4.1   Create a configuration file

SeSQL will import its configuration file with import sesql_config statements. So it should be somewhere in your PYTHONPATH, called sesql_config.py. The best place to put is usually alongside with the settings.py file.

To create it, the best is to copy the config.py.sample file (which is provided with SeSQL) as sesql_config.py on your chosen location.

If you are in Django project, you can define an alternative module from where to load the sesql_config, useful when you cannot alter your PYTHONPATH, with this setting in your project configuration:

SESQL_CONFIG_PATH = 'my_app.sesql_configuration'

4.2   Adjust the text search configuration

The sample configuration file is tuned for the french language. If you want to use another language, replace french inside simple_french and ascii_french for TS_CONFIG_NAME and STOPWORDS_FILE by the name of the language you want to use.

4.3   Provide a stopwords file

PostgreSQL (and therefore SeSQL) requires a stopwords file. Stop words are words like "the", "a", "of", ... in english, that should be ignore in searches because they are too frequent to have any value.

You've two ways to generate this stopwords file :

1. Your language is among those handled by PostgreSQL. Then, you have a file called <language>.stop in your PostgreSQL installation; for example on Debian it would be inside /usr/share/postgresql/8.4/tsearch_data/. In that case, you just need to create a plain ASCII version of the file, with a command line:

LC_ALL=fr_FR.UTF-8 iconv -f utf-8 -t ascii//TRANSLIT \
/usr/share/postgresql/8.4/tsearch_data/french.stop > \
/usr/share/postgresql/8.4/tsearch_data/ascii_french.stop

2. Your language isn't supported by PostgreSQL. In that case, you'll need to write the stopwords file yourself, just put one word per line, in pure ascii.

4.4   Create the type map

The type map has two purposes : select which models to index and which to ignore, and to group models that are often searched at the same time for better performances.

If you're not sure about it, you can just index all models into one big table. That should work very well unless you have several millions of rows in your database. If you have such a high volume, you should consider having a finer type map.

The simplest type map possible would look like:

from django.db import models
TYPE_MAP = ((models.Model, "sesql_default"),)

4.5   Define your fields

Then you need to define your fields (that is, the indexes on which you'll query and/or sort SeSQL). This is done with the FIELDS variable.

The first two items of the sample configuration file, the classname and id ones, are required and should be kept as they are.

For the rest, adjust the fields on the example to match what you want to index. Refer to the datamodel documentation for details.

4.6   Fine-tune configuration

Now, you can adjust other parameters of the configuration file. Refer to the install documentation for more details on various options you can tweak.