RETS

A pure python RETS client for real estate data. Make requests to the MLS server to get real estate listings, media, and metadata.

Installation

The easiest way to install is through pip. pip install rets

If you need to build the package locally, it can be downloaded from github and installed through setuptools.

git clone https://github.com/refindlyllc/rets.git
cd python-rets
python setup.py install

You can now import the rets module within Python.

Quickstart

After installing the rets package locally, make requests to an MLS server for data.

>>> from rets import Session
>>> login_url = 'http://retrsprovider.com/login'
>>> username = 'user123'
>>> password = 'a48a*32fa$5'
>>> rets_client = Session(login_url, username, password)
>>> rets_client.login()
>>> system_data = rets_client.get_system_metadata()
>>> system_data
{'version': '1.11.76004', 'system_description': 'MLS-RETS', 'system_id': 'MLS-RETS'}
>>> resources = rets_client.get_resource_metadata((resource='Agent')
>>> resources
{'ClassCount': '1',
 'ClassDate': '2016-04-20T15:17:13Z',
 'ClassVersion': '1.00.00023',
 'Date': '2016-12-08T16:15:15Z',
 'Description': 'Agent',
 'EditMaskDate': '2013-03-26T00:10:01Z',
 'EditMaskVersion': '1.00.00000',
 'KeyField': 'unique_id',
 'LookupDate': '2016-05-06T17:05:40Z',
 'LookupVersion': '1.00.00369',
 'ObjectDate': '2014-06-20T14:15:57Z',
 'ObjectVersion': '1.00.00001',
 'ResourceID': 'Agent',
 'SearchHelpDate': '2013-03-26T00:10:01Z',
 'SearchHelpVersion': '1.00.00000',
 'StandardName': 'Agent',
 'TableName': 'AGENT',
 'UpdateHelpDate': '2013-03-26T00:10:01Z',
 'UpdateHelpVersion': '1.00.00000',
 'ValidationExpressionDate': '2013-03-26T00:10:01Z',
 'ValidationExpressionVersion': '1.00.00000',
 'ValidationExternalDate': '2013-03-26T00:10:01Z',
 'ValidationExternalVersion': '1.00.00000',
 'ValidationLookupDate': '2013-03-26T00:10:01Z',
 'ValidationLookupVersion': '1.00.00000',
 'Version': '1.11.73255',
 'VisibleName': 'Agent'}

>>> search_results = rets_client.search(resource='Property', resource_class='RES', limit=1, dmql_query='(ListPrice=150000+)')
>>> for result in search_results:
...     result

 {'Acres': '0.0000',
  'ActiveOpenHouseCount': '',
  'AdditionalRooms': 'LAINRE,SCPOLA',
  'AmenRecFreq': '',
  'Amenities': 'BASKET,CLUBHO,COMPOO,COMSPA,EXEROO,EXTSTO,PRIMEM,PUTGRE,SAUNA,SIDEWA,STREET,TENCOU,UNDUTI',
  'AmenityRecFee': '0.00',
  'ApplicationFee': '100.00',
  'ApproxLivingArea': '1946',
  'AssociationMngmtPhone': '',
  'BathsFull': '2',
  'BathsHalf': '0',
  'BathsTotal': '2.00',
  'BedroomDesc': '',
  'Bedrooms': '3',
  'BedsTotal': '3',
  ...
   }
>>> rets_client.logout()

The Session Object

All requests to a RETS server must be authenticated. The login credential fields must be passed to the Session object at instantiation. As some RETS servers limit the number of concurrent requests, it is also ideal to logout when requests to the RETS server are complete.

