Query Methods

Although most operations you will do involve directly interacting with known buckets and keys, there are additional ways to get information out of Riak.

Secondary Indexes

Objects can be tagged with secondary index entries. Those entries can then be queried over the bucket for equality or across ranges.:

bucket = client.bucket("index_test")

# Tag an object with indexes and save
sean = bucket.new("seancribbs")
sean.add_index("fname_bin", "Sean")
sean.add_index("byear_int", 1979)
sean.store()

# Performs an equality query
seans = bucket.get_index("fname_bin", "Sean")

# Performs a range query
eighties = bucket.get_index("byear_int", 1980, 1989)

Secondary indexes are also available via MapReduce.

Streaming and Paginating Indexes

Sometimes the number of results from such a query is too great to process in one payload, so you can also stream the results:

for keys in bucket.stream_index("bmonth_int", 1):
    # keys is a list of matching keys
    print(keys)

Both the regular get_index() method and the stream_index() method allow you to return the index entry along with the matching key as tuples using the return_terms option:

bucket.get_index("byear_int", 1970, 1990, return_terms=True)
# => [(1979, 'seancribbs')]

You can also limit the number of results using the max_results option, which enables pagination:

results = bucket.get_index("fname_bin", "S", "T", max_results=20)

Optionally you can use paginate_index() or paginate_stream_index() to create a generator of paged results:

for page in bucket.paginate_stream_index("maestro_bin", "Cribbs"):
    for key in page:
        do_something(key)
    page.close()

All of these features are implemented using the IndexPage class, which emulates a list but also supports streaming and capturing the continuation, which is a sort of pointer to the next page of results:

# Detect whether there are more results
if results.has_next_page():

    # Fetch the next page of results manually
    more = bucket.get_index("fname_bin", "S", "T", max_results=20,
                            continuation=results.continuation)

    # Fetch the next page of results automatically
    more = results.next_page()

MapReduce

RiakMapReduce allows you to construct query-processing jobs that are performed mostly in-parallel around the Riak cluster. You can think of it as a pipeline, where inputs are fed in one end, they pass through a number of map and reduce phases, and then are returned to the client.

Constructing the query

Inputs

The first step is to identify the inputs that should be processed. They can be:

1. An entire bucket
2. An entire bucket, with the keys filtered by criteria
3. A list of bucket/key pairs or bucket/key/data triples
4. A fulltext search query
5. A secondary-index query

Adding inputs always returns the RiakMapReduce object so that you can chain the construction of the query job.

Phases

The second step is to add processing phases to the query. map phases load and process individual keys, returning one or more results, while reduce phases operate over collections of results from previous phases. link phases are a special type of map phase that extract matching links from the object, usually so they can be used in a subsequent map phase.

Any number of phases can return results directly to the client by passing keep=True.

Phase shortcuts

A number of commonly-used phases are also available as shortcut methods:

Execution

Query results can either be executed in one round-trip, or streamed back to the client. The format of results will depend on the structure of the map and reduce phases the query contains.

Shortcut constructors

RiakObject contains some shortcut methods that make it more convenient to begin constructing RiakMapReduce queries.

Riak Search 2.0 (Yokozuna)

With Riak 2.0 came the introduction of Riak Search 2.0, a.k.a Yokozuna (the top rank in sumo). Riak Search 2.0 is an integration of Solr (for indexing and querying) and Riak (for storage and distribution). It allows for distributed, scalable, fault-tolerant, transparent indexing and querying of Riak values. After connecting a bucket (or bucket type) to a Apache Solr index, you simply write values (such as JSON, XML, plain text, Data Types, etc.) into Riak as normal, and then query those indexed values using the Solr API. Unlike traditional Riak data, however, Solr needs to know the format of the stored data so it can index it. Solr is a document-based search engine so it treats each value stored in Riak as a document.

Creating a schema

The first thing which needs to be done is to define a Solr schema for your data. Riak Search comes bundled with a default schema named _yz_default. It defaults to many dynamic field types, where the suffix defines its type. This is an easy path to start development, but we recommend in production that you define your own schema.

You can find information about defining your own schema at Search Schema, with a short section dedicated to the default schema.

Here is a brief example of creating a custom schema with create_search_schema():

content = """<?xml version="1.0" encoding="UTF-8" ?>
<schema name="test" version="1.5">
<fields>
   <field name="_yz_id" type="_yz_str" indexed="true" stored="true"
    multiValued="false" required="true" />
   <field name="_yz_ed" type="_yz_str" indexed="true" stored="true"
    multiValued="false" />
   <field name="_yz_pn" type="_yz_str" indexed="true" stored="true"
    multiValued="false" />
   <field name="_yz_fpn" type="_yz_str" indexed="true" stored="true"
    multiValued="false" />
   <field name="_yz_vtag" type="_yz_str" indexed="true" stored="true"
    multiValued="false" />
   <field name="_yz_rk" type="_yz_str" indexed="true" stored="true"
    multiValued="false" />
   <field name="_yz_rb" type="_yz_str" indexed="true" stored="true"
    multiValued="false" />
   <field name="_yz_rt" type="_yz_str" indexed="true" stored="true"
    multiValued="false" />
   <field name="_yz_err" type="_yz_str" indexed="true"
    multiValued="false" />
</fields>
<uniqueKey>_yz_id</uniqueKey>
<types>
    <fieldType name="_yz_str" class="solr.StrField"
     sortMissingLast="true" />
</types>
</schema>"""
schema_name = 'jalapeno'
client.create_search_schema(schema_name, content)

