Installing Search Engines


Official Download Location:

Solr is Java but comes in a pre-packaged form that requires very little other than the JRE and Jetty. It’s very performant and has an advanced featureset. Haystack suggests using Solr 6.x, though it’s possible to get it working on Solr 4.x+ with a little effort. Installation is relatively simple:

For Solr 6.X:

curl -LO
tar -C solr -xf solr-X.Y.0.tgz --strip-components=1
cd solr
./bin/solr create -c tester -n basic_config

By default this will create a core with a managed schema. This setup is dynamic but not useful for haystack, and we’ll need to configure solr to use a static (classic) schema. Haystack can generate a viable schema.xml and solrconfig.xml for you from your application and reload the core for you (once Haystack is installed and setup). To do this run: ./ build_solr_schema --configure-directory=<CoreConfigDif> --reload-core. In this example CoreConfigDir is something like ../solr-6.5.0/server/solr/tester/conf, and --reload-core is what triggers reloading of the core. Please refer to build_solr_schema in the management-commands for required configuration.

For Solr 4.X:

curl -LO
tar xvzf solr-4.10.2.tgz
cd solr-4.10.2
cd example
java -jar start.jar

You’ll need to revise your schema. You can generate this from your application (once Haystack is installed and setup) by running ./ build_solr_schema. Take the output from that command and place it in solr-4.10.2/example/solr/collection1/conf/schema.xml. Then restart Solr.


Please note; the template filename, the file YOU supply under TEMPLATE_DIR/search_configuration has changed to schema.xml from solr.xml. The previous template name solr.xml was a legacy holdover from older versions of solr.

You’ll also need a Solr binding, pysolr. The official pysolr package, distributed via PyPI, is the best version to use (2.1.0+). Place somewhere on your PYTHONPATH.


pysolr has its own dependencies that aren’t covered by Haystack. See for the latest documentation. Simplest approach is to install using pip install pysolr

More Like This

To enable the “More Like This” functionality in Haystack, you’ll need to enable the MoreLikeThisHandler. Add the following line to your solrconfig.xml file within the config tag:

<requestHandler name="/mlt" class="solr.MoreLikeThisHandler" />

Spelling Suggestions

To enable the spelling suggestion functionality in Haystack, you’ll need to enable the SpellCheckComponent.

The first thing to do is create a special field on your SearchIndex class that mirrors the text field, but uses FacetCharField. This disables the post-processing that Solr does, which can mess up your suggestions. Something like the following is suggested:

class MySearchIndex(indexes.SearchIndex, indexes.Indexable):
    text = indexes.CharField(document=True, use_template=True)
    # ... normal fields then...
    suggestions = indexes.FacetCharField()

    def prepare(self, obj):
        prepared_data = super(MySearchIndex, self).prepare(obj)
        prepared_data['suggestions'] = prepared_data['text']
        return prepared_data

Then, you enable it in Solr by adding the following line to your solrconfig.xml file within the config tag:

<searchComponent name="spellcheck" class="solr.SpellCheckComponent">

  <str name="queryAnalyzerFieldType">text_general</str>
  <lst name="spellchecker">
    <str name="name">default</str>
    <str name="field">text</str>
    <str name="classname">solr.DirectSolrSpellChecker</str>
    <str name="distanceMeasure">internal</str>
    <float name="accuracy">0.5</float>
    <int name="maxEdits">2</int>
    <int name="minPrefix">1</int>
    <int name="maxInspections">5</int>
    <int name="minQueryLength">4</int>
    <float name="maxQueryFrequency">0.01</float>

Then change your default handler from:

<requestHandler name="/select" class="solr.SearchHandler">
  <lst name="defaults">
    <str name="echoParams">explicit</str>
    <int name="rows">10</int>

... to ...:

<requestHandler name="/select" class="solr.SearchHandler">
  <lst name="defaults">
    <str name="echoParams">explicit</str>
    <int name="rows">10</int>

    <str name="spellcheck.dictionary">default</str>
    <str name="spellcheck">on</str>
    <str name="spellcheck.extendedResults">true</str>
    <str name="spellcheck.count">10</str>
    <str name="spellcheck.alternativeTermCount">5</str>
    <str name="spellcheck.maxResultsForSuggest">5</str>
    <str name="spellcheck.collate">true</str>
    <str name="spellcheck.collateExtendedResults">true</str>
    <str name="spellcheck.maxCollationTries">10</str>
    <str name="spellcheck.maxCollations">5</str>
   <arr name="last-components">

Be warned that the <str name="field">suggestions</str> portion will be specific to your SearchIndex classes (in this case, assuming the main field is called text).


Official Download Location:

Elasticsearch is Java but comes in a pre-packaged form that requires very little other than the JRE. It’s also very performant, scales easily and has an advanced featureset. Haystack currently only supports Elasticsearch 1.x and 2.x. Elasticsearch 5.x is not supported yet, if you would like to help, please see #1383.

Installation is best done using a package manager:

# On Mac OS X...
brew install elasticsearch

# On Ubuntu...
apt-get install elasticsearch

# Then start via:
elasticsearch -f -D es.config=<path to YAML config>

# Example:
elasticsearch -f -D es.config=/usr/local/Cellar/elasticsearch/0.90.0/config/elasticsearch.yml

You may have to alter the configuration to run on localhost when developing locally. Modifications should be done in a YAML file, the stock one being config/elasticsearch.yml:

# Unicast Discovery (disable multicast) false [""]

# Name your cluster here to whatever.
# My machine is called "Venus", so...
  name: venus


  logs: /usr/local/var/log
  data: /usr/local/var/data

You’ll also need an Elasticsearch binding: elasticsearch (NOT pyes). Place elasticsearch somewhere on your PYTHONPATH (usually python install or pip install elasticsearch).


elasticsearch has its own dependencies that aren’t covered by Haystack. You’ll also need urllib3.


Official Download Location:

Whoosh is pure Python, so it’s a great option for getting started quickly and for development, though it does work for small scale live deployments. The current recommended version is 1.3.1+. You can install via PyPI using sudo easy_install whoosh or sudo pip install whoosh.

Note that, while capable otherwise, the Whoosh backend does not currently support “More Like This” or faceting. Support for these features has recently been added to Whoosh itself & may be present in a future release.


Official Download Location:

Xapian is written in C++ so it requires compilation (unless your OS has a package for it). Installation looks like:

curl -O
curl -O

unxz xapian-core-1.2.18.tar.xz
unxz xapian-bindings-1.2.18.tar.xz

tar xvf xapian-core-1.2.18.tar
tar xvf xapian-bindings-1.2.18.tar

cd xapian-core-1.2.18
sudo make install

cd ..
cd xapian-bindings-1.2.18
sudo make install

Xapian is a third-party supported backend. It is not included in Haystack proper due to licensing. To use it, you need both Haystack itself as well as xapian-haystack. You can download the source from Installation instructions can be found on that page as well. The backend, written by David Sauve (notanumber), fully implements the SearchQuerySet API and is an excellent alternative to Solr.