mirror of
https://github.com/gsi-upm/senpy
synced 2024-11-23 00:22:28 +00:00
118 lines
4.2 KiB
ReStructuredText
118 lines
4.2 KiB
ReStructuredText
Developing new plugins
|
|
----------------------
|
|
|
|
There are two types of files that were needed by senpy for loading a plugin:
|
|
|
|
- *.senpy: this file is the builder of the plugin.
|
|
- *.py: this file is the interface of the plugin.
|
|
|
|
Plugins Builder
|
|
================
|
|
|
|
The structure of this files is similar to a python dictionary, where the data representation consists on attribute-value pairs.
|
|
The principal attributes are:
|
|
|
|
* name: plugin name used in senpy to call the plugin.
|
|
* module: name of the file where the interface is written (*.py)
|
|
|
|
.. code:: python
|
|
|
|
{
|
|
"name" : "senpyPlugin",
|
|
"module" : "{python file}"
|
|
}
|
|
|
|
You can use another attributes such as `description`, `author`, `version`, etc.
|
|
|
|
|
|
Plugins Interface
|
|
=================
|
|
|
|
The basic methods in a plugin are:
|
|
|
|
* __init__
|
|
* activate: used to load memory-hungry resources
|
|
* deactivate: used to free up resources
|
|
* analyse: called in every user requests. It takes in the parameters supplied by a user and should return a senpy Results.
|
|
|
|
Plugins are loaded asynchronously, so don't worry if the activate method takes too long. The plugin will be marked as activated once it is finished executing the method.
|
|
|
|
F.A.Q.
|
|
======
|
|
If I'm using a classifier, where should I train it?
|
|
???????????????????????????????????????????????????
|
|
|
|
Training a classifier can be time time consuming. To avoid running the training unnecessarily, you can use ShelfMixin to store the classifier. For instance:
|
|
|
|
.. code:: python
|
|
|
|
from senpy.plugins import ShelfMixin, SenpyPlugin
|
|
|
|
class MyPlugin(ShelfMixin, SenpyPlugin):
|
|
def train(self):
|
|
''' Code to train the classifier
|
|
'''
|
|
# Here goes the code
|
|
# ...
|
|
return classifier
|
|
|
|
def activate(self):
|
|
if 'classifier' not in self.sh:
|
|
classifier = self.train()
|
|
self.sh['classifier'] = classifier
|
|
self.classifier = self.sh['classifier']
|
|
|
|
def deactivate(self):
|
|
self.close()
|
|
|
|
You can speficy a 'shelf_file' in your .senpy file. By default the ShelfMixin creates a file based on the plugin name and stores it in that plugin's folder.
|
|
|
|
Where can I define extra parameters to be introduced in the request to my plugin?
|
|
?????????????????????????????????????????????????????????????????????????????????
|
|
|
|
You can add these parameters in the *.senpy file under the attribute "extra_params" : "{param_name}". The name of the parameter is going to act as another python dictionary with the next attributes:
|
|
|
|
* aliases: the different names which can be used in the request to use the parameter.
|
|
* required: this option is a boolean and indicates if the parameters is binding in operation plugin.
|
|
* options: the different values of the paremeter.
|
|
* default: the default value which can have the parameter, this is useful in case the paremeter is required and you want to have a default value.
|
|
|
|
.. code:: python
|
|
|
|
"extra_params": {
|
|
"language": {
|
|
"aliases": ["language", "l"],
|
|
"required": true,
|
|
"options": ["es"],
|
|
"default": "es"
|
|
}
|
|
}
|
|
|
|
This example shows how to introduce a parameter associated language.
|
|
The extraction of this paremeter is used in the analyse method of the Plugin interface.
|
|
|
|
.. code:: python
|
|
|
|
lang = params.get("language")
|
|
|
|
Where can I set up variables for using them in my plugin?
|
|
?????????????????????????????????????????????????????????
|
|
|
|
You can add these variables in the *.senpy with: {variable_name} : {variable_value}.
|
|
|
|
Once you have added your variables, the next step is to extract them in the plugin. The plugin's __init__ method has a parameter called `info` where you can extract the values of the variables. This info parameter has the structure of a python dictionary.
|
|
|
|
Can I activate a DEBUGG mode for my plugin?
|
|
???????????????????????????????????????????
|
|
|
|
You can activate the DEBUGG mode by the command-line tool using the option -d.
|
|
|
|
.. code:: bash
|
|
|
|
python -m senpy -d
|
|
|
|
Where can I find more code examples?
|
|
????????????????????????????????????
|
|
|
|
See: `<http://github.com/gsi-upm/senpy-plugins-community>`_.
|