Skip to main content
Python Software Foundation 20th Year Anniversary Fundraiser  Donate today!

Extensions to standard Python's heapq for performance applications

Project description

Fast CPython extensions to Python standard library with focus on performance.

This library provides CPython native extensions to mimic some of the well known built-in types. The implementation relies on enforced protocol - all the objects and abstract data types are implemented in C/C++ to provide highly effective manipulation.

Extended heapq - fext.ExtHeapQueue

The extended heap queue acts as a min-heap queue from the standard Python library. It uses a hash table for storing information about indexes (where values sit in the min-heap queue) to optimize removals from the heap to O(log(N)).

https://raw.githubusercontent.com/thoth-station/fext/master/fig/fext_extheapq.png

Using fext in a C++ project

The design of this library allows you to use sources in your C++ project as well. The eheapq.hpp file defines the extended heap queue and edict.hpp the extended dictionary. Python files then act as a bindings to their respective Python interfaces. Mind the API design for the templated classes - it was meant to be used with pointers to objects (so avoid possible copy constructors).

Building the extensions

To build extensions, install the following packages (Fedora):

dnf install -y python3-devel g++ gcc

Now you can build extensions:

python3 setup.py build

If you would like to produce binaries with debug information:

CFLAGS='-Wall -O0 -ggdb' python3 setup.py build

Check sections below for more info on testing the C/C++ parts of extensions.

Reference count and memory leak checks

You can find Makefile in the Git repo. This repo defines targets to perform leak checks and reference count checks. Note they use different Python interpreters (with/without debug information) so make sure you do not mix virtual environments when running the tests.

make check

Developing the extension

First, prepare your environment:

dnf install -y make
make deps

To develop or adjust sources, simply change sources and verify your change is accepted by the test suite:

make check

The check target will run the actual test suite (see also make test). Besides it, the test suite will be executed two more times to check test suite and its behaviour with respect to Python object reference counting (python3-debug dependency will be automatically installed with the provided make deps). This part of the test suite can be executed using make check-refcount. The last part of the test suite runs valgrind against the test suite - you can explicitly trigger this part by calling make check-leaks.

Mind make check-refcount and make check-leaks will take some time given the checks and processing that is done on the background. To verify your changes more iterativelly, make test should do the trick (don’t forget to do make check after that though).

To clean up your environment, perform:

make clean

Building and releasing

The release can be done from a containerized environment:

podman run --rm --workdir /io --entrypoint bash -it --volume `pwd`:/io:Z quay.io/pypa/manylinux2014_x86_64 -c "yum install -y make && make all"

To check what’s happening, let’s run a containerized environment - this can be helpful when you are testing or developing the extension:

podman run --rm --workdir /io --entrypoint bash -it --volume `pwd`:/io:Z quay.io/pypa/manylinux2014_x86_64

The following commands (run in the container stated above) will install all the necessary tools:

yum install -y make
make deps

Once tests pass, clean the environment:

make clean

Now we should be ready to produce bdist_wheel and sdist distribution for PyPI:

python3 setup.py bdist_wheel
python3 setup.py sdist

Finally, upload artifacts to PyPI:

auditwheel repair fext/*.whl
twine upload wheelhouse/*.whl

Alternativelly you can let make all happen.

Installation

The project is hosted on PyPI. You can install it via pip or Pipenv:

pipenv install fext
# pip3 install fext

If there is no release conforming your system, a build process is triggered during the installation - requires python3-devel and gcc/g++.

Usage

These data structures were designed for Thoth’s adviser - for data kept in resolver’s internal state as well as in the reinforcement learning part.

Project details


Download files

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

Files for fext, version 0.2.0
Filename, size File type Python version Upload date Hashes
Filename, size fext-0.2.0-cp36-cp36m-manylinux2014_x86_64.whl (174.3 kB) File type Wheel Python version cp36 Upload date Hashes View
Filename, size fext-0.2.0-cp38-cp38-manylinux2014_x86_64.whl (189.9 kB) File type Wheel Python version cp38 Upload date Hashes View
Filename, size fext-0.2.0.tar.gz (10.6 kB) File type Source Python version None Upload date Hashes View

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring DigiCert DigiCert EV certificate Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page