Smart Plugins¶

Pyrofork embeds a smart, lightweight yet powerful plugin system that is meant to further simplify the organization of large projects and to provide a way for creating pluggable (modular) components that can be easily shared across different Pyrofork applications with minimal boilerplate code.

Introduction¶

Prior to the Smart Plugin system, pluggable handlers were already possible. For example, if you wanted to modularize your applications, you had to put your function definitions in separate files and register them inside your main script after importing your modules, like this:

myproject/
    handlers.py
    main.py

• handlers.py

  async def echo(client, message):
      await message.reply(message.text)
  
  
  async def echo_reversed(client, message):
      await message.reply(message.text[::-1])
  

• main.py

  from pyrogram import Client, filters
  from pyrogram.handlers import MessageHandler
  
  from handlers import echo, echo_reversed
  
  app = Client("my_account")
  
  app.add_handler(
      MessageHandler(
          echo,
          filters.text & filters.private))
  
  app.add_handler(
      MessageHandler(
          echo_reversed,
          filters.text & filters.private),
      group=1)
  
  app.run()
  

This is already nice and doesn’t add too much boilerplate code, but things can get boring still; you have to manually import, manually add_handler() and manually instantiate each MessageHandler object because you can’t use decorators for your functions. So, what if you could? Smart Plugins solve this issue by taking care of handlers registration automatically.

Using Smart Plugins¶

Setting up your Pyrofork project to accommodate Smart Plugins is pretty straightforward:

1. Create a new folder to store all the plugins (e.g.: “plugins”, “handlers”, …).

2. Put your python files full of plugins inside. Organize them as you wish.

3. Enable plugins in your Client.

myproject/
    plugins/
        handlers.py
    main.py

• plugins/handlers.py

  from pyrogram import Client, filters
  
  
  @Client.on_message(filters.text & filters.private)
  async def echo(client, message):
      await message.reply(message.text)
  
  
  @Client.on_message(filters.text & filters.private, group=1)
  async def echo_reversed(client, message):
      await message.reply(message.text[::-1])
  

• main.py

  from pyrogram import Client
  
  plugins = dict(root="plugins")
  
  Client("my_account", plugins=plugins).run()
  

The first important thing to note is the new plugins folder. You can put any python file in any subfolder and each file can contain any decorated function (handlers) with one limitation: within a single module (file) you must use different names for each decorated function.

The second thing is telling Pyrofork where to look for your plugins: you can use the Client parameter “plugins”; the root value must match the name of your plugins root folder. Your Pyrofork Client instance will automatically scan the folder upon starting to search for valid handlers and register them for you.

Then you’ll notice you can now use decorators. That’s right, you can apply the usual decorators to your callback functions in a static way, i.e. without having the Client instance around: simply use @Client (Client class) instead of the usual @app (Client instance) and things will work just the same.

Specifying the Plugins to include¶

By default, if you don’t explicitly supply a list of plugins, every valid one found inside your plugins root folder will be included by following the alphabetical order of the directory structure (files and subfolders); the single handlers found inside each module will be, instead, loaded in the order they are defined, from top to bottom.

This default loading behaviour is usually enough, but sometimes you want to have more control on what to include (or exclude) and in which exact order to load plugins. The way to do this is to make use of include and exclude directives in the dictionary passed as Client argument. Here’s how they work:

• If both include and exclude are omitted, all plugins are loaded as described above.

• If include is given, only the specified plugins will be loaded, in the order they are passed.

• If exclude is given, the plugins specified here will be unloaded.

The include and exclude value is a list of strings. Each string containing the path of the module relative to the plugins root folder, in Python notation (dots instead of slashes).

E.g.: subfolder.module refers to plugins/subfolder/module.py, with root="plugins".

You can also choose the order in which the single handlers inside a module are loaded, thus overriding the default top-to-bottom loading policy. You can do this by appending the name of the functions to the module path, each one separated by a blank space.

E.g.: subfolder.module fn2 fn1 fn3 will load fn2, fn1 and fn3 from subfolder.module, in this order.

Examples¶

Given this plugins folder structure with three modules, each containing their own handlers (fn1, fn2, etc…), which are also organized in subfolders:

myproject/
    plugins/
        subfolder1/
            plugins1.py
                - fn1
                - fn2
                - fn3
        subfolder2/
            plugins2.py
                ...
        plugins0.py
            ...
    ...

• Load every handler from every module, namely plugins0.py, plugins1.py and plugins2.py in alphabetical order (files) and definition order (handlers inside files):

  plugins = dict(root="plugins")
  
  Client("my_account", plugins=plugins).run()
  

• Load only handlers defined inside plugins2.py and plugins0.py, in this order:

  plugins = dict(
      root="plugins",
      include=[
          "subfolder2.plugins2",
          "plugins0"
      ]
  )
  
  Client("my_account", plugins=plugins).run()
  

• Load everything except the handlers inside plugins2.py:

  plugins = dict(
      root="plugins",
      exclude=["subfolder2.plugins2"]
  )
  
  Client("my_account", plugins=plugins).run()
  

• Load only fn3, fn1 and fn2 (in this order) from plugins1.py:

  plugins = dict(
      root="plugins",
      include=["subfolder1.plugins1 fn3 fn1 fn2"]
  )
  
  Client("my_account", plugins=plugins).run()
  

Load/Unload Plugins at Runtime¶

In the previous section we’ve explained how to specify which plugins to load and which to ignore before your Client starts. Here we’ll show, instead, how to unload and load again a previously registered plugin at runtime.

Each function decorated with the usual on_message decorator (or any other decorator that deals with Telegram updates) will be modified in such a way that a special handlers attribute pointing to a list of tuples of (handler: Handler, group: int) is attached to the function object itself.

• plugins/handlers.py

  @Client.on_message(filters.text & filters.private)
  async def echo(client, message):
      await message.reply(message.text)
  
  print(echo)
  print(echo.handlers)
  

• Printing echo will show something like <function echo at 0x10e3b6598>.

• Printing echo.handlers will reveal the handlers, that is, a list of tuples containing the actual handlers and the groups they were registered on [(<MessageHandler object at 0x10e3abc50>, 0)].

Unloading¶

In order to unload a plugin, all you need to do is obtain a reference to it by importing the relevant module and call remove_handler() Client’s method with your function’s handler instance:

• main.py

  from plugins.handlers import echo
  
  handlers = echo.handlers
  
  for h in handlers:
      app.remove_handler(*h)
  

The star * operator is used to unpack the tuple into positional arguments so that remove_handler will receive exactly what is needed. The same could have been achieved with:

handlers = echo.handlers
handler, group = handlers[0]

app.remove_handler(handler, group)

Loading¶

Similarly to the unloading process, in order to load again a previously unloaded plugin you do the same, but this time using add_handler() instead. Example:

• main.py

  from plugins.handlers import echo
  
  ...
  
  handlers = echo.handlers
  
  for h in handlers:
      app.add_handler(*h)