rawes

About

rawes is an elasticsearch driver for Python. It provides a small level of abstraction above the requests library - enough abstraction to be useful, but not so much to obscure elasticsearch's great native api

Features

• elasticsearch native API support
• Python 3 support
• gzip over HTTP support
• HTTPS support
• Thrift support

Installation

$ pip install rawes

Usage

Create a connection to elasticsearch

import rawes
es = rawes.Elastic('localhost:9200')

Search for a document

es.get('tweets/tweet/_search', data={
    'query' : {
        'match_all' : {}
    }
})

The rawes.Elastic constructor takes the following parameters (defaults shown):

rawes.Elastic(
    url='http://localhost:9200',    # Protocol, host, and port of elasticsearch service. Can be a list of hosts.
                                    # Valid protocols: http, https, thrift
                                    # Default protocol is http, unless port is in range 9500-9600, then thrift
                                    # Default ports: http=9200, https=443, thrift=9500
    timeout=30,                     # Timeout in seconds,
    **kwrgs                         # http(s) only: additional parameters you wish to pass
                                    # to the python 'requests' library (for example, basic auth)
)

Constructor examples:

es = rawes.Elastic()                        # will connect to: http://localhost:9200
es = rawes.Elastic('https://localhost')     # will connect to: https://localhost:443
es = rawes.Elastic('thrift://localhost')    # will connect to: thrift://localhost:9500
es = rawes.Elastic('https://example.org:8443', auth=('user','pass'))  # https with basic auth connection to: https://example.org:8443
es = rawes.Elastic(['http://host1:9200', 'http://host2:9200', 'http://host3:9200'])    # Each call will be issued to a different host, default is RoundRobin strategy

An instance of rawes.Elastic ('es' in this case) has methods for get, post, put, delete, and head (for each http verb). Each method takes the following parameters (defaults shown):

es.get(
    path='',    # http URL path
    data='',    # http body.  can be either a string or a python dictionary (will automatically be converted to JSON)
    params={},  # http URL params passed as a python dictionary
    headers={}, # http headers as a python dictionary
    **kwargs    # additional parameters you wish to pass to the python 'requests' library or the thrift RestRequest
                # Examples: headers, basic auth
)

Examples

Create a new document in the twitter index of type tweet with id 1

es.put('tweets/tweet/1', data={
    'user' : 'dwnoble',
    'post_date' : '2012-8-27T08:00:30Z',
    'message' : 'Tweeting about elasticsearch'
})
es.put('blogs/post/2', data={
    'user' : 'dan',
    'post_date' : '2012-8-27T09:30:03Z',
    'title' : 'Elasticsearch',
    'body' : 'Blogging about elasticsearch'
})

Search for a document, specifying http params

es.get('tweets/tweet/_search', data={
    'query' : {
        'match_all' : {}
    }
}, params= {
    'size': 2
})

Search for a document with a JSON string

es.get('tweets,blogs/_search', data="""
{
    "query" : {
        "match_all" : {}
    }
}
""")

Update a document

es.put('someindex/sometype/123', data={
    'value' : 100,
    'other' : 'stuff'
})
es.post('someindex/sometype/123/_update', data={
    'script' : 'ctx._source.value += value',
    'params' : {
        'value' : 50
    }
})

Delete a document

es.delete('tweets/tweet/1')

Bulk load

bulk_body = '''
{"index" : {}}
{"key":"value1"}
{"index" : {}}
{"key":"value2"}
{"index" : {}}
{"key":"value3"}
'''

es.post('someindex/sometype/_bulk', data=bulk_body)

bulk_list = [
    {"index" : {}},
    {"key":"value4"},
    {"index" : {}},
    {"key":"value5"},
    {"index" : {}},
    {"key":"value6"}
]

# Remember to include the trailing \n character for bulk inserts
bulk_body_2 = '\n'.join(map(json.dumps, bulk_list))+'\n'

es.post('someindex/sometype/_bulk', data=bulk_body_2)

Alternate Syntax

Instead of setting the first argument of a es.<http verb> call to the HTTP URL path, you can also use python attributes and item accessors to build up the url path. For example:

es.post('tweets/tweet/', data={
    'user' : 'dwnoble',
    'post_date' : '2012-8-27T09:15:59',
    'message' : 'More tweets about elasticsearch'
})

Becomes:

es.tweets.tweet.post(data={
    'user' : 'dwnoble',
    'post_date' : '2012-8-27T09:15:59',
    'message' : 'More tweets about elasticsearch'
})

Or using item accessors ([] notation). This can be useful for characters that are not allowed in python attributes:

es['tweets']['tweet'].post(data={
    'user' : 'dwnoble',
    'post_date' : '2012-8-27T09:15:59',
    'message' : 'More tweets about elasticsearch'
})

More examples:

Searching the "tweets" index for documents of type "tweets"

es.tweets.tweet._search.get(data={'query' : {'match_all' : {} }})

Searching the "tweets" and "blogs" index for documents of any type using a JSON strings

es['tweets,blogs']._search.get(data='{"query" : {"match_all" : {}}}')

JSON Encoding

By default, rawes will encode datetimes (timezone required!) to UTC ISO8601 strings with 'second' precision before handing the JSON off to elasticsearch. If elasticsearch has no mapping defined, this will result in the default mapping of 'dateOptionalTime.' Timezones are required for this automatic serialization: you may want to use a python module like python-dateutil (Python 2.x only) or pytz to make your life easier.

