Skip to main content

A plugin manager based on setuptools entry points mechanism

Project description

https://github.com/aiidateam/reentry/workflows/ci/badge.svg https://codecov.io/gh/aiidateam/reentry/branch/develop/graph/badge.svg

Reentry

A plugin manager based on setuptools entry points with 10x the speed

Archival notice

Python 3.8 saw the introduction of importlib.metadata into the Python standard library. It is both fast to import importlib.metadata and fast to scan for entry points using the entry_points() function.

The separate importlib-metadata package backports this functionality to python >=3.6, and as of version 4.6.3 (or earlier), scanning is even faster than with the standard library.

Below are timings recorded on a MacBook Air M1 (2020) in a python 3.9.1 conda environment with 248 packages installed (~30 of which register entry points).

pkg_resources: ~147 msec in total:

$ python -mtimeit -n1 -r1 'import pkg_resources'
1 loop, best of 1: 146 msec per loop
$ python -mtimeit -s 'import pkg_resources as pkg' 'pkg.iter_entry_points("aiida.calculations")'
500000 loops, best of 5: 435 nsec per loop

importlib.metadata: ~39 msec in total:

$ python -mtimeit -n1 -r1 'import importlib.metadata'
1 loop, best of 1: 17.8 msec per loop
$ python -mtimeit -s 'import importlib.metadata as im' 'im.entry_points()'
50 loops, best of 5: 21.4 msec per loop

importlib-metadata package (v4.6.3): ~40 msec in total:

$ python -mtimeit -n1 -r1 'import importlib_metadata'
1 loop, best of 1: 33.8 msec per loop
$ python -mtimeit -s 'import importlib_metadata as im' 'im.entry_points()'
50 loops, best of 5: 5.94 msec per loop

reentry (v1.3.1): 25 msec in total:

$ python -mtimeit -n1 -r1 'import reentry'
1 loop, best of 1: 23.8 msec per loop
$ python -mtimeit  -s 'from reentry.default_manager import PluginManager as p' 'p().get_entry_map()'
200 loops, best of 5: 1.07 msec per loop

The advent of faster (solid-state) disks, together with the faster importlib implementations have led to a substantial reduction of the speed benefit that reentry provided (down to ~15 ms or ~40% in the benchmark above). While there may still be certain edge cases where reentry is useful, we will stop using it going forward and are thus archiving this repository.

Users interested in continuing the maintenance of reentry are encouraged to open an issue on the issue tracker.

Features

  • finding plugins: reentry keeps a map of entry points in a file
  • speed: reentry provides an EntryPoint implementation that trades extra checks for search and load speed
  • automatic registering: use reentry_register: True in your setup.py to automatically register plugins

Note that reentry_register creates a build-time dependency on reentry. The suggested way to resolve that is using the method described in PEP518, for which support has been added in pip 10: next to setup.py, put a file pyproject.toml containing:

[build-system]
# Minimum requirements for the build system to execute.
requires = ["setuptools", "wheel", "reentry"]

An alternative way for specifying a build dependency is to put:

setup(
   ...
   setup_requires=[reentry],
   ...
)

in your setup.py. This alternative works with all versions of pip, but fails on systems, where python is linked to old SSL libraries (such as the system python for some versions of OS X).

Limitations

  • entry points with extra dependencies (name = module_name:attrs [extras]) are still supported. Trying to load them, however, will lead to importing pkg_resources and forego the speedup.

Quickstart

Use the following in your plugins’s setup.py:

setup(
   ...
   setup_requires=['reentry'],
   reentry_register=True,
   entry_points={
      'my_plugins': ['this_plugin = this_package.subpackage:member'],
      ...
   }

And iterate over installed plugin from the host package:

from reentry import manager
available_plugins = manager.iter_entry_points(group='my_plugins')
for plugin in available_plugins:
   plugin_object = plugin.load()
   plugin_object.use()

The syntax is consistent with setuptools’s pkg_resources, so you may use it as a fallback:

try:
   from reentry import manager as entry_pt_manager
except:
   import pkg_resources as entry_pt_manager

entry_pt_manager.iter_entry_points(...)
...

