Novelty Plastic Moustache


    0.1.7 • Public • Published


    Build Status


    mekano(1) – maintain, update and regenerate groups of files.


    • is a make-like update tool: you have a bunch of files in some directories, you want to generate other files from them, fast (no unnecessary work);
    • liberally aims to lessen the frustration that might occur working with GNU make(1) on small projects;
    • tries to be balanced between speed and convenience;
    • works best with a powerful shell (like bash & co.), that it does not supplant;
    • is not tied to any specific technology and may be used to compile C/C++, build a web application Javascript/CSS assets, or brew your coