gamla 69

No project description provided

גamla is a performant functional programming library for python which supports async.

Installation

pip install gamla

Debugging anonymous compositions

gamla.debug

It is sometimes hard to debug pipelines because you can't place ordinary breakpoints. For this gamla.debug and gamla.debug_exception were created.

gamla.debug can be used within pipelines and provide a pdb breakpoint prompt where the value at this position can be referenced by x.

def increment(x):
    return x + 1

increment_twice = gamla.compose_left(increment, gamla.debug, increment)

increment_twice(1)

The above code will break with x being 2.

When you have a long pipeline and want to debug at each step of the way, you can use gamla.debug_compose and gamla.debug_compose_left.

gamla.debug_exception

In some cases tracking down an exception involves inspecting code that runs many times. Consider the following example:

def increment(x):
    return x + 1

def sometimes_has_a_bug(x):
    if x == 666:
        raise Exception
    return x

increment_with_bug = gamla.map(gamla.compose_left(increment, sometimes_has_a_bug))

tuple(inrement_with_bug(range(1000)))

Adding a gamla.debug here can be quite tedious, because the code will break many times.

Instead we can use gamla.debug_exception to break only in the case the inner function raises, at which case we would get a breakpoint prompt, and be able to inspect the value causing the exception, use the name x. This would like this:

increment_with_bug = gamla.map(gamla.compose_left(increment, gamla.debug_exception(sometimes_has_a_bug)))

One can also use gamla.debug_exception using a decorator.

@gamla.debug_exception
def sometimes_has_a_bug(x):
    if x == 666:
        raise Exception
    return x

Debug mode

gamla.compose(x, y, z) produces a new function which doesn't have a proper name. If x raises an exception, it is sometimes hard to figure out where this occurred. To overcome this, set the env variable GAMLA_DEBUG_MODE (to anything) to get more useful exceptions. This is turned on only by flag because it incurs significant overhead so things might get slow.

Mixing asynchronous and synchronous code

Most functions in this lib will work seamlessly with async and regular functions, and allow the developer to focus on the logic instead of deciding where to place an await.

For example:

import asyncio

import gamla


def increment(i):
    return i + 1


async def increment_async(i):
    asyncio.sleep(1)
    return i + 1


async def run():
    mixed_composition = gamla.compose_left(increment, increment_async, increment)
    return await mixed_composition(0)  # returns 3!

Migrating from toolz

The main problems - toolz is slow and does not support async functions.

Why are curried functions and composition in toolz slow?

These functions use an expensive inspect call to look at a function’s arguments, and doing so at each run.

Why does gamla not suffer from this problem?

Two reasons:

1. It does not have binary signatures on things like map, so it doesn’t need to infer anything (these are higher order functions in gamla).
2. The gamla.curry function eagerly pays for the signature inspection in advance, and remembers its results for future runs.

Function mapping and common gotchas:

Most functions are drop in replacements. Here are some examples:

• curried.(filter|map|valmap|itemmap|keymap) -> gamla.$1 (make sure the call is with a single argument)
• toolz.identity -> gamla.identity
• toolz.contains -> gamla.contains
• toolz.lt -> gamla.greater_than
• toolz.gt -> gamla.less_than
• toolz.ge -> gamla.less_equals
• toolz.le -> gamla.greater_equals
• toolz.filter(None) -> gamla.filter(gamla.identity)
• toolz.excepts(a, b, c) -> gamla.excepts(a, c, b)
• toolz.excepts(a, b) -> gamla.excepts(a, gamla.just(None), b) (following the “data-last” currying convention)

Releasing a new version

Increment the version in master, and pypi will automatically update.

Updating documentation after change in README.md

While in gamla directory:

1. Install md-to-rst converter: pip install m2r
2. Convert README.md to README.rst: m2r README.md
3. Move README.rst to docs/source folder instead of existing one: mv README.rst docs/source