C++20 library for ordered parallel execution of tasks.

Library implements similar behavior to make or ninja. User defines tasks that shall be processed and library takes care of executing those tasks in lazy manner, with respect to dependencies.

More details on API in Documentation

Recommended way of installing is to use CMakes FetchContent, or use git submodules.

FetchContent_Declare(
  joque
  GIT_REPOSITORY https://github.com/koniarik/joque
  GIT_TAG v0.1
)
FetchContent_MakeAvailable(joque)

Library gives users ability to define a set of tasks. The set is represented by a tree - each set may contain tasks or another subsets.

Each task or set has a name, and these are composed into a full path string. For set M, which contains task A, full name of the task is //M/A.

task has attributes that defines it's dependencies, run after relationships, and resources that it needs. The rules are as follows:

• if task A depends on task B, than A has to be executed after B, and if B is invalidated, so is A
• if task A should run after B, than A has to be executed after B
• if tasks A and B relies on resource X, than only one of those can be executed at once

Each task also defines what should be done for it's execution and how it becames invalidated. This is represented by it's job attribute. Library supports two types of jobs:

• Callables, these are always invalidated, library just executes the callable object
• process, represents a subprocess that shall be executed, it becames invalidated if some of input files are more recent than output files.

The set of tasks is given to exec function which builds internal DAG representing the task set and executes the tasks in parallel, respecting dependencies and resource management.

#include "joque/exec.hpp"
#include "joque/task.hpp"

joque::task_set ts;

std::filesystem::path input_filename{ "in.txt" };
std::filesystem::path output_filename{ "out.txt" };

// create task with process
ts.tasks["A"] = joque::task{
    .job =
        joque::process{
            .cmd =
                { "my_script.sh",
                  "--in",
                  input_filename.string(),
                  "--out",
                  output_filename.string() },
            .input  = { input_filename },
            .output = { output_filename },
        },
    .hidden = true,  // won't be visible in output
};

// use smart deduction for process task
ts.tasks["B"] = joque::task{
    .job = joque::process::derive(
        "my_script.sh",
        "--in",
        input_filename,
        "--out",
        joque::out( output_filename ) ),
    .depends_on = { ts["A"] },
};

// create task with lambda function
ts.tasks["A"] = joque::task{
    .job =
        []( const joque::task& ) {
                std::cout << "A is running" << std::endl;
        },
    .depends_on = { ts["B"] },
};

// execute the tasks with one worker thread
joque::exec( ts, 1 );