lml seamlessly finds the lml based plugins from your current python environment but loads your plugins on demand. It is designed to support plugins that have external dependencies, especially bulky and/or memory hungry ones. lml provides the plugin management system only and the plugin interface is on your shoulder.

lml enabled applications helps your customers [1] in two ways:

1. Your customers could cherry-pick the plugins from pypi per python environment. They could remove a plugin using pip uninstall command.
2. Only the plugins used at runtime gets loaded into computer memory.

When you would use lml to refactor your existing code, it aims to flatten the complexity and to shrink the size of your bulky python library by distributing the similar functionalities across its plugins. However, you as the developer need to do the code refactoring by yourself and lml would lend you a hand.

The following code tries to get you started quickly with non-lazy loading.

from lml.plugin import PluginInfo, PluginManager


@PluginInfo("cuisine", tags=["Portable Battery"])
class Boost(object):
    def make(self, food=None, **keywords):
        print("I can cook %s for robots" % food)


class CuisineManager(PluginManager):
    def __init__(self):
        PluginManager.__init__(self, "cuisine")

    def get_a_plugin(self, food_name=None, **keywords):
        return PluginManager.get_a_plugin(self, key=food_name, **keywords)


if __name__ == '__main__':
    manager = CuisineManager()
    chef = manager.get_a_plugin("Portable Battery")
    chef.make()

At a glance, above code simply replaces the Factory pattern should you write them without lml. What's not obvious is, that once you got hands-on with it, you can start work on how to do lazy loading.

You can install lml via pip:

$ pip install lml

or clone it and install it:

$ git clone https://github.com/python-lml/lml.git
$ cd lml
$ python setup.py install

Beyond the documentation above, here is a list of projects using lml:

1. pyexcel
2. pyecharts
3. moban

lml is available on these distributions:

1. ARCH linux
2. Conda forge
3. OpenSuse

New BSD