Data Driven Configuration

This page describes how QMK's data driven JSON configuration system works. It is aimed at developers who want to work on QMK itself.

History

Historically QMK has been configured through a combination of two mechanisms- rules.mk and config.h. While this worked well when QMK was only a handful of keyboards we've grown to encompass nearly 4000 supported keyboards. That extrapolates out to 6000 configuration files under keyboards/ alone! The freeform nature of these files and the unique patterns people have used to avoid duplication have made ongoing maintenance a challenge, and a large number of our keyboards follow patterns that are outdated and sometimes harder to understand.

We have also been working on bringing the power of QMK to people who aren't comformable with a CLI, and other projects such as VIA are working to make using QMK as easy as installing a program. These tools need information about how a keyboard is laid out or what pins and features are available so that users can take full advantage of QMK. We introduced info.json as a first step towards this. The QMK API is an effort to combine these 3 sources of information- config.h, rules.mk, and info.json- into a single source of truth that end-user tools can use.

Now we have support for generating rules.mk and config.h values from info.json, allowing us to have a single source of truth. This will allow us to use automated tooling to maintain keyboards saving a lot of time and maintenance work.

Overview

On the C side of things nothing changes. When you need to create a new rule or define you follow the same process:

Add it to docs/config_options.md
Set a default in the appropriate core file
Add your ifdef statements as needed

You will then need to add support for your new configuration to info.json. The basic process is:

Add it to the schema in data/schemas/keyboards.jsonschema
Add a mapping in data/maps
(optional and discouraged) Add code to extract/generate it to:

lib/python/qmk/info.py
lib/python/qmk/cli/generate/config_h.py
lib/python/qmk/cli/generate/rules_mk.py

Adding an option to info.json

This section describes adding support for a config.h/rules.mk value to info.json.

Add it to the schema

QMK maintains jsonschema files in data/schemas. The values that go into keyboard-specific info.json files are kept in keyboard.jsonschema. Any value you want to make available to end users to edit must go in here.

In some cases you can simply add a new top-level key. Some examples to follow are keyboard_name, maintainer, processor, and url. This is appropriate when your option is self-contained and not directly related to other options.

In other cases you should group like options together in an object. This is particularly true when adding support for a feature. Some examples to follow for this are indicators, matrix_pins, and rgblight. If you are not sure how to integrate your new option(s) open an issue or join #cli on Discord and start a conversation there.

Add a mapping

In most cases you can add a simple mapping. These are maintained as JSON files in data/mappings/info_config.hjson and data/mappings/info_rules.hjson, and control mapping for config.h and rules.mk, respectively. Each mapping is keyed by the config.h or rules.mk variable, and the value is a hash with the following keys:

info_key: (required) The location within info.json for this value. See below.
value_type: (optional) Default raw. The format for this variable's value. See below.
to_json: (optional) Default true. Set to false to exclude this mapping from info.json
to_c: (optional) Default true. Set to false to exclude this mapping from config.h
warn_duplicate: (optional) Default true. Set to false to turn off warning when a value exists in both places

Info Key

We use JSON dot notation to address variables within info.json. For example, to access info_json["rgblight"]["split_count"] I would specify rgblight.split_count. This allows you to address deeply nested keys with a simple string.

Under the hood we use Dotty Dict, you can refer to that documentation for how these strings are converted to object access.

Value Types

By default we treat all values as unquoted "raw" data. If your value is more complex you can use one of these types to intelligently parse the data:

array: A comma separated array of strings
array.int: A comma separated array of integers
int: An integer
hex: A number formatted as hex
list: A space separate array of strings
mapping: A hash of key/value pairs
str: A quoted string literal

Add code to extract it

Most use cases can be solved by the mapping files described above. If yours can't you can instead write code to extract your config values.

Whenever QMK generates a complete info.json it extracts information from config.h and rules.mk. You will need to add code for your new config value to lib/python/qmk/info.py. Typically this means adding a new _extract_<feature>() function and then calling your function in either _extract_config_h() or _extract_rules_mk().

If you are not sure how to edit this file or are not comfortable with Python open an issue or join #cli on Discord and someone can help you with this part.

Add code to generate it

The final piece of the puzzle is providing your new option to the build system. This is done by generating two files:

.build/obj_<keyboard>_<keymap>/src/info_config.h
.build/obj_<keyboard>_<keymap>/src/rules.mk

These two files are generated by the code here:

lib/python/qmk/cli/generate/config_h.py
lib/python/qmk/cli/generate/rules_mk.py

For config.h values you'll need to write a function for your rule(s) and call that function in generate_config_h().

If you have a new top-level info.json key for rules.mk you can simply add your keys to info_to_rules at the top of lib/python/qmk/cli/generate/rules_mk.py. Otherwise you'll need to create a new if block for your feature in generate_rules_mk().

6.4 KiB Raw Permalink Blame History