redisearch

RediSearch Python Client

Deprecation notice

As of redis-py 4.0.0 this library is deprecated. It's features have been merged into redis-py. Please either install it from pypy or the repo.

---

This is a Python search engine library that utilizes the RediSearch Redis Module API.

It is the "official" client of RediSearch, and should be regarded as its canonical client implementation.

Features

RediSearch is a source avaliable (RSAL), high performance search engine implemented as a Redis Module. It uses custom data types to allow fast, stable and feature rich full-text search inside Redis.

This client is a wrapper around the RediSearch API protocol, that allows you to utilize its features easily.

RediSearch's features include:

• Full-Text indexing of multiple fields in documents.
• Incremental indexing without performance loss.
• Document ranking (provided manually by the user at index time) and field weights.
• Auto-complete suggestions (with fuzzy prefix suggestions).
• Exact Phrase Search.
• Stemming based query expansion in many languages (using Snowball).
• Limiting searches to specific document fields (up to 8 fields supported).
• Numeric filters and ranges.
• Automatically index existing HASH keys as documents.

For more details, visit http://redisearch.io

Examples

Creating a client instance

When you create a redisearch-py client instance, the only required argument is the name of the index.

from redisearch import Client

client = Client("my-index")

To connect with a username and/or password, pass those options to the client initializer.

client = Client("my-index", password="my-password")

Using core Redis commands

Every instance of Client contains an instance of the redis-py Client as well. Use this object to run core Redis commands.

import datetime

from redisearch import Client

START_TIME = datetime.datetime.now().strftime("%Y-%m-%d-%H:%M.%S")

client = Client("my-index")

client.redis.set("start-time", START_TIME)

Checking if a RediSearch index exists

To check if a RediSearch index exists, use the FT.INFO command and catch the ResponseError raised if the index does not exist.

from redis import ResponseError
from redisearch import Client

client = Client("my-index")

try:
    client.info()
except ResponseError:
    # Index does not exist. We need to create it!

Defining a search index

Use an instance of IndexDefinition to define a search index. You only need to do this when you create an index.

RediSearch indexes follow Hashes in your Redis databases by watching key prefixes. If a Hash whose key starts with one of the search index's configured key prefixes is added, updated, or deleted from Redis, RediSearch will make those changes in the index. You configure a search index's key prefixes using the prefix parameter of the IndexDefinition initializer.

NOTE: Once you create an index, RediSearch will continuously index these keys when their Hashes change.

IndexDefinition also takes a schema. The schema specifies which fields to index from within the Hashes that the index follows. The field types are:

• TextField
• TagField
• NumericField
• GeoField

For more information on what these field types mean, consult the RediSearch documentation on the FT.CREATE command.

With redisearch-py, the schema is an iterable of Field instances. Once you have an IndexDefinition instance, you can create the instance by passing a schema iterable to the create_index() method.

from redis import ResponseError
from redisearch import Client, IndexDefinition, TextField

SCHEMA = (
    TextField("title", weight=5.0),
    TextField("body")
)

client = Client("my-index")

definition = IndexDefinition(prefix=['blog:'])

try:
    client.info()
except ResponseError:
    # Index does not exist. We need to create it!
    client.create_index(SCHEMA, definition=definition)

Indexing a document

A RediSearch 2.0 index continually follows Hashes with the key prefixes you defined, so if you want to add a document to the index, you only need to create a Hash with one of those prefixes.

# Indexing a document with RediSearch 2.0.
doc = {
    'title': 'RediSearch',
    'body': 'Redisearch adds querying, indexing, and full-text search to Redis'
}
client.redis.hset('doc:1', mapping=doc)

Past versions of RediSearch required that you call the add_document() method. This method is deprecated, but we include its usage here for reference.

# Indexing a document for RediSearch 1.x
client.add_document(
    "doc:2",
    title="RediSearch",
    body="Redisearch implements a search engine on top of redis",
)

Querying

Basic queries

Use the search() method to perform basic full-text and field-specific searches. This method doesn't take many of the options available to the RediSearch FT.SEARCH command -- read the section on building complex queries later in this document for information on how to use those.

res = client.search("evil wizards")

Result objects

Results are wrapped in a Result object that includes the number of results and a list of matching documents.

>>> print(res.total)
2
>>> print(res.docs[0].title)
"Wizard Story 2: Evil Wizards Strike Back"

Building complex queries

You can use the Query object to build complex queries:

q = Query("evil wizards").verbatim().no_content().with_scores().paging(0, 5)
res = client.search(q)

For an explanation of these options, see the RediSearch documentation for the FT.SEARCH command.

Query syntax

The default behavior of queries is to run a full-text search across all TEXT fields in the index for the intersection of all terms in the query.

So the example given in the "Basic queries" section of this README, client.search("evil wizards"), run a full-text search for the intersection of "evil" and "wizard" in all TEXT fields.

Many more types of queries are possible, however! The string you pass into the search() method or Query() initializer has the full range of query syntax available in RediSearch.

For example, a full-text search against a specific TEXT field in the index looks like this:

# Full-text search
res = client.search("@title:evil wizards")

Finding books published in 2020 or 2021 looks like this:

client.search("@published_year:[2020 2021]")

To learn more, see the RediSearch documentation on query syntax.

