Skip to content

Easily store, index, and modify Python dicts in Redis (with flexible searching)

License

Notifications You must be signed in to change notification settings

kenjyco/redis-helper

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Note: as of redis-helper v0.4.0, version 3.0 of redis-py is in use, which has backwards incompatible changes withe redis-py 2.x. See https://github.com/redis/redis-py/tree/70ef9ec68f9163c86d4cace2941e2f0ae4ce8525#upgrading-from-redis-py-2x-to-30

About

Install redis-helper, create an instance of redis_helper.Collection (the args/kwargs define the model) and use the add, get, update, delete, and find methods to:

  • quickly store/retrieve/modify Python dicts in Redis
  • filter through indexed fields with simple/flexible find arguments
  • power real-time dashboards with metrics at a variety of time ranges
  • super-charge event logging and system debugging
  • build FAST prototypes and simulators
  • greatly simplify data access patterns throughout application

See the request logging demo and urls demo (with unique_field defined). The examples they reference are short and easy to read.

The redis-helper project evolved from a reference Python project that would be easy to teach and follow many practical best practices and useful patterns. Main purpose was to have something that was super easy to configure (a single ~/.config/redis-helper/settings.ini file for multiple application environments) that did cool things with Redis.

The redis-helper package provides a Collection class that was designed to be easy to interact with in the shell (for exploration, experimentation, and debugging). Most methods on a Collection help minimize typing (passing multiple arguments in a single delimited string when appropriate) and do "the most reasonable thing" whenever possible.

The first time that redis_helper is imported, the sample settings.ini file will be copied to the ~/.config/redis-helper directory.

[default]
image_version = 6-alpine

[dev]
container_name = redis-helper
port = 6379
rm = False
redis_url = redis://localhost:6379/1

[test]
container_name = redis-helper-test
port = 6380
rm = True
redis_url = redis://localhost:6380/9

If docker is installed to your system and your user has permission to use it, the bg-helper docker tools will be used to start a redis container for development or running tests, if Redis is not already installed locally.

(Optionally) install Redis and start server locally

% sudo apt-get install -y redis-server

or

% brew install redis
% brew services start redis

Install redis-helper

Intro

Redis is a fast in-memory data structure server, where each stored object is referenced by a key name. Objects in Redis correspond to one of several basic types, each having their own set of specialized commands to perform operations. The redis Python package provides the StrictRedis class, which contains methods that correspond to all of the Redis server commands.

When initializing Collection objects, you must specify the "namespace" and "name" of the collection (which are used to create the internally used _base_key property). All Redis keys associated with a Collection will have a name pattern that starts with the _base_key.

import redis_helper as rh


request_logs = rh.Collection(
    'log',
    'request',
    index_fields='status, uri, host',
    json_fields='request, response, headers'
)

urls = rh.Collection(
    'web',
    'url',
    unique_field='name',
    index_fields='domain, _type'
)

notes = rh.Collection(
    'input',
    'note',
    index_fields='topic, tag',
    insert_ts=True
)

sample = rh.Collection(
    'ns',
    'sample',
    unique_field='name',
    index_fields='status',
    json_fields='data',
    rx_name='\S{4,6}',
    rx_status='(active|inactive|cancelled)',
    rx_aws='[a-z]+\-[0-9a-f]+',
    insert_ts=True
)

uses_sample = rh.Collection(
    'ns',
    'uses_sample',
    index_fields='z',
    rx_thing='\S{4,6}',
    reference_fields='thing--ns:sample'
)
  • a unique_field can be specified on a collection if items in the collection should not contain duplicate values for that particular field
    • the unique_field cannot also be included in json_fields or pickle_fields
    • if you specify a unique_field, that field must exist on each item you add to the collection
  • use index_fields to specify which fields you will want to filter on when using the find method
    • the values for data fields being indexed MUST be simple strings or numbers
    • the values for data fields being indexed SHOULD NOT be long strings, as the values themselves are part of the index keys
  • use json_fields to specify which fields should be JSON encoded before insertion to Redis (using the very fast ujson library)
  • use rx_{field} to specify a regular expression for any field with strict rules for validation
  • use reference_fields to specify fields that reference the unique_field of another collection
    • uses field--basekey combos
  • use pickle_fields to specify which fields should be pickled before insertion to Redis
  • set insert_ts=True to create an additional index to store insert times
    • only do this if you are storing items that you are likely to update and also likely to want to know the original insert time
      • each time an object is updated, the score associated with the hash_id (at the _ts_zset_key) is updated to the current timestamp
      • the score associated with the hash_id (at the _in_zset_key) is never updated

Essentially, you can store a Python dict in a Redis hash and index some of the fields in Redis sets. The collection's _ts_zset_key is the Redis key name for the sorted set containing the hash_id of every hash in the collection (with the score being a utc_float corresponding to the UTC time the hash_id was added or modified).

