Skip to main content
This is a pre-production deployment of Warehouse. Changes made here affect the production instance of PyPI (
Help us improve Python packaging - Donate today!

an unobtrusive and light reactive system

Project Description
author:Alberto Berti
license:GNU General Public License version 3 or later

An unobtrusive and light reactive system


This package implements a framework for functional reactive programming which wants to be simple to use and extend without imposing complex concepts of streams, channels and so on that are typical of dataflow programming.

To explain reactive programming just think about a spreadsheet where you have a value cell and a formula cell. The latter updates automatically just when it’s appropriate.

This package implement just that. No, wait, not a spreadsheet, but a way to express that a block of code (a function) that creates a result (a calculated value) or a side effect depends on some other value so that when the value changes, the block of code is automatically re-run.

It has been inspired by Javascript meteor’s “tracker” package but it diverges from in order to be more pythonic.


Let’s see a small example:

cur_temp_fahrenheit = 40

def cur_temp_celsius(t_fahrenheit):
    return (t_fahrenheit - 32) / 1.8

log = []

def log_temp_celsius():


assert log == [4.444444444444445]

This is a small piece of code with a function that converts Fahrenheit degrees to Celsius and then logs them to a list, but you can think of it as any kind of side effect.

Now, we suppose that cur_temp_fahrenheit changes and we want to log it whenever it does so.

To do that we need to trasform cur_temp_fahrenheit into a reactive value and have the tracker track the dependencies between that value and the computation that uses it. This way, when the value is changed, our log_temp_celsius() can be re-run and it will do its work. So we change the code a bit mostly by using a getter and a setter to change the temp variable and add some code when this happens and then instruct the tracker to run the log function so that it knows what to re-run. Let’s see:

from metapensiero import reactive

tracker = reactive.get_tracker()
dep = tracker.dependency()

# this is just to handle setting a global var
cur_temp_fahrenheit = [40]

def get_temp_f():
    return cur_temp_fahrenheit[0]

def set_temp_f(new):
    if new != cur_temp_fahrenheit[0]:
    cur_temp_fahrenheit[0] = new

def cur_temp_celsius(t_fahrenheit):
    return (t_fahrenheit - 32) / 1.8

log = []

def log_temp_celsius(handle):

handle = tracker.reactive(log_temp_celsius)

assert log == [4.444444444444445]


assert log == [4.444444444444445, 10.0]
assert cur_temp_fahrenheit == 50



assert log == [4.444444444444445, 10.0]
assert cur_temp_fahrenheit == 60

As you can see, when we set the current temperature to a new value, log_temp_celsius is re-run and a new entry is added to the log list. we can still use the function(s) without using the tracker, in which case we will have the default, normal, non-reactive behavior. When we use tracker.reactive() all the defined dependencies on reactive-aware data sources are tracked by running the given function immediately. Next, when the reactive source changes, the tracker re-executes the function, thus re-tracking the dependencies that may be different. tracker.reactive() returns an handle, a Computation object that can be used to stop the reactive behavior when it’s no more necessary. The same object is given as parameter to the tracked function.

The example proposed is indeed silly, but shows you the power of the framework:

  • code changes are minimal;
  • the new concepts to learn are very few and simple;
  • the reactive functions can be run alone without tracker involvement and they will run as normal code, without the need to refactor them.

Tracked functions can use tracker.reactive() themselves, in which case the inner trackings will be stopped when the outer is re-run.

The code above is a bit ugly due to the usage of the getter and setter, how can we avoid that? Here is the same example using the Value class:

from metapensiero import reactive

tracker = reactive.get_tracker()
cur_temp_fahrenheit = reactive.Value(40)

def cur_temp_celsius(t_fahrenheit):
    return (t_fahrenheit - 32) / 1.8

log = []

def log_temp_celsius(handle):

handle = tracker.reactive(log_temp_celsius)

assert log == [4.444444444444445]

cur_temp_fahrenheit.value = 50

assert log == [4.444444444444445, 10.0]


cur_temp_fahrenheit.value = 60
assert log == [4.444444444444445, 10.0]

Value class can be used also be used as a method decorator in a way similar to the builtin property decorator but with only a getter function.

Another way to use the Value class is just as a value container, by using its value to get or set the value, or just as any other data member in a class body.

a = Value()

a.value = True
assert a.value == True

class Foo(object):

    bar = Value()

    def zoo(self):
        # ... calc something useful

foo = Foo() = True

assert == True

animal = foo.zoo

When used in class’ body a Value saves a triplet of (Dependency, Computation, value) per instance so you have to take that into account. Value uses weak references in order to avoid keeping instances alive.

There is also a constructor to build reactive namedlist classes.

The framework is also compatible with gevent and asyncio in order to batch computation’s recalculation in another Greenlet or Task, respectively. As all the invalidated calculations are recomputed sequentially, it’s important to avoid having suspension points in the reactive code, like calls to sleep() functions or the execution of yield from and await statements. If this is unavoidable, a manual suspension context manager is avaliable in computations, named suspend(). Using that, the block of code inside a with statement runs isolated, and tracking is reinstated afterwards.

For all those features, please have a look at code and tests for now.


To run the tests you should run the following at the package root:

python test

To test both gevent with Python 2.7 and asyncio with Python 3.5, run:

pip install tox

Build status


0.1 (2016-02-05)

  • Initial effort.
  • Added testing with tox and gevent and asyncio.
  • Firt cut of the docs.

0.2 (2016-02-05)

  • small doc fixes.

0.3 (2016-02-10)

  • more tests.
  • allow __set__ if generator is not defined.
  • refactoring of Value’s code.
  • Fix behavior if Value’s accessed when tracking isn’t active.
  • Provide a mechanism to halt tracking while computing if system suspension is needed (gevent, asyncio).
  • Updates to the doc.
  • Code is now tested in a pre-production environment.

0.4 (2016-02-11)

  • fix a small bug in __delete__().

0.5 (2016-02-11)

  • require flush only when there are dependents.

0.6 (2016-02-13)

  • documentation updates.
  • add a @computation decorator.
Release History

Release History

This version
History Node


History Node


Download Files

Download Files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

File Name & Checksum SHA256 Checksum Help Version File Type Upload Date
metapensiero.reactive-0.6.tar.gz (12.4 kB) Copy SHA256 Checksum SHA256 Source Feb 13, 2016

Supported By

WebFaction WebFaction Technical Writing Elastic Elastic Search Pingdom Pingdom Monitoring Dyn Dyn DNS Sentry Sentry Error Logging CloudAMQP CloudAMQP RabbitMQ Heroku Heroku PaaS Kabu Creative Kabu Creative UX & Design Fastly Fastly CDN DigiCert DigiCert EV Certificate Rackspace Rackspace Cloud Servers DreamHost DreamHost Log Hosting