Aggregations

This library contains a programmatic interface to run aggregation queries with RediSearch.

Making an aggregation query

To make an aggregation query, pass an instance of the AggregateRequest class to the search() method of an instance of Client.

For example, here is what finding the most books published in a single year looks like:

from redisearch import Client
from redisearch import reducers
from redisearch.aggregation import AggregateRequest

client = Client('books-idx')

request = AggregateRequest('*').group_by(
    '@published_year', reducers.count().alias("num_published")
).group_by(
    [], reducers.max("@num_published").alias("max_books_published_per_year")
)

result = client.aggregate(request)

A redis-cli equivalent query

The aggregation query just given is equivalent to the following FT.AGGREGATE command entered directly into the redis-cli:

FT.AGGREGATE books-idx *
    GROUPBY 1 @published_year
      REDUCE COUNT 0 AS num_published
    GROUPBY 0
      REDUCE MAX 1 @num_published AS max_books_published_per_year

The AggregateResult object

Aggregation queries return an AggregateResult object that contains the rows returned for the query and a cursor if you're using the cursor API.

from redisearch.aggregation import AggregateRequest, Asc

request = AggregateRequest('*').group_by(
    ['@published_year'], reducers.avg('average_rating').alias('average_rating_for_year')
).sort_by(
    Asc('@average_rating_for_year')
).limit(
    0, 10
).filter('@published_year > 0')

...


In [53]: resp = c.aggregate(request)
In [54]: resp.rows
Out[54]:
[['published_year', '1914', 'average_rating_for_year', '0'],
 ['published_year', '2009', 'average_rating_for_year', '1.39166666667'],
 ['published_year', '2011', 'average_rating_for_year', '2.046'],
 ['published_year', '2010', 'average_rating_for_year', '3.125'],
 ['published_year', '2012', 'average_rating_for_year', '3.41'],
 ['published_year', '1967', 'average_rating_for_year', '3.603'],
 ['published_year', '1970', 'average_rating_for_year', '3.71875'],
 ['published_year', '1966', 'average_rating_for_year', '3.72666666667'],
 ['published_year', '1927', 'average_rating_for_year', '3.77']]

Reducer functions

Notice from the example that we used an object from the reducers module. See the RediSearch documentation for more examples of reducer functions you can use when grouping results.

Reducer functions include an alias() method that gives the result of the reducer a specific name. If you don't supply a name, RediSearch will generate one.

Grouping by zero, one, or multiple fields

The group_by statement can take a single field name as a string, or multiple field names as a list of strings.

AggregateRequest('*').group_by('@published_year', reducers.count())

AggregateRequest('*').group_by(
    ['@published_year', '@average_rating'],
    reducers.count())

To run a reducer function on every result from an aggregation query, pass an empty list to group_by(), which is equivalent to passing the option GROUPBY 0 when writing an aggregation in the redis-cli.

AggregateRequest('*').group_by([], reducers.max("@num_published"))

NOTE: Aggregation queries require at least one group_by() method call.

Sorting and limiting

Using an AggregateRequest instance, you can sort with the sort_by() method and limit with the limit() method.

For example, finding the average rating of books published each year, sorting by the average rating for the year, and returning only the first ten results:

from redisearch import Client
from redisearch.aggregation import AggregateRequest, Asc

c = Client()

request = AggregateRequest('*').group_by(
    ['@published_year'], reducers.avg('average_rating').alias('average_rating_for_year')
).sort_by(
    Asc('@average_rating_for_year')
).limit(0, 10)

c.aggregate(request)

NOTE: The first option to limit() is a zero-based offset, and the second option is the number of results to return.

Filtering

Use filtering to reject results of an aggregation query after your reducer functions run. For example, calculating the average rating of books published each year and only returning years with an average rating higher than 3:

from redisearch.aggregation import AggregateRequest, Asc

req = AggregateRequest('*').group_by(
    ['@published_year'], reducers.avg('average_rating').alias('average_rating_for_year')
).sort_by(
    Asc('@average_rating_for_year')
).filter('@average_rating_for_year > 3')

Installing

1. Install RediSearch
2. Install the Python client:

$ pip install redisearch

Developing

1. Create a virtualenv to manage your python dependencies, and ensure it's active. virtualenv -v venv
2. Install pypoetry to manage your dependencies. pip install --user poetry
3. Install dependencies. poetry install

Note: Due to an interaction between and python 3.10, you may need to run the following, if you receive a JSONError while installing packages.

poetry config experimental.new-installer false

Testing

Testing can easily be performed using using Docker. Run the following:

make -C test/docker test PYTHON_VER=3

(Replace PYTHON_VER=3 with PYTHON_VER=2 to test with Python 2.7.)

Alternatively, use the following procedure:

First, run:

PYTHON_VER=3 ./test/test-setup.sh

This will set up a Python virtual environment in venv3 (or in venv2 if PYTHON_VER=2 is used).

Afterwards, run RediSearch in a container as a daemon:

docker run -d -p 6379:6379 redislabs/redisearch:2.0.0

Finally, invoke the virtual environment and run the tests:

. ./venv3/bin/activate
REDIS_PORT=6379 python test/test.py
REDIS_PORT=6379 python test/test_builder.py