Developer Instructions

This file is aimed at developers of light, or developers who want to implement "drivers" (enumerators) for their own hardware.

Coding Standards

Light is a small project, which helps keep it clean. However we still would like to see a consistent styling throughout the project, as well as in third-party enumerator implementations. The actual source code may not be fully "up to code" yet, but it's getting there.

We use 4 spaces for indentation. We have an empty line at the top and bottom of each file.

The following two sources should be clear enough examples of our coding style:

Header files

    #pragma once
    #include <stdbool.h>
    #include <stdint.h>
    #include <stdfoo.h> /* foo_type_t */
    typedef struct _some_struct_t some_struct_t;
    struct _some_struct_t 
        uint64_t    id;
        foo_type_t  my_foo_thing;
        foo_type_t  *my_foo_ptr;
    /* Describe what the purpose of this function is, what it does with/to foo_struct, and what it returns. */
    bool do_some_stuff(some_struct_t *foo_struct);

Source files

The second line of each source file should be the include to the corresponding header file, followed by an empty line.

Internal/static functions are always prefixed with an underscore (_).

#include "some_struct.h"

static void _increment_one(uint64_t *out_value)
    *out_value += 1;

bool do_some_stuff(some_struct_t *foo_struct)
    if(foo_struct->id > 33)
        return false;
    if(foo_struct->my_foo_ptr != NULL)
    foo_struct->my_foo_ptr = malloc(sizeof(foo_type_t));
    return true;

Implementing an enumerator

Implementing your own devices through an enumerator is pretty easy. The required steps are as follows:

Step 1

Create a headerfile and a corresponding sourcefile under the impl folder, Call them foo.h and foo.c. Open up the sysfs.c and sysfs.h files for reference implementations.

Step 2

In the header, you need to first do a #pragma once (obviously), then #include "light.h" to get access to some struct declarations, then at the bare minimum declare 6 functions. If you need to store your own data for each device or device-target, you will need to declare structs for these as well.

#pragma once 

#include "light.h"

// Implementation of the foo enumerator
// Enumerates devices for quacking ducks

// Device target data 
struct _impl_foo_data_t
    int32_t internal_quack_id;

typedef struct _impl_foo_data_t impl_foo_data_t;

bool impl_foo_init(light_device_enumerator_t *enumerator);
bool impl_foo_free(light_device_enumerator_t *enumerator);

bool impl_foo_set(light_device_target_t *target, uint64_t in_value);
bool impl_foo_get(light_device_target_t *target, uint64_t *out_value);
bool impl_foo_getmax(light_device_target_t *target, uint64_t *out_value);
bool impl_foo_command(light_device_target_t *target, char const *command_string);

Step 3

In the sourcefile, you need to implement the 6 methods. Make sure to return true on success and false on failure. If you do not actually implement a function (for example impl_foo_command), just return true.

The job of the enumerator is to identify/enumerate a bunch of different devices (or just one, or even zero if it doesnt find any). You are also responsible to create the device targets for them (i.e, the things that you actually write to on the device). You do this by setting the devices and targets up in impl_foo_init. You are not required to do anything in impl_foo_free, any allocated memory will be automatically free'd by light, including device/target data that you allocate yourself. You may use impl_foo_free to free resources you allocate outside of the light API.

#include "impl/foo.h"

#include "light.h"
#include "helpers.h"

bool impl_foo_init(light_device_enumerator_t *enumerator)
    /* Lets create a single device, with a single target, for simplicity */
    /* Create a new device called new_device_name, we dont need any userdata so pass NULL to the device_data parameter */
    light_device_t *new_device = light_create_device(enumerator, "new_device_name", NULL)
    /* Setup userdata specific to the target we will create*/ 
    /* Useful to for example reference an ID in a third-party API or likewise */
    /* NOTE: The userdata will be free()'d automatically on exit, so you do not need to free it yourself */
    impl_foo_data_t *custom_data = malloc(sizeof(impl_foo_data_t));
    custom_data->internal_quack_id = 333;
    /* Create a new device target called new_target_name, and pass in the functions and userdata that we just allocated */
    light_create_device_target(new_device, "new_target_name", impl_foo_set, impl_foo_get, impl_foo_getmax, impl_foo_command, custom_data)
    /* Return true because we didnt get any errors! */
    return true;

bool impl_foo_free(light_device_enumerator_t *enumerator)
    /* We dont need to do anything here, but if we want to, we can free some third-party API resources */
    return true;

/* Implement the other functions to do their thing. Get, Set and GetMax should be self-explanatory. Command is reserved for future use, but basically will allow the user to run custom commands on a target. */

Step 4

Now that you have implemented your enumerator, it is time to inject it to the application itself. You will be able to compile your enumerator into a plugin in the future, but for now, locate the light_initialize function inside light.c. You will see some calls (perhaps just one call) to light_create_enumerator inside of this function. Add one more call to this function to register your enumerator in the application:

The first argument is the application context, the second is the name that your enumerator will get, and the last two are the init and free functions that we implemented.

light_create_enumerator(new_ctx, "foo", &impl_foo_init, &impl_foo_free);

Once you do this, you should be able to find your device target when running light -L, and it should be called something like foo/new_device_name/new_target_name if you followed this guide.

The only thing left now is to create a pull request so that the rest of the world can share the functionality that you just implemented!


If you run into any issues, feel free to create a new Github issue, even if you are just asking for "support" or likewise.