is it support chinese search? like "你好".

Hi, we currently have Soulmate setup in our rails app and we are starting to hit high enough traffic where we want to separate soulmate out into its own process (prevent queries from being blocked behind unicorn)

Starting soulmate-web is simple enough but I cannot figure out how to query the separate Sinatra server from within our Rails app. Is this currently possible?

I have removed the `require: 'soulmate/server'`` dependency from the Gemfile but the Soulmate::Server middleware seems to still be present. Is there a way of configuring an external soulmate-web server rather than mounting it within the app?

After running bundle install with gems
gem 'rack-contrib'
gem 'soulmate', :require => 'soulmate/server'
in the gem file of my rails app (5.0.1) I keep getting this

LoadError: cannot load such file -- rack/showexceptions
/var/lib/gems/2.3.0/gems/sinatra-1.0/lib/sinatra/showexceptions.rb:1:in require' /var/lib/gems/2.3.0/gems/sinatra-1.0/lib/sinatra/showexceptions.rb:1:in <top (required)>'
/var/lib/gems/2.3.0/gems/sinatra-1.0/lib/sinatra/base.rb:6:in require' /var/lib/gems/2.3.0/gems/sinatra-1.0/lib/sinatra/base.rb:6:in <top (required)>'
/var/lib/gems/2.3.0/gems/soulmate-1.1.0/lib/soulmate/server.rb:1:in require' /var/lib/gems/2.3.0/gems/soulmate-1.1.0/lib/soulmate/server.rb:1:in <top (required)>'
/home/shivraj/Projects/fast-autocomplete/config/application.rb:7:in <top (required)>' /home/shivraj/Projects/fast-autocomplete/Rakefile:4:in require_relative'
/home/shivraj/Projects/fast-autocomplete/Rakefile:4:in <top (required)>' /var/lib/gems/2.3.0/gems/rake-12.0.0/exe/rake:27:in <top (required)>'
(See full trace by running task with --trace)
Error.
after running rake db:migrate.
Could anybody please tell me what the issue is about?

I recently upgrade to redis-rb 3 on Heroku and soulmate is not returning any results.

This is broken... using autocomplete gets 400 error:
Failed to load resource: the server responded with a status of 400 (HTTP/2.0 400)

{"message":"missing required parameter: q","meta":{"status":400},"status":400}

Probably the word 'term' needs to be changed to 'q'

After looking at the source for OmniAuth and later looking at the source of this project and noticing the serialization to JSON, I thought maybe multi_json would be a good choice for this project. It enables using yajl-ruby which IMO is in the spirit of redis, as its core is implemented in C, but will still work on JRuby because it falls back to using another library when yajl-ruby isn't found. I also noticed that ripple, the ruby client for riak, uses multijson.

The matcher uses a normalize function which only allows letters from a to z. Hence any Umlaute like ü,ö,ä etc. are ommited.

Thanks for the otherwise great gem,
Paul

Can we use custom filter to do the soulmate search.

For Ex: If we would like to add the city query in our data { city_id: 1 }. can we add it to soulmate?

Please let me know.

We regenerate our term data nightly and update the soulmate db. Due to vagaries of our setup, it's hard to keep id constant for a given record from night to night, so "soulmate add TYPE" would lead to duplicate entries. It would be simplest to have a "soulmate drop TYPE" command which allowed us to easily blitz and reload the whole collection. ("soulmate remove TYPE" requires a list of ids).

Thanks.

I have a large database of topics that users are allowed to search (via soulmate). Each topic has reputation on the website so that translates nicely to a soulmate score. However, when a user types a short word often longer words that include this short word will show up first. For example if the user types 'band', there are many topics in the world with the word band included in their name. I'm not sure what the best way to implement this is, but would it be possible to allow weighting of shorter words higher? Such that if someone searches 'band' the topic 'band' will be more likely to show up than 'dave matthew's band'? Maybe take into account the score as well.

I'm open to all ideas. Thanks!

Searching for english words is not case sensitive, but searching for russian words is case sensitive, i.e when the search term is "москва" nothing's found, while the search term "Москва" returns the correct result.

I have a large (8.25 million) JSON file that I want to import. This is what is causing the issue (bin/soulmate:44):

items = $stdin.read.split("\n").map { |l| MultiJson.decode(l) }

Needless to say this is very inefficient and creates a lot of ruby objects.

A better way would be to take a file argument and read the lines sequentially. I'll try to submit a pull request if I have time.

When I have the following terms in Soulmate:
["Citrix", "Citrix SaaS", "Citrix UK User Group", "Citrix User Group Norway", "Citrix VDI-In-A-Box (Kaviza) User Group", "Dutch Citrix User Group", "Freelancer Citrix Certified Trainer/ Consultant", "Remote Citrix Experts", "SBC Solutions your Citrix solution", "Xenergix - Expertos en Citrix"]

When I search Soulmate using the term "Citrix", I get these terms:
["Citrix SaaS", "Freelancer Citrix Certified Trainer/ Consultant", "Remote Citrix Experts", "Citrix", "SBC Solutions your Citrix solution", "Citrix VDI-In-A-Box (Kaviza) User Group", "Dutch Citrix User Group", "Citrix User Group Norway", "Citrix UK User Group", "Xenergix - Expertos en Citrix"]

Why isn't the exact match the first in the result?