If you would like to retrieve the current XML Solr schema, get_search_schema() is available:

schema = client.get_search_schema('jalapeno')

Solr indexes

Once a schema has been created, then a Solr index must also be created. This index represents a collection of similar data that you use to perform queries. When creating an index with create_search_index(), you can optionally specify a schema. If you do not, the default schema will be used:

client.create_search_index('nacho')

Likewise you can specify a schema, e.g. the index "nacho" is associated with the schema "jalapeno":

client.create_search_index('nacho', 'jalapeno')

Just as easily you can delete an index with delete_search_index():

client.delete_search_index('jalapeno')

A single index can be retrieved with get_search_index() or all of them with list_search_indexes():

index = client.get_search_index('jalapeno')
name = index['name']
schema = index['schema']
indexes = client.list_search_indexes()
first_nval = indexes[0]['n_val']

Note

Note that index names may only be ASCII values from 32-127 (spaces, standard punctuation, digits and word characters). This may change in the future to allow full unicode support.

More discussion about Riak Search 2.0 Indexes can be found at Indexes.

Linking a bucket type to an index

The last step to setting up Riak Search 2.0 is to link a Bucket Type to a Solr index. This lets Riak know when to index values. This can be done via the command line:

riak-admin bucket-type create spicy '{"props":{"search_index":"jalapeno"}}'
riak-admin bucket-type activate spicy

Or simply create an empty Bucket Type:

riak-admin bucket-type create spicy '{"props":{}}'
riak-admin bucket-type activate spicy

Then change the bucket properties on the associated bucket or Bucket Type:

b = client.bucket('peppers')
b.set_property('search_index', 'jalapeno')
btype = client.bucket_type('spicy')
btype.set_property('search_index', 'jalapeno')

Querying an index

Once the schema, index and bucket properties have all been properly configured, adding data is as simple as writing to Riak. Solr is automatically updated.

To query, on the other hand, is as easy as writing Solr queries. This allows for the full use of existing Solr tools as well as its rich semantics.

Here is a brief example of loading and querying data::

bucket = self.client.bucket('peppers')
bucket.new("bell", {"name_s": "bell", "scoville_low_i": 0,
                    "scoville_high_i": 0}).store()
bucket.new("anaheim", {"name_s": "anaheim", "scoville_low_i": 1000,
                       "scoville_high_i": 2500}).store()
bucket.new("chipotle", {"name_s": "chipotle", "scoville_low_i": 3500,
                        "scoville_high_i": 10000}).store()
bucket.new("serrano", {"name_s": "serrano", "scoville_low_i": 10000,
                       "scoville_high_i": 23000}).store()
bucket.new("habanero", {"name_s": "habanero", "scoville_low_i": 100000,
                        "scoville_high_i": 350000}).store()
results = bucket.search("name_s:/c.*/", index='jalapeno')
# Yields single document 'chipotle'
print(results['docs'][0]['name_s'])
results = bucket.search("scoville_high_i:[20000 TO 500000]")
# Yields two documents
for result in results['docs']:
    print(result['name_s'])
results = bucket.search('name_s:*', index='jalapeno',
                        sort="scoville_low_i desc")
# Yields all documents, sorted in descending order. We take the top one
print("The hottest pepper is {0}".format(results['docs'][0]['name_s']))

The results returned by search() is a dictionary with lots of search metadata like the number of results, the maxium Lucene Score as well as the matching documents.

When querying on Data Types the datatype is the name of the field used in Solr since they do not fit into the default schema, e.g.:

riak-admin bucket-type create visitors '{"props":{"datatype": "counter}}'
riak-admin bucket-type activate visitors

client.create_search_index('website')
bucket = client.bucket_type('visitors').bucket('hits')
bucket.set_property('search_index', 'website')

site = bucket.new('bbc.co.uk')
site.increment(80)
site.store()
site = bucket.new('cnn.com')
site.increment(150)
site.store()
site = bucket.new('abc.net.au')
site.increment(24)
site.store()

results = bucket.search("counter:[10 TO *]", index='website',
                        sort="counter desc", rows=5)

# Assume you have a bucket-type named "profiles" that has datatype
# "map". Let's create and search an index containing maps.
client.create_search_index('user-profiles')
bucket = client.bucket_type('profiles').bucket('USA')
bucket.set_property('search_index', 'user-profiles')

brett = bucket.new()
brett.registers['fname'].assign("Brett")
brett.registers['lname'].assign("Hazen")
brett.sets['emails'].add('spam@basho.com')
brett.counters['visits'].increment()
brett.maps['pages'].counters['homepage'].increment()
brett.update()

# Note that the field name in the index/schema is the field name in
# the map joined with its type by an underscore. Deeply embedded
# fields are joined with their parent field names by an underscore.
results = bucket.search('lname_register:Hazen AND pages_map_homepage_counter:[1 TO *]',
                        index='user-profiles')

Details on querying Riak Search 2.0 can be found at Querying.