Reentry Configuration

Reentry supports getting information from a configuration file. The file will be searched at the following paths:

  • <HOME>/.reentryrc
  • <HOME>/.config/reentry/config

The configuration file has an ini format and supports the following keys:

[general]
datadir=/path/to/data/dir
data_filename=name

The datadir is the folder in which reentry stores the data file that contains the information about the registered entry points. If the config file doesn’t exist in one of the above paths, the datadir is set to <HOME>/.config/reentry/data. data_filename is the name of the data file, in case you want to pick the name by your own instead of letting reentry choose it. Warning: By default, reentry creates a separate data file for every python interpreter in order not to mix entry points between different python environments on your system. Setting a data_filename in the configuration file tells reentry to always use this data file and may result in unexpected behavior if you use reentry in multiple python environments.

You can also set configuration options for reentry via environment variables:

  • datadir can be defined by REENTRY_DATADIR.
  • data_filename can be defined by REENTRY_DATA_FILENAME.

Environment variables take precedence over the configuration file.

What for?

To make entry points usable for plugins in time-critical situations such as command line interfaces!

Setuptool’s entry point system is convenient to use for plugin-based python applications. It allows separate python packages to act as plugins to a host package (or to each other), making it easy for the host to find and iterate over the relevant data structures from plugins.

However, the time spent on importing setuptools scales badly with the number of installed distributions and can easily reach 0.5 seconds for moderately complex environments. Finding and loading of plugins can be time-critical, for example in command line tools that need to load subcommands, where 100 ms are a noticeable delay.

Importing setuptools’s pkg_resources takes time, because it verifies that dependencies are installed correctly for all distributions present in the environment. This allows entry points to have additional dependencies or “extras” (entry_point = module_name:attrs [extras]).

Reentry forgoes this dependency check for entry points without ‘extras’ and thereby manages to be fast and scale better with the number of plugins installed.

Standalone Manager Usage

Sometimes it might be necessary to update the cached entry points, for example

  • after uninstalling a plugin (there are no uninstall hooks by setuptools at the moment)
  • after installing a plugin that does not use install hooks
  • while developing a plugin / plugin host

for those cases reentry has a commandline interface:

$ reentry --help
Usage: reentry [OPTIONS] COMMAND [ARGS]...

  manage your reentry python entry point cache

Options:
  --help  Show this message and exit.

Commands:
  clear  Clear entry point map.
  dev    Development related commands.
  map    Print out a map of cached entry points
  scan   Scan for python entry points to cache for faster loading.
$ reentry scan --help
Usage: reentry scan [OPTIONS] PATTERN

   Scan for python entry points to cache for faster loading.

   Scan only for specific PATTERNs or leave empty to scan all

Options:
   -r, --regex  Treat PATTERNs as regular expresions
   --help       Show this message and exit.
$ reentry map --help
Usage: reentry map [OPTIONS]

Options:
  --dist TEXT   limit map to a distribution
  --group TEXT  limit map to an entry point group
  --name TEXT   limit map to entrypoints that match NAME
  --help        Show this message and exit.

Note: Where needed (e.g. in jupyter notebooks), these operations also be performed in python using the reentry manager, e.g.:

from reentry import manager
manager.scan()

CLI Example

Reentry provides a drop-in replacement for iter_entry_points:

import click
from click_plugins import with_plugins
from reentry.manager import iter_entry_points

@with_plugins(iter_entry_points('cli_plugins'))
@click.group()
def cli():
   """
   command with subcommands loaded from plugin entry points
   """

For this to work, reentry has to be installed and must have been used to scan for entry points in the ‘cli_plugins’ group once.

Development

Running the tests:

tox

Creating a release:

tox -e py39-release

Project details


Download files

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

Source Distribution

reentry-1.3.3.tar.gz (18.7 kB view hashes)

Uploaded source

Built Distribution

reentry-1.3.3-py3-none-any.whl (17.6 kB view hashes)

Uploaded py3

Supported by

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