I'm on Rails 3.2.13 and tried installing soulmate and it doesn't work. It says it is installed but the soulmate command doesn't work. It also looks like putting
gem 'rack-contrib'
gem 'soulmate', :require => 'soulmate/server'
in my gemfile and bundling doesn't seem to add them to my app.

I'm trying to run Soulmate directly via the Soulmate module. How do I do incremental updates of my index? Soulmate::Loader.new seems to blow away the existing index every time it's called.

(Also, will Soulmate work if used directly like this?)

I'm doing some stress tests of this gem to evaluate it for usage in my application. I don't know redis very well, so I don't fully understand code in Matcher#matches_for_term. I see that it accepts an option named cache with true as default value, and I'm benchmarking it's behavior.

I've integrated soulmate into a Contact class which produces the following hash for any given contact:

[21] pry(main)> Contact.completions 'ab'
=> [{"id"=>2085, "term"=>"Ms. Abe King", "score"=>9988},
 {"id"=>1762, "term"=>"Mr. Abner Bode", "score"=>9986},
 {"id"=>495, "term"=>"Ms. Hank Abbott", "score"=>9985},
 {"id"=>927, "term"=>"Mrs. Abbey Stehr", "score"=>9984},
 {"id"=>893, "term"=>"Mr. Tobin Abbott", "score"=>9984},
 {"id"=>787, "term"=>"Mrs. Otis Abbott", "score"=>9984},
 {"id"=>438, "term"=>"Mr. Abagail Kihn", "score"=>9984},
 {"id"=>199, "term"=>"Mr. Liana Abshire", "score"=>9983},
 {"id"=>1807, "term"=>"Dr. Frida Abshire", "score"=>9983},
 {"id"=>960, "term"=>"Miss Lucile Abbott", "score"=>9982}]

#completions is a simple wrapper for Soulmate::Matcher.new('contact').matches_for_term, and in my database there are about 2200 records.

Now I'm running the following benchmark:

# Initial implementation
def random_string(l)
  ('a'..'z').to_a.shuffle[0,l].join
end

# Already populated redis instance
matcher = Soulmate::Matcher.new('contact')
# Match Procs
cache = proc { |l| matcher.matches_for_term random_string(l), cache: true }
no_cache = proc { |l| matcher.matches_for_term random_string(l), cache: false }

Benchmark.bmbm do |x| 
  x.report('With cache on (l = 2)') { 1000.times { cache.call(2) } }
  x.report('With cache off (l = 2)') { 1000.times { no_cache.call(2) } }
  x.report('With cache on (l = 5)') { 1000.times { cache.call(5) } }
  x.report('With cache off (l = 5)') { 1000.times { no_cache.call(5) } }
end

And I'm getting these results

# with 2118 contacts

Rehearsal ----------------------------------------------------------
With cache on (l = 2)    2.630000   0.080000   2.710000 (  2.723404)
With cache off (l = 2)   2.180000   0.060000   2.240000 (  2.236404)
With cache on (l = 5)    2.600000   0.060000   2.660000 (  2.657957)
With cache off (l = 5)   1.920000   0.050000   1.970000 (  1.966270)
------------------------------------------------- total: 9.580000sec

                             user     system      total        real
With cache on (l = 2)    2.530000   0.050000   2.580000 (  2.589633)
With cache off (l = 2)   2.100000   0.050000   2.150000 (  2.148585)
With cache on (l = 5)    2.500000   0.070000   2.570000 (  2.555852)
With cache off (l = 5)   1.810000   0.040000   1.850000 (  1.858565)

With cache on the completion engine is slower than with cache turned off. Since my benchmarks use a random string and most of them don't produce results I tried a different implementation of random_string method to ensure completion results. The new implementation is the following:

# Generate a set of exisiting prefixes
def prefix_set(l)
  @prefixes_set ||= {}
  @prefixes_set[l] ||= begin
    Set.new.tap do |set|
      Contact.all.map(&:full_name).each do |name|
        # Build an entry for each prefix found in full names
        set.merge(name.split(/\W+/).reject {|s| s.size < 2 }.map { |w| w.first(l) })
      end
    end
  end
end

# Generate prefixes of size 2 from existing data
def existing_prefixes(l)
  @prefixes ||= {}
  @prefixes[l] ||= prefix_set(l).to_a
end

def random_string(l)
  existing_prefixes(l).sample
end

# warmup exisiting prefixes and remove prefixes without completions
existing_prefixes(2).reject! { |p| Contact.completions(p).empty?  }
existing_prefixes(5).reject! { |p| Contact.completions(p).empty?  }