  • if insert_ts=True was passed in when initializing the Collection (or sub-class), then the collection will also define self.in_zset_key to be the Redis key name for the sorted set (for hash_id and utc_float of insert time)
request_logs.add(
    method='get',
    status=400,
    host='blah.net',
    uri='/info',
    request={'x': 50, 'y': 100},
    response={'error': 'bad request'},
)

urls.add(
    name='redis-helper github',
    url='https://github.com/kenjyco/redis-helper',
    domain='github.com',
    _type='repo',
)

The get method is a wrapper to hash commands hget, hmget, or hgetall. The actual hash command that gets called is determined by the number of fields requested.

  • a Python dict is typically returned from get
  • if item_format is specified, a string will be returned matching that format instead
request_logs.get('log:request:1')
request_logs.get('log:request:1', 'host,status')
request_logs.get('log:request:1', item_format='{status} for {host}{uri}')
request_logs.get_by_position(0, item_format='{status} for {host}{uri}')
urls.get_by_position(-1, 'domain,url')
urls.get_by_unique_value('redis-helper github', item_format='{url} points to a {_type}')
  • the get_by_position and get_by_unique_value methods are wrappers to get
    • the get_by_unique_value method is only useful if a unique_field was set on the Collection

The find method allows you to return data for items in the collection that match some set of search criteria. Multiple search terms (i.e. index_field:value pairs) maybe be passed in the terms parameter, as long as they are separated by one of , ; |. Any fields specified in the get_fields parameter are passed along to the get method (when the actual fetching takes place).

  • when using terms, all terms that include the same field will be treatead like an "or" (union of related sets), then the intersection of different sets will be computed
  • see the Redis set commands and sorted set commands

There are many options for specifying time ranges in the find method including:

  • since and until when specifying num:unit strings (i.e. 15:seconds, 1.5:weeks, etc)
  • start_ts and end_ts when specifying timestamps with a form between YYYY and YYYY-MM-DD HH:MM:SS.f
  • start and end when specifying a utc_float
  • for since, until, start_ts, and end_ts, multiple values may be passed in the string, as long as they are separated by one of , ; |.
    • when multiple time ranges are specified, the find method will determine all reasonable combinations and return a result-set per combination (instead of returning a list of items, returns a dict of list of items)

If count=True is specified, the number of results matching the search criteria are returned instead of the actual results

  • if there are multiple time ranges specified, counts will be returned for each combination
request_logs.find('status:400, host:blah.net', get_fields='uri,error')
request_logs.find(since='1:hr, 30:min', until='15:min, 5:min')
request_logs.find(count=True, since='1:hr, 30:min', until='15:min, 5:min')
urls.find(count=True, since='1:hr, 30:min, 10:min, 5:min, 1:min')
urls.find(start_ts='2017-02-03', end_ts='2017-02-03 7:15:00')
urls.find(start_ts='2017-02-03', item_format='{_ts} -> {_id}')

The update method allows you to change values for some fields (modifying the unique_field, when it is specified, is not allowed).

  • every time a field is modified for a particular hash_id, the previous value and score (timestamp) are stored in a Redis hash
  • the old_data_for_hash_id or old_data_for_unique_value methods can be used to retrieve the history of all changes for a hash_id
urls.update('web:url:1', _type='fancy', notes='this is a fancy url')
urls.old_data_for_hash_id('web:url:1')
urls.old_data_for_unique_value('redis-helper github')

The load_ref_data option on get, get_by_unique_value, or find methods allow you to load the referenced data object from the other collection (where reference_fields are specified)

In [1]: sample.add(name='hello', aws='ami-0ad5743816d822b81', status='active')
Out[1]: 'ns:sample:1'

In [2]: uses_sample.add(thing='hello', z=500, y=True)
Out[2]: 'ns:uses_sample:1'

In [3]: uses_sample.get('ns:uses_sample:1')
Out[3]: {'thing': 'hello', 'z': 500, 'y': True}

