RedGrease

RedGrease is a Python package and set of tools to facilitate development against Redis in general and Redis Gears in particular.

It may help you create:

• Advanced analytical queries,
• Event based and streaming data processing,
• Custom Redis commands and interactions,
• And much, much more...

... all written in Python and running distributed ON your Redis nodes.

Read the Documentation

Installation

• Client - I.e. Package used by application and/or during dev.
• Runtime - I.e. Package installed on the Redis engine.

Quick Example:

import redgrease as rdgr

conn = rdgr.RedisGears(host='localhost', port=30001)

# count for each genre how many times it appears

results = (
    rdgr.KeysReader()
    .keys(type='hash')
	.map(lambda key: rdgr.cmd.hget(key, 'genres'))
	.filter(lambda x:x != '\\N')
    .flatmap(lambda x: x.split(','))
	.map(str.strip)
	.countby()
	.run(on=conn)
)


for res in results:
    print(f"{res['key']::<15s}{res['value']}")

RedGrease consists of the followinig, components:

1. A Redis / Redis Gears client, which is an extended version of the redis client, but with additional pythonic functions, mapping closely (1-to-1) to the Redis Gears command set (e.g. RG.PYEXECUTE, RG.GETRESULT, RG.TRIGGER, RG.DUMPREGISTRATIONS etc), outlined here

import redgrease

gear_script = ... # Some vanilla Gear script string, a GearFunction object or a script file path.

rg = redgrease.RedisGears()
rg.gears.pyexecute(gear_script)  # <--

2. Wrappers for the runtime functions (e.g. GearsBuilder, GB, atomic, execute, log etc) that are automatically loaded into the server runtime environment. These placeholder versions provide docstrings, auto completion and type hints during development, and does not clash with the actual runtime, i.e does not require redgrease to be installed on the server.

3. Server side Redis commands, allowing for all Redis (v.6) commands to be executed on serverside as if using a Redis 'client' class, instead of 'manually' invoking the execute(). It is basically the redis client, but with execute_command() rewired to use the Gears-native execute() instead under the hood.

import redgrease

def download_image(annotation):
    img_id = annotation["image_id"]
    img_key = f"image:{img_id}"
    if redgrease.cmd.hexists(img_key, "image_data"):  # <- hexists
        # image already downloaded
        return img_key
    redgrease.log(f"Downloadign image for annotation: {annotation}")
    image_url = redgrease.cmd.hget(img_key, "url")  # <- hget
    response = requests.get(image_url)
    redgrease.cmd.hset(img_key, "image_data", bytes(response.content))  # <- hset
    return img_key


# Redis connection (with Gears)
connection = redgrease.RedisGears()

# Automatically download corresponding image, whenever an annotation is created.
image_keys = (
    redgrease.KeysReader()
    .values(type="hash", event="hset")
    .foreach(download_image, requirements=["requests"])
    .register("annotation:*", on=connection)
)

4. 'First class' GearFunction objects, inspired by the remote builders of the official redisgears-py client, but with some differences.

records = redgrease.KeysReader().records(type="hash")  # <- fun 1 : hash-recods

record_listener = ( # <- fun 2, "extents" fun 1
    records.foreach(schedule)
    .register(key_pattern, eventTypes=["hset"])
)  

count_by_status = (   # <- fun 3, also "extends" fun 1
    records.countby(lambda r: r.value.get("status", "unknown"))
    .map(lambda r: {r["key"]: r["value"]})
    .aggregate({}, lambda a, r: dict(a, **r))
    .run(key_pattern)
)

process_records = (   # <- fun 4
    redgrease.StreamReader()
    .values()  # <- Sugar
    .foreach(process, requirements=["numpy"])
    .register("to_be_processed")
)

# Instantiate client objects
server = redgrease.RedisGears()

# Different ways of executing
server.gears.pyexecute(record_listener)
process_records.on(server)
count = count_by_status.on(server)

⚠️ As with the official package, this require that the server runtime Python version matches the client runtime that defined the function. As of writing this is Only for Python 3.7. But don't worry, both "vanilla" and RedGrease flavoured Gear functions can still be excuted by file.

5. CLI tool running and or loading of Gears scripts onto a Redis Gears cluster. Particularls useful for "trigger-based" CommandReader Gears. It also provides a simple form of 'hot-reloading' of Redis Gears scripts, by continously monitoring directories containing Redis Gears scripts and automatically 'pyexecute' them on a Redis Gear instance if it detects modifications. The purpose is mainly to streamline development of 'trigger-style' Gear scripts by providing a form of hot-reloading functionality.

redgrease --server 10.0.2.21 --watch scripts/

This will 'pyexecute' the gears scripts in the 'scripts' directory against the server. It will also watch the directors for changes and re-execute the scripts if they have been modified.

6. Other boilerplate or otherwise functions and utilities (redgrease.utils), that are commonly used in gears. e.g:
   • A record record function that can be used to transform the default KeysReader dict to an Records object with the appropriate attributes.
   • CommandReader + Trigger function decorator, that makes custom Redis Commands really easy.
   • Helpers to aid debugging and/ or testing of gears.
   • ...

Additional feature suggestions appriciated.

7. Docker images with redgrease pre-installed as well as, hopefully, variants for all python versions 3.6 or later (Work-in-Progress).

