recent examples that I remember:

• Searching for influx (https://hackage.haskell.org/packages/search?terms=influx) returns no results even though there is an influxdb package (https://hackage.haskell.org/package/influxdb).
• Searching for postgres (https://hackage.haskell.org/packages/search?terms=postgres) doesn't return postgresql-simple (https://hackage.haskell.org/package/postgresql-simple), which is the most popular PostgreSQL package.

That being said, I'm not holding my breath for a fix. Hackage's search appears to be based on keywords and word stems. And search issues tend to stay open for ... a while.

• Hackage search "compiler" misses purescript #474
• Search by partial package name #208
• Use tags in search #27
• Add "keyword in context" feature for search results #21
• Support search in module names #20

Have you tried searching on Stackage? It does a better job for all the queries given here:

• https://www.stackage.org/lts-9.3/hoogle?q=rethink
• https://www.stackage.org/lts-9.3/hoogle?q=influx
• https://www.stackage.org/lts-9.3/hoogle?q=postgres

[gbaz]

Hackage is under active development at the moment, and search is definitely a place we want to tackle! thanks for the links to related issues, but I wouldn't be so negative about the possibility for progress. More hands and more hackers are always welcome :-)

[gbaz]

So I think the simplest thing to do is ad-hoc improve our stemmer since it seems to work just fine except for "weird" computery stems like dropping "db" or "ql" etc.

[adamgundry]

It might be worth switching Hackage to use the full-text-search package. This was derived from the Hackage search implementation (so it provides a very similar interface and switching should be fairly trivial), but it has a few extra features, notably autocomplete search.

[gbaz]

Good idea. Any volunteers? :-P

[adamgundry]

I'm hoping to get a chance to look at the full-text-search integration at some point (but no promises, and no objections to someone else jumping in!).

On the original issue, another case I just happened across (see haskell/binary#92) is that searching for "serialisation" yields binary whereas searching for "serialization" yields cereal.

[gbaz]

@adamgundry do you think we should always change "z" to "s" in all our stemming? Might be the simplest way to prevent the pitfalls of britishised spellings (or britishized if you prefer).

[gbaz]

I do worry about other internationalisation issues that have a different flavour...

[gbaz]

Ad-hoc stemming now works. That means postgres and influx and rethink searches work. I think that may suffice to close this issue?

[gbaz]

I'll make a different issue for full-text-search