Session Parameters

• login_url: The login URL for the RETS feed
• username: The username for the RETS feed
• password: The password for the RETS feed
• version: The RETS version is typically provided from the server at login. You can set the version here to override the value provided by the server
• user_agent: The useragent for the RETS feed. Not all servers require this.
• user_agent_password: The useragent password for the RETS feed. Not all servers require this.
• follow_redirects: Follow HTTP redirects. The default True.
• use_post_method: Use HTTP POST method when making requests instead of GET. The default is True
• metadata_format: COMPACT_DECODED or STANDARD_XML. The client will attempt to set this automatically based on response codes from the RETS server.
• session_id_cookie_name: The session cookie name returned by the RETS server. Default is RETS-Session-ID

Context Manager

If you don't want to manually call the session's login and logout methods, the Session object can be opened in a context manager that logs the client in and out automatically.

with Session(rets_client = Session(login_url, username, password) as s:
    print('Now logged in')
    system_metadata = s.get_system_metadata()
    search_results = s.search(resource='Property', resource_class='RES', limit=100, dmql_query='(ListPrice=150000+)')
print('Now logged out')
# do stuff with the search results

Metadata Methods

The session object can get RETS metadata through the following methods:

rets_client.get_system_metadata()

Returns the METADATA-SYSTEM information in a dictionary.

rets_client.get_resource_metadata(resource=None)

Returns the METADATA-RESOURCE information in a list of dicts. The resource argument can be supplied to this method to limit the returned value to just the dict containing that resource.

rets_client.get_class_metadata(resource)

Returns the METADATA-CLASS information for a given resource in a list of dicts.

rets_client.get_table_metadata(resource, class)

Returns the METADATA-TABLE information for a resource and class in a list of dicts.

rets_client.get_object_metadata(resource)

Returns the METADATA-OBJECT information for a resource in a list of dicts

rets_client.get_lookup_values(resource, lookup_name)

Returns the METADATA-LOOKUP_TYPE information for a field of a resource

Object Methods

The session can get RETS Objects through the GetObject request. There are two methods for obtaining objects.

rets_client.get_preferred_object(resource, object_type, content_id, location=0)

Returns a dict containing information on the preferred object for a given content_id.

rets_client.get_object(resource, object_type, content_ids, object_ids='*', location=0)

Returns a list of dicts containing information on objects for one or more content_ids. The content_ids can be passed as a list if there are multiple content_ids. The object_ids variable limits the objects returned to the index number of each object on the server. This can be useful when getting a single object or subset of total objects. Each dict contains a key of content_md5 that contains the md5 checksum for the object. This should help users identify duplicates supplied by the RETS servers or compare the objects against their previously saved objects.

Here is an example of getting an object's images and saving them to file:

with Session(rets_client = Session(login_url, username, password) as s:
    unique_listing_id = '123456789'
    object_dict_list = s.get_object(
        resource='Property',
        object_type='LargePhoto',
        content_ids=unique_listing_id
    )

    for ob in object_dict_list:
        # Save the images individually
        file_name = "{}_{}.jpg".format(unique_listing_id, ob['content_id'])
        with open(file_name, 'wb') as f:
            f.write(ob['content'])

Searching

Use the client's search method to search for real estate data. All searches must have the resource, class, and search query. The query can be sent as either a Data Mining Query Language string or a search filter dictionary.

The search method takes the following parameters:

• resource: The resource that contains the class to search
• resource_class: The class to search
• search_filter=None: The query as a dict
• dmql_query=None: The query in dmql format
• limit=None: Limit search values count
• offset=None: Offset for RETS request. Useful when RETS limits number of results or transactions
• optional_parameters=None: Values for option paramters
• query_type: The query type to submit as. Defaults to DMQL2
• standard_names: Boolean for if the search uses standard names. Defaults to 0 indicating the search uses system field names
• response_format: The format of the response you would like back, defaults to COMPACT-DECODED

The resource and resource_class parameters are required. You must also provide either the search_filter parameter or the dmql_query parameter.

The dmql query is what RETS is expecting and the search_filter dict ends up creating the dmql to be sent to rets.

>>> search_res = rets_client.search('Property', 'RES', dmql_query='(Status=A)')
>>> the_same_res = rets_client.search('Property', 'RES', search_filter={'Status': 'A"})

Many RETS servers limit the number of results returned with a search request. You may pass the limit and/or offset parameters to the search method to better control the result set.

>>> small_res = rets_client.search('Property', 'RES', search_filter={'Status': 'A"}, limit=1)

The small_res just has a single listing returned.