In [4]: uses_sample.get('ns:uses_sample:1', load_ref_data=True)
Out[4]:
{'thing': {'name': 'hello',
  'aws': 'ami-0ad5743816d822b81',
  'status': 'active',
  '_id': 'ns:sample:1',
  '_ts': 20201028210044.875},
 'z': 500,
 'y': True}

In [5]: uses_sample.add(thing='byebye', z=100, y=True)
Out[5]: 'ns:uses_sample:2'

In [6]: uses_sample.get('ns:uses_sample:2', load_ref_data=True)
Out[6]: {'thing': 'byebye', 'z': 100, 'y': True}

Tip

There may be times where you want to use redis-helper (if it's already installed), but don't want to make it an explicit requirement of your project. In cases like this you can do the following:

try:
    import redis_helper as rh
    from redis import ConnectionError as RedisConnectionError
except (ImportError, ModuleNotFoundError):
    SomeCollection = None
else:
    try:
        SomeCollection = rh.Collection(
            ...
        )
    except RedisConnectionError:
        SomeCollection = None

Then in whatever function, you can just do:

def some_func():
    if SomeCollection is None:
        return

    # Do stuff with SomeCollection

Local development setup

% git clone https://github.com/kenjyco/redis-helper
% cd redis-helper
% ./dev-setup.bash

The dev-setup.bash script will create a virtual environment in the ./venv directory with extra dependencies (ipython, pdbpp, pytest), then copy settings.ini to the ~/.config/redis-helper directory.

Running tests in development setup

The setup.cfg file contains the options for py.test, currently -vsx -rs --pdb.

The -vsx -rs --pdb options will run tests in a verbose manner and output the reason why tests were skipped (if any were skipped). If there are any failing tests, py.test will stop on the first failure and drop you into a pdb++ debugger session.

See the debugging section of the README for tips on using the debugger and setting breakpoints (in the actual project code, or in the test code).

% venv/bin/py.test

or

% venv/bin/python3 setup.py test

Note: This option requires setuptools to be installed.

Usage

The rh-download-examples, rh-download-scripts, rh-notes, and rh-shell scripts are provided.

$ venv/bin/rh-download-examples --help
Usage: rh-download-examples [OPTIONS] [DIRECTORY]

  Download redis-helper example files from github

Options:
  --help  Show this message and exit.

$ venv/bin/rh-download-scripts --help
Usage: rh-download-scripts [OPTIONS] [DIRECTORY]

  Download redis-helper script files from github

Options:
  --help  Show this message and exit.

$ venv/bin/rh-notes --help
Usage: rh-notes [OPTIONS] [TOPIC]

  Prompt user to enter notes (about a topic) until finished; or review notes

Options:
  -c, --ch TEXT  string appended to the topic (default "> ")
  -s, --shell    Start an ipython shell to inspect the notes collection
  --help         Show this message and exit.

$ venv/bin/rh-shell --help
Usage: rh-shell [OPTIONS]

  Interactively select a Collection model and start ipython shell

Options:
  --help  Show this message and exit.
>>> import redis_helper as rh
>>> collection = rh.Collection(..., index_fields='field1, field3')
>>> hash_id = collection.add(field1='', field2='', field3='', ...)
>>> collection.add(...)
>>> collection.add(...)
>>> collection.update(hash_id, field1='', field4='', ...)
>>> change_history = collection.old_data_for_hash_id(hash_id)
>>> data = collection.get(hash_id)
>>> some_data = collection.get(hash_id, 'field1, field3')
>>> results = collection.find(...)
>>> results2 = collection.find('field1:val, field3:val', ...)
>>> results3 = collection.find(..., get_fields='field2, field4')
>>> counts = collection.find(count=True, ...)
>>> top_indexed = collection.index_field_info()
>>> collection.delete(hash_id, ...)

Basics - Part 1 (request logging demo)

Demo bookmarks:

  • 1:10 is when the ipython session is started with venv/bin/ipython -i request_logs.py
  • 3:14 is when a second ipython session is started (in a separate tmux pane) to simulate a steady stream of requests with slow_trickle_requests(randomsleep=True, show=True)
  • 4:22 is when the index_field_info method is used to get the latest counts of top indexed items
  • 6:11 is when slow_trickle_requests(.001) is run to simulate a large quick burst in traffic
  • 7:00 is when multiple values are passed in the since argument of find... request_logs.find(count=True, since='5:min, 1:min, 30:sec')
  • 8:37 is when get and get_by_position methods are used with a variety of arguments to change the structure of what's returned
  • 10:33 is when the redis_helper.ADMIN_TIMEZONE is changed at run time from America/Chicago to Europe/London
  • 11:27 is when find is used with a variety of arguments to change the structure of what's returned
  • 14:30 is when find is used with multiple search terms and multiple since values... request_logs.find('host:dogs.com, uri:/breeds', count=True, since='5:min, 1:min, 10:sec')
  • 15:54 is when the update method is used to modify data and change history is retrieved with the old_data_for_hash_id method

