Skip to main content

Yet another meta build-system.

Project description

yambs (3.0.4)

python Build Status codecov PyPI - Status Dependents (via

Yet another meta build-system.



Python Version Support

This package is tested with the following Python minor versions:

Platform Support

This package is tested on the following platforms:

  • ubuntu-latest
  • macos-latest
  • windows-latest


It's undeniable that a software engineer must implement some project-workflow tooling as a coming-of-age endeavor.

Your current workflow is slow and cumbersome? Time to roll up your sleeves and improve it!

What does every build system on the block advertise? Speed of course!

What else do they advertise? Scalability! Compile source repositories that only fleets of data centers can even fathom checking out entire source trees of. Compile your whole company's software stack in only eight hours! (Down from ten from the previous build-system we used!!!)

Most build systems optimize for technical problem solving, and leave user-facing configuration and recurring-interaction mechanisms an afterthought.

Want to use our awesome tool that solves every hard computer-science problem known to mankind? Sorry but you need to learn a totally bespoke programming language to get started. What's that? Does it have linting + formatting + static-analysis tooling? Your text editor doesn't even ship with syntax highlighting out of the box? C'mon just configure your build. Buy some books on Amazon and spend a few days reading the manual cover-to-cover.

This project aims at bringing sanity to the user-facing parts of build systems: the commands you enter, and the coherence of configuration data that actually warrants human management (and not automated generation).

Prior Art

  • vmklib - a "Makefile library" with a Python extensibility interface, great for generic project workflow tasks, but not specific enough for meta build-system purposes
  • datazen - a solution for connecting the awesomeness of Jinja to data and template directories, eliminating a need for writing bespoke code to generate common boilerplate when a bit of configuration data is sufficient
  • rcmpy - a more targeted datazen, focused on a "put templates and configuration data here" approach to generating userspace configuration files (for e.g. your developer workstation editor, shell and other configurations)
  • userfs - a tool for performing arbitrary fresh-system bootstrapping tasks (cloning repositories, building and installing software from source) at the userspace level, rather than as a privileged user


yambs ain't wheel reinvention.

This isn't a build-system alternative - it generates build instructions for ninja.

What's that? Have I heard of CMake? Meson perhaps? What about Bazel? GNU's Make is a timeless classic.

Build System Cardinal Sins

Infinite Configurability and Flexibility

It's understandable that almost every build system implementation begins with the objective of compiling sources and linking programs from an existing source tree, with probably a lot of inconsistent organization conventions.

This drives complexity into the build system's interface for being told what to do. It's rare for any build system to have an out-of-the-box notion of where to look for things and what work to do.

Infinite Scalability

Isn't any modern piece of software junk if it can't scale to millions of concurrent network connections / disk read-write tasks / computations?

That's what the current software engineering zeitgeist will tell you. Why define conventions for small/medium projects, trivial to spawn the N+1'th instance when the need arises, when you can spend all of your time making an infinitely scalable build system for every line of code that thousands of people are contributing every day?


  1. Sane out-of-the-box conventions: put your code here (and organize it by following a specific convention), put your configuration data here.

Command-line Options

$ ./venv3.12/bin/mbs -h

usage: mbs [-h] [--version] [-v] [-q] [--curses] [--no-uvloop] [-C DIR]
           {compile_config,dist,download,gen,native,uf2conv,noop} ...

Yet another meta build-system.

