A plugin consists of two files, the source and the descriptor .ini. Let’s say we’d like to provide a plugin for imshow using matplotlib. We’ll call our plugin mpl:
skimage/io/_plugins/mpl.py
skimage/io/_plugins/mpl.ini
The name of the .py and .ini files must correspond. Inside the .ini file, we give the plugin meta-data:
[mpl] <-- name of the plugin, may be anything
description = Matplotlib image I/O plugin
provides = imshow <-- a comma-separated list, one or more of
imshow, imsave, imread, _app_show
The “provides”-line lists all the functions provided by the plugin. Since our plugin provides imshow, we have to define it inside mpl.py:
# This is mpl.py
import matplotlib.pyplot as plt
def imshow(img):
plt.imshow(img)
Note that, by default, imshow is non-blocking, so a special function _app_show must be provided to block the GUI. We can modify our plugin to provide it as follows:
[mpl]
provides = imshow, _app_show
# This is mpl.py
import matplotlib.pyplot as plt
def imshow(img):
plt.imshow(img)
def _app_show():
plt.show()
Any plugin in the _plugins directory is automatically examined by skimage.io upon import. You may list all the plugins on your system:
>>> import skimage.io as io
>>> io.plugins()
{'gtk': ['imshow'],
'matplotlib': ['imshow', 'imsave'],
'pil': ['imread'],
'qt': ['imshow'],
'test': ['imsave', 'imshow', 'imread']}
or only those already loaded:
>>> io.plugins(loaded=True)
{'pil': ['imread']}
A plugin is loaded using the use_plugin command:
>>> import skimage.io as io
>>> io.use_plugin('pil') # Use all capabilities provided by PIL
or
>>> io.use_plugin('pil', 'imread') # Use only the imread capability of PIL
Note that, if more than one plugin provides certain functionality, the last plugin loaded is used.
To query a plugin’s capabilities, use plugin_info:
>>> io.plugin_info('pil')
>>>
{'description': 'Image reading via the Python Imaging Library',
'provides': 'imread'}