Example usage:

import redgrease
import redgrease.utils

relevant_usr_fields = {
    "active": bool,
    "permissions": redgrease.utils.list_parser(str),
}

# # "Open" Gear function
# Extracting a dict for every 'active' user
active_users = (
    redgrease.KeysOnlyReader()
    .map(lambda key: redgrease.cmd.hmget(key, *relevant_usr_fields.keys()))
    .map(
        lambda udata: redgrease.utils.to_dict(
            udata, keys=relevant_usr_fields.keys(), val_transform=relevant_usr_fields
        )
    )
    .filter(lambda usr: usr["active"])
)
# # "Open" Gear function re-use
# Count the number of active users
active_user_count = active_users.count()

# Get all the distinct user permissions
all_issued_permissions = active_users.flatmap(lambda usr: usr["permissions"]).distinct()

# # Redis Client w. Gears-features can be created separately
r = redgrease.RedisGears()

# # Two ways of running:
# With 'pyexecute' ...
count = r.gears.pyexecute(active_user_count.run("user:*"))
# ... or using the 'on' method
permissions = all_issued_permissions.run("user:*").on(r)

# Result values are directly accessible
print(f"Count: {count}")
if count > 100:
    print("So many users!")
print(permissions)
if "root" in permissions:
    print("Someone has root permissions")

# Errors can be accessed too
if count.errors:
    print(f"Errors counting users: {count.errors}")
if permissions.errors:
    print(f"Errors collecting permissions: {permissions.errors}")

Installation

Redgrease may be installed either as a developmet tool only, a client library and/or as a runtime library on the Redis Gears server. It can be installed with different 'extras' dependencies depending on preferred usage.

Client Installation

Installaton for development and/or client application environment.

In the environment where you develop your Gears scripts, simply install 'redgrease' with pip3, as usual:

python3 -m pip install redgrease[all]

This installs all the dependencies, allowing for the full features set.

reagrease[cli] : Installs dependencies for the CLI

Runtime Installation

Installation for the Redis Gears Runtime / Server.

Server Requirements

Firstly, the Redis Engine must be running the Redis Gears module. The easiest way to get started is to run published Docker images:

1. Single engine setup

docker run -p 6379:6379 redislabs/redisgears

2. Three master node cluster setup

docker run -p 30001:30001 -p 30002:30002 -p 30003:30003 redislabs/rgcluster

For more advanced deployments of Redis with Gears, please refer to the official Gears installation documentation.

Redgrease Runtime Package

As for RedGrease, it is recommended to use the redgrease[runtime] package as a server side dependency. This installs dependencies for the all the server side features such as server side Redis commands and the runtime for gears constructed with the Remote Gears Builder.

The easiest way of installing the latest stable runtime package is by simply setting the enforce_redgrease argument of pyexecute to True. Redgrease will also be automatically enforced when executing any dynamically constructed GearFunction objects.

import redgrease

rg = redgrease.RedisGears()
rg.gears.pyexecute(enforce_redgrease=True)

NOTE: Its possible to pin the versions inside enforce_redgrese, see redgrease.requirements.PackageOption for more details.

Extras

The current 3rd party packages in the runtime extras are:

• attrs - For parsing composite Gears response structures into attrs objects
• redis - Needed for server-side Redis commands
• cludpickle - Needed for the "Remote" gears (similar to the official redisgears-py client)

Note on the runtime environment

If the only thing you need is the redgrease client, documentation, type hints and the loader CLI for development of conventional Gears scripts (i.e that only use the standard commands and loaded as strings), then the RedGrease package is not strictly required to be installed in the Redis Gears Python Runtime environment (i.e. on the server).

You can in this case, simply remove the redgrease import clause from your script, after development, but before pyexecuting them as per the Redis Gears documentation. Such scripts will still run perfectly fine without redgrease in Redis Gears Environment.

A minimal install, without any 3rd party dependencies, which is pretty much only the syntactic sugar and runtime placeholders, can be installed using the bare redgrease package.

This might be useful if you really don't want the 3rd party packages in the server runtime but still want to use the redgrease sugar.

You can also use the RedGrease watcher or loader CLI to automate loading your scripts as well as requirements from a normal requirements.txt files, as outlined here

Usage / Documentation

The documention is work-in-progress, but the latest and greatest version is available here:

https://redgrease.readthedocs.io

Go read the docs!

Testing

Tests are separate from the package, but are available in the GitHub repo.

git clone https://github.com/lyngon/redgrease

In order to run the tests, Docker is required to be installed in order to spin up fresh Redis instances, on demand for the tests.

Pytest and quite a number of its add-ons, as well as tox, is also needed to run the tests properly. All tests, including development, requirements are best installed through:

cd redgrease/
python3 -m venv .venv
source .venv/bin/activate
pip install -r src/requirements-dev.txt

Then the test can be run with:

pytest 

NOTE: Running the tests takes very long, because a new fresh Redis instance is spun up for each test, just to ensure no risk of cross-contamination. This may be optimized, as many tests are actually independent, but it's left like this for now.

Why This?

The need for this arose from wanting to prototype the concepts for a new Redis module, using Gears CommandReaders and triggers instead of having to write a full fledged module in C.

Initially RedGrease has been just a very simple module, with placeholders for the default Redis gears runtime functions, with type hints and docstrings, just to make it more convenient and less error prone to write Gears functions.

Then the loader CLI was created, in order to further speed up the rapid development cycle.

Then the server-side Redis client commands function was added to minimize errors (e.g. misspelled command strings).

Then the client was added ... and before long it started to get a life of its own.

Note that this means RedGrease package is primarily intended to be an aid for development of Gears scripts, and was not originally intended to be used in any "production" software.

This intent has now changed, and the new goal is now to make Redgrease a production grade package for Redis Gears. Granted, there is still quite some way to go to get there, so your support and feedback is greatly appreciated.

If you like this project, or want professional support, please consider sponsoring.