  -h, --help            show this help message and exit
  --version             show program's version number and exit
  -v, --verbose         set to increase logging verbosity
  -q, --quiet           set to reduce output
  --curses              whether or not to use curses.wrapper when starting
  --no-uvloop           whether or not to disable uvloop as event loop driver
  -C DIR, --dir DIR     execute from a specific directory

                        set of available commands
    compile_config      load configuration data and write results to a file
    dist                create a source distribution
    download            download GitHub release assets
    gen                 poll the source tree and generate any new build files
    native              generate build files for native-only target projects
    uf2conv             convert to UF2 or flash directly
    noop                command stub (does nothing)

Sub-command Options


$ ./venv3.12/bin/mbs compile_config -h

usage: mbs compile_config [-h] [-i INCLUDES_KEY] [-u] [-e]
                          output inputs [inputs ...]

positional arguments:
  output                file to write
  inputs                files to read

  -h, --help            show this help message and exit
  -i INCLUDES_KEY, --includes-key INCLUDES_KEY
                        top-level key to use for included files (default:
  -u, --update          whether or not to use the 'update' merge strategy
                        (instead of 'recursive')
  -e, --expect-overwrite
                        allow configuration files to overwrite data when


$ ./venv3.12/bin/mbs dist -h

usage: mbs dist [-h] [-c CONFIG] [-s]

  -h, --help            show this help message and exit
  -c CONFIG, --config CONFIG
                        the path to the top-level configuration file (default:
  -s, --sources         set this flag to only capture source files


$ ./venv3.12/bin/mbs download -h

usage: mbs download [-h] [-o OWNER] [-r REPO] [-O OUTPUT] [-p PATTERN]

  -h, --help            show this help message and exit
  -o OWNER, --owner OWNER
                        repository owner (default: 'vkottler')
  -r REPO, --repo REPO  repository name (default: 'toolchains')
  -O OUTPUT, --output OUTPUT
                        output directory (default: '.')
  -p PATTERN, --pattern PATTERN
                        a pattern to use to select project specifications
                        filtered by name


$ ./venv3.12/bin/mbs gen -h

usage: mbs gen [-h] [-c CONFIG] [-i] [-w] [-s] [-n]

  -h, --help            show this help message and exit
  -c CONFIG, --config CONFIG
                        the path to the top-level configuration file (default:
  -i, --single-pass     only run a single watch iteration
  -w, --watch           whether or not to continue watching for source tree
  -s, --sources         whether or not to only re-generate source manifests
  -n, --no-build        whether or not to skip running 'ninja'


$ ./venv3.12/bin/mbs native -h

usage: mbs native [-h] [-c CONFIG] [-i] [-w] [-s] [-n]

  -h, --help            show this help message and exit
  -c CONFIG, --config CONFIG
                        the path to the top-level configuration file (default:
  -i, --single-pass     only run a single watch iteration
  -w, --watch           whether or not to continue watching for source tree
  -s, --sources         whether or not to only re-generate source manifests
  -n, --no-build        whether or not to skip running 'ninja'


$ ./venv3.12/bin/mbs uf2conv -h

usage: mbs uf2conv [-h] [-b BASE] [-f FAMILY] [-o FILE] [-d DEVICE_PATH] [-l]
                   [-c] [-D] [-w] [-C] [-i]

positional arguments:
  INPUT                 input file (HEX, BIN or UF2)

  -h, --help            show this help message and exit
  -b BASE, --base BASE  set base address of application for BIN format
                        (default: 0x2000)
  -f FAMILY, --family FAMILY
                        specify familyID - number or name (default: 0x0)
  -o FILE, --output FILE
                        write output to named file; defaults to "flash.uf2" or
                        "flash.bin" where sensible
                        select a device path to flash
  -l, --list            list connected devices
  -c, --convert         do not flash, just convert
  -D, --deploy          just flash, do not convert
  -w, --wait            wait for device to flash
  -C, --carray          convert binary file to a C array, not UF2
  -i, --info            display header information from UF2, do not convert

Internal Dependency Graph

A coarse view of the internal structure and scale of yambs's source. Generated using pydeps (via mk python-deps).

yambs's Dependency Graph

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

yambs-3.0.4.tar.gz (48.3 kB view hashes)

Uploaded Source

Built Distribution

yambs-3.0.4-py3-none-any.whl (67.4 kB view hashes)

Uploaded Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page