But even with this approach performances are directly comparable:

Rehearsal ----------------------------------------------------------
With cache on (l = 2)    2.870000   0.080000   2.950000 (  2.947495)
With cache off (l = 2)   2.900000   0.070000   2.970000 (  2.986641)
With cache on (l = 5)    2.800000   0.060000   2.860000 (  2.849990)
With cache off (l = 5)   2.800000   0.060000   2.860000 (  2.865739)
------------------------------------------------ total: 11.640000sec

                             user     system      total        real
With cache on (l = 2)    2.790000   0.060000   2.850000 (  2.861272)
With cache off (l = 2)   2.810000   0.060000   2.870000 (  2.884098)
With cache on (l = 5)    2.750000   0.070000   2.820000 (  2.811039)
With cache off (l = 5)   2.730000   0.060000   2.790000 (  2.795882)

Is it worth to keep the cache functionality into the match method?

I spent the last hour or so trying to figure out why I couldn't send a cross-domain request to a Soulmate sever, even though the docs + commits said JSONP was added in May.

I finally realized that the 0.0.3 gem didn't include the commits from 2 months ago and got it working once I manually built the gem from the github repo.

Can the hosted gem be updated to 0.0.4 to include the latest commits?

It will likely save other dev's the trouble I went through.

I read that Seatgeek no longer uses Soulmate and I see that a fork of this project is born.

I find this project very useful when one needs only autocompletion and not full text search, however I'd like to know if it's worth to rely on this project anymore.

What's the current project status? Will it be updated?

Please advice

When I add Soulmate to my Gemfile and run bundle install, I get the following error:

soulmate (>= 0) x86-mingw32 depends on
  json (~> 1.4.6) x86-mingw32

rails (~> 3.1.1) x86-mingw32 depends on
  json (1.6.3)

E.g., I want the query "beef" to match the result "The Notorious B.I.G. – What's Beef?" (not just "Beef Cake")

when i come to load soulmate into my rails app give me that error uninitialized constant Soulmate.

I've noticed a couple of issues with the gemspec. The first is that soulmate depends on itself. That means that presently I cannot install soulmate using gem install soulmate. Also, jeweler is adding a lot of duplicate dependencies (in fact it adds them every time you run rake gemspec:generate.

I'm not sure what's gone on with the gemspec, but there's obviously some broken stuff there.

Redis 2.8.9 introduced a new command called ZRANGEBYLEX that soulmate could use internally to improve performances and reduce memory usage significantly.

An example of implementation of autocomplete using the new command is here: http://autocomplete.redis.org/

Note: sometimes the example is down because I've not enough memory in the virtual machine running the example.

If so, how would one do it? I'd like to get the soulmate search at a url like /sm/search...

I've tried:

mount Soulmate::Server, :at => "/sm"

In my routes file with no luck. I also have this in an initializer:

require 'soulmate/server'

Thanks!

Gem file pointing to v1.0.0 is slightly different from master branch which causes failures when assigning existing instance of Redis on initialization.

The autocomplete doesn't work with UTF-8.
For example, if I add the value "בדיקה", the autocomplete just won't work.

How can I solve this problem?

Thanks

How just show the words result instead of show on the left the term?

Add some items to the index:

$ echo '{"id":1, "term":"bun"}' | soulmate add test
Adding items of type test...
Loaded a total of 1 items
$ echo '{"id":2, "term":"buy"}' | soulmate add test
Adding items of type test...
Loaded a total of 1 items

redis> KEYS *
1) "soulmate-data:test"
2) "soulmate-index:test:buy"
3) "soulmate-index:test:bun"
4) "soulmate-index:test"
5) "soulmate-index:test:bu"

Remove one of those items:

$ echo '{"id":2, "term":"buy"}' | soulmate remove test
Removing items of type test...
Removed a total of 1 items

redis> KEYS *
1) "soulmate-data:test"
2) "soulmate-index:test:bun"
3) "soulmate-index:test"
4) "soulmate-index:test:bu"

Recreate the collection:

$ echo '{"id":2, "term":"eat"}' | soulmate load test
Loading items of type test...
Loaded a total of 1 items

redis> KEYS *
1) "soulmate-data:test"
2) "soulmate-index:test:eat"
3) "soulmate-index:test:ea"
4) "soulmate-index:test"
5) "soulmate-index:test:bu"

The key soulmate-index:test:bu was orphaned by the removal of buy.

Having massive trouble demangling all the rack, etc. to get this working with Rails 5. Anyone have a baranch/fork that will offer this support?

This is a pretty self-explanatory performance improvement.

Any way the multi json dependency can be upped to 1.1.0? Rabl and the Twitter gem both require 1.1.0 while soulmate requires 1.0.3.

Or is there another way around this that I don't know about?