title: How To --- summary: Covers the most common things plugins would do and how to implement it. --- sort_key: 20 --- body: So you want to build a plugin but it's not quite clear how you would structure a plugin to accomplish this? This part of the documentation should help you find answers to those questions. ## Expose Functionality Most plugins will provide functionality of sorts. There are two places where functionality is typically needed: in templates or for other plugins to use. Plugins can import from each other just like this, but functionality exposed into templates should follow some basic rules: * modify `jinja_env.globals` or `jinja_env.filters` and do not use `process-template-context` unless you absolutely know what you are doing. * expose as few global variables as possible. If you have a lot you want to provide then consider registering an object with the templates and to attach multiple attributes on it. * use clear names for template globals. This is a shared space that all plugins and Lektor itself modify so do not provide a function called `get_script` but for instance call it `get_my_plugin_script`. A simple example of a plugin that implements [Gravatar :ext](https://en.gravatar.com/) support: ```python from hashlib import md5 from werkzeug.urls import url_encode from lektor.pluginsystem import Plugin BASE_URL = 'https://secure.gravatar.com/avatar/' def get_gravatar(email, **options): fn = md5(email.lower().strip().encode('utf-8')).hexdigest() return '%s/%s?%s' % (BASE_URL, fn, url_encode(options)) class GravatarPlugin(Plugin): name = 'Gravatar' def on_setup_env(self, **extra): self.env.jinja_env.filters['gravatar'] = get_gravatar ``` ## Configure Plugins Plugins can come with their own config files and it's encouraged that plugins take advantage of this. Each plugin has exactly one INI file called `.ini` inside the `configs` folder. To get access to the plugin the `get_config` function can be used which returns a dict like object to access the ini file. ```python config = self.get_config() value = config.get('section.key', 'default_value') ``` This would correspond to this config in `configs/my-plugin.ini`: ```ini [section] key = the value ``` ## Dependency Tracking While a lot of dependencies are tracked automatically, when you develop a plugin you probably will discover that sometimes you need to track your own ones. There are different ways in which dependency tracking can work and depending on the situation you might have to use two different mechanisms. 1. **Dependency tracking by association**: while a build of a source object into an artifact is active more dependencies for that artifact can be registered with the [record_dependency :ref](../../api/build/context/record-dependency/) method of the context. It takes a filename that should be recorded as additional dependency for the current artifact 2. **Dependency tracking as artifact source**: if you build your own artifact you need to define the source files that make up the artifact (if you have such a thing). For instance if you build a thumbnail you will need to track those source files that are the source images. This can be done through the [sub_artifact :ref](../../api/build/context/sub-artifact/) method which declares a new artifact. Here is an example of both ways in one plugin: ```python import os from flask import json from lektor.pluginsystem import Plugin def dump_exif(image): ctx = get_ctx() path = posixpath.join(image.path, '-exif.json') @ctx.sub_artifact(path, sources=[image.source_filename]) def include_file(artifact): ctx.record_dependency(__file__) with artifact.open('wb') as f: json.dump(image.exif.to_dict(), f) return path class ExifDumpPlugin(Plugin): def setup_env(self, **extra): self.env.jinja_env.globals['dump_exif'] = dump_exif ``` This dumps out the EXIF data into a JSON file and returns the artifact name. The source image is tracked as direct source for the artifact and within the function we also track the plugin's filename to rebuild if the plugin changes. ## Adding New Field Types Let's say you want to add an "asciidoc" [field type :ref](../../api/db/types/) so you can write with [AsciiDoc](http://www.methods.co.nz/asciidoc) markup. First [install AsciiDoc](http://www.methods.co.nz/asciidoc/INSTALL.html) so its command-line program is available. Then update `blog-post.ini` from the [blog guide :ref](../../guides/blog/) like so: ```ini [fields.body] label = Body type = asciidoc # Custom type. ``` In a blog post's `contents.lr`, write some AsciiDoc like: ``` body: == Header 1 Some text. ----- code here ----- ``` You can add your "asciidoc" type to Lektor with a plugin: ```python from subprocess import PIPE, Popen from lektor.pluginsystem import Plugin from lektor.types import Type def asciidoc_to_html(text): # The "-" at the end tells asciidoc to read from stdin. p = Popen( ['asciidoc', '--no-header-footer', '--backend=html5', '-'], stdin=PIPE, stdout=PIPE, stderr=PIPE) out, err = p.communicate(text) if p.returncode != 0: raise RuntimeError('asciidoc: "%s"' % err) return out # Wrapper with an __html__ method prevents # Lektor from escaping HTML tags. class HTML(object): def __init__(self, html): self.html = html def __html__(self): return self.html class AsciiDocType(Type): widget = 'multiline-text' def value_from_raw(self, raw): return HTML(asciidoc_to_html(raw.value or u'')) class AsciiDocPlugin(Plugin): name = u'AsciiDoc' description = u'Adds AsciiDoc field type to Lektor.' def on_setup_env(self, **extra): # Derives type name "asciidoc" from class name. self.env.add_type(AsciiDocType) ``` ### Base Classes In the general case, new types should inherit from `lektor.types.base.Type`. However, when creating a new type that will use a "single-line" input widget in the Admin GUI, inheriting from `lektor.types.primitives.SingleInputType` will enable the use of the `addon_label` [field option :ref](/docs/models/#fields).