from datetime import datetime
from dateutil import tz
eastern_timezone = tz.gettz('America/New_York')

es.put('tweets/tweet/99', data={
    'user' : 'dwnoble',
    'post_date' : datetime(2012, 8, 27, 8, 0, 30, tzinfo=eastern_timezone),
    'message' : 'Tweeting about elasticsearch'
})

es.get('tweets/tweet/99')['_source']['post_date']
# Returns:
u'2012-08-27T12:00:30Z'

Alternatively, you can specify a custom JSON encoder using the json_encoder parameter:

from datetime import datetime
from dateutil import tz
eastern_timezone = tz.gettz('America/New_York')

def encode_custom(obj):
    if isinstance(obj, datetime):
        return obj.astimezone(tz.tzutc()).strftime('%Y-%m-%d')
    raise TypeError(repr(obj) + " is not JSON serializable")

es.put('tweets/tweet/445', data={
    'user' : 'dwnoble',
    'post_date' : datetime(2012, 11, 12, 9, 45, 45, tzinfo=eastern_timezone),
    'message' : 'Tweeting about elasticsearch'
}, json_encoder=encode_custom)

es.get('tweets/tweet/445')['_source']['post_date']
# Returns:
u'2012-11-12'

Additionally, a default JSON encoder can be specified in the rawes.Elastic constructor:

import rawes
from datetime import datetime
from dateutil import tz
eastern_timezone = tz.gettz('America/New_York')

def encode_custom(obj):
    if isinstance(obj, datetime):
        return obj.astimezone(tz.tzutc()).strftime('%Y-%m-%d')
    raise TypeError(repr(obj) + " is not JSON serializable")

es = rawes.Elastic("http://localhost:9200", json_encoder=encode_custom)

es.put('tweets/tweet/445', data={
    'user' : 'dwnoble',
    'post_date' : datetime(2012, 11, 12, 9, 45, 45, tzinfo=eastern_timezone),
    'message' : 'Tweeting about elasticsearch'
})

es.get('tweets/tweet/445')['_source']['post_date']
# Returns:
u'2012-11-12'

JSON Decoding

Like with JSON encoding, a custom JSON decoder may be specified as well to parse elasticsearch results. A common use case here may be parsing ISO8601 dates to python datetime objects.

Index a document with a ISO8601 formatted date string:

import rawes
es = rawes.Elastic()

es.put('blogs/post/3', data={
    'user' : 'dan',
    'post_date' : '2013-7-04T23:14:53Z',
    'title' : 'Elasticsearch 2',
    'body' : 'More blogging about elasticsearch'
})

Define a custom JSON decoder:

import json
import pytz
import dateutil.parser

class Iso8601JsonDecoder(json.JSONDecoder):
    """
    Automatically decode ISO8601 strings with key "post_date" to python datetime objects in UTC timezone
    """
    def __init__(self):
        json.JSONDecoder.__init__(self, object_hook=self.dict_to_object)

    def dict_to_object(self, d):
        for k,v in d.iteritems():
            if k == "post_date":
                d[k] = dateutil.parser.parse(v)
        return d


iso8601_json_decoder = Iso8601JsonDecoder()

Now retrieve this document using our JSON decoder

es.get("blogs/post/3")["_source"]["post_date"]
# returns:
# u'2013-7-04T23:14:53Z'
es.get("blogs/post/3",json_decoder=iso8601_json_decoder.decode)["_source"]["post_date"]
# returns:
# datetime.datetime(2013, 7, 4, 23, 14, 53, tzinfo=tzutc())
es_default_decoder = rawes.Elastic(json_decoder=iso8601_json_decoder.decode)
es_default_decoder.get("blogs/post/3")["_source"]["post_date"]
# returns:
# datetime.datetime(2013, 7, 4, 23, 14, 53, tzinfo=tzutc())

Error Handling

As of version 0.5, the rawes.Elastic constructor throws a rawes.elastic_exception.ElasticException any time elasticsearch returns an http status code of 400 or greater.

from rawes.elastic_exception import ElasticException
es = rawes.Elastic('localhost:9200')
try:
    es.get('invalid_index/invalid_type/123')
except ElasticException as e:
    # since our index is invalid, this exception handler will run
    print e.result
    # prints: {u'status': 404, u'error': u'IndexMissingException[[invalid_index] missing]'}
    print e.status_code
    # prints: 404

Thrift support

Thrift is supported for Python 2.x versions only. Before thrift will work with rawes, you must install the thrift python module

$ pip install thrift

By default, connections on ports between 9500 and 9600 will use thrift

import rawes
es_thrift = rawes.Elastic('localhost:9500')

If you are using thrift on a non standard port, specify a 'thrift://' url

import rawes
es_thrift = rawes.Elastic('thrift://localhost:8500')

Connection Pooling

rawes supports connection pooling of elasticsearch hosts:

import rawes
es = rawes.Elastic(['http://host1:9200', 'http://host2:9200', 'http://host3:9200'])

The class cycles through these hosts in a round robin fashion with each request method call. You can change the connection pool strategy, or other connection pool parameters, by passing options to the rawes.Elastic constructor's connection_pool_kwargs argument:

import rawes
es = rawes.Elastic(['http://host1:9200', 'http://host2:9200', 'http://host3:9200'], connection_pool_kwargs={
    "selector_class" : rawes.connection_pool.RandomSelector
})

See rawes/connection_pool.py for details on all available parameters.

The conncetion pooling implementation used is from the elasticsearch-py project.

Run Unit Tests

rawes' unit tests require the python thrift and mock modules to run:

$ pip install thrift
$ pip install mock

Run tests:

$ python -m unittest tests

Run tests for Python 3 (no thrift tests, no need to "pip install thrift")

$ python3 -m unittest tests.py3k

License

Apache 2.0 License

Contact

@dwnoble