>>> first_res = rets_client.search('Property', 'RES', search_filter={'Status': 'A"})

The RETS server only returned the first 10,000 results from this query. Do a second query to get the rest of the results.

>>> second_res = rets_client.search('Property', 'RES', search_filter={'Status': 'A"}, offset=10000)

Lastly, if there are any other parameters to send to the Search end point, you may provide them in the optional_parameters dict.

Filters

Complex queries in DQML can be troublesome to read and maintain. Creating these queries as search_filter dictionaries can make this a little better.

The following logical operators are parsed by client.

• $gte: numeric or datetime values greater than or equal to this.
• $lte: numeric or datetime values less than or equal than to this.
• $contains: a string contains these characters anywhere.
• $begins: a string begins with these characters.
• $ends: a string ends with these characters.
• $in: a list of possible values a field can contain.
• $nin: a list of values a field cannot contain.
• $neq: the value must not equal this.

Additionally, all date, datetime, and time objects passed to the search_filter are converted to the appropriate format expected by RETS server.

Examples Search Filters

Active listings in the past 48 hours.

>>> two_days_ago = datetime.today() - datetime.timedelta(days=2)
>>> filter = {
        "Status": "Active",
        "CreatedDatetime": {
            "$gte": two_days_ago
            }
        }
>>> results = rets_client.search('Property', 'RES', search_filter=filter)

Expensive properties that have been on the market over 5 months

>>> five_months_ago = datetime.today() - datetime.timedelta(months=5)
>>> filter = {
        "Status": "Active",
        "CreatedDatetime": {
            "$lte": five_months_ago
            }
        }
    }
>>> results = rets_client.search('Property', 'RES', search_filter=filter)

Listings on a "Main" street in a neighborhood that contains "Quail West". (Some RETS use legal descriptions of neighborhood data or allow brokers to enter inconsistent neighborhood names)

>>> filter = {
        "Status": "Active",
        "StreetName": {
            "$begins": "Main S"
        },
        "DevelopmentName": {
            "$contains": "Quail West"
        }
>>> results = rets_client.search('Property', 'RES', search_filter=filter)

At least four bedrooms, two to three bathrooms, under $150,000.

>>> filter = {
        "Status": "Active",
        "Bedrooms": {
            "$gte": 4
        },
        "Bathrooms": {
            "$in": [2, 3]
        },
        "ListPrice": {
            "$lte": 150000
        }
    }
>>> results = rets_client.search('Property', 'RES', search_filter=filter)

Search Results

Searches with the RETS client return a list of dictionaries that represents listings of a search result.

Custom Results Parser

Some RETS server return non-standard search result responses. In these cases it is useful to create your own parser class. This class must define a method generator that takes a single argument of the rets server response. A simple example of this can be found in the CREA Test file.

When the Session is instantiated, pass the and instance of the class as the search_parser class.

RETS Exceptions

There are many RETS Reply Codes that can be returned from the server. As a rule, this rets library raises a rets.exceptions.RETSException for all reply codes that are non-zero. The reply_code and reply_text are set as parameters for the exception to make it easier for applications to catch and respond to specific reply codes.

Contributing

This RETS client has a long way to go, and keeping up with new RESO Standards , RETS 2.0, and other features will require ongoing maintenance. Please feel free to fork this repo and make pull requests to the development branch if you wish to contribute. Ensure that all new code has accompanying tests. Travis-CI will run your code through the current and new tests when you make a pull request.

All pull requests should reference an Github issue. Features and bugs should be discussed in the issue rather than be discussed in a pull request.

Many thanks to the passive contribution of @troydavisson for his work on PHRETS. We shamelessly used many of his great conventions to make this project successful.

Testing

If you wish to test the code prior to contribution use tox to test on python 2 and 3.

tox

Helpful RETS Links

• http://www.reso.org/glossary/
• https://www.flexmls.com/developers/rets/tutorials/example-rets-session/
• http://www.realtor.org/retsorg.nsf/pages/docs