The first demo walks through the following:

  • creating a virtual environment, installing redis-helper, and downloading example files

    $ python3 -m venv venv
    $ venv/bin/pip3 install redis-helper ipython
    $ venv/bin/rh-download-examples
    $ cat ~/.config/redis-helper/settings.ini
    $ venv/bin/ipython -i request_logs.py
    
  • using the sample Collection defined in request_logs.py to

    • show values of some properties on a Collection
      • redis_helper.Collection._base_key
      • redis_helper.Collection.now_pretty
      • redis_helper.Collection.now_utc_float
      • redis_helper.Collection.keyspace
      • redis_helper.Collection.size
      • redis_helper.Collection.first
      • redis_helper.Collection.last
    • show values of some settings from redis_helper
      • redis_helper.APP_ENV
      • redis_helper.REDIS_URL
      • redis_helper.REDIS
      • redis_helper.SETTINGS_FILE
      • redis_helper.ADMIN_TIMEZONE
    • show output from some methods on a Collection
      • redis_helper.Collection.index_field_info()
      • redis_helper.Collection.find()
      • redis_helper.Collection.find(count=True)
      • redis_helper.Collection.find(count=True, since='30:sec')
      • redis_helper.Collection.find(since='30:sec')
      • redis_helper.Collection.find(since='30:sec', admin_fmt=True)
      • redis_helper.Collection.find(count=True, since='5:min, 1:min, 30:sec')
      • redis_helper.Collection.find('index_field:value')
      • redis_helper.Collection.find('index_field:value', all_fields=True, limit=2)
      • redis_helper.Collection.find('index_field:value', all_fields=True, limit=2, admin_fmt=True, item_format='{_ts} -> {_id}')
      • redis_helper.Collection.find('index_field:value', get_fields='field1, field2', include_meta=False)
      • redis_helper.Collection.find('index_field1:value1, index_field2:value2', count=True)
      • redis_helper.Collection.find('index_field1:value1, index_field2:value2', count=True, since='5:min, 1:min, 10:sec')
      • redis_helper.Collection.get(hash_id)
      • redis_helper.Collection.get(hash_id, 'field1,field2,field3')
      • redis_helper.Collection.get(hash_id, include_meta=True)
      • redis_helper.Collection.get(hash_id, include_meta=True, fields='field1, field2')
      • redis_helper.Collection.get(hash_id, include_meta=True, item_format='{_ts} -> {_id}')
      • redis_helper.Collection.get_by_position(0)
      • redis_helper.Collection.get_by_position(0, include_meta=True, admin_fmt=True)
      • redis_helper.Collection.update(hash_id, field1='value1', field2='value2')
      • redis_helper.Collection.old_data_for_hash_id(hash_id)

Basics - Part 2 (urls demo, with unique field)

Demo bookmarks:

  • TODO

The second demo walks through the following:

  • using the sample Collection defined in urls.py to
    • TODO

Settings, environments, testing, and debugging

To trigger a debugger session at a specific place in the project code, insert the following, one line above where you want to inspect

import pdb; pdb.set_trace()

To start the debugger inside test code, use

pytest.set_trace()
  • use (l)ist to list context lines
  • use (n)ext to move on to the next statement
  • use (s)tep to step into a function
  • use (c)ontinue to continue to next break point (i.e. set_trace() lines in your code)
  • use sticky to toggle sticky mode (to constantly show the currently executing code as you move through with the debugger)
  • use pp to pretty print a variable or statement

If the redis server at redis_url (in the test section of ~/.config/redis-server/settings.ini) is not running or is not empty, redis server tests will be skipped.

Use the APP_ENV environment variable to specify which section of the settings.ini file your settings will be loaded from. Any settings in the default section can be overwritten if explicity set in another section.

  • if no APP_ENV is explicitly set, dev is assumed
  • the APP_ENV setting is overwritten to be test no matter what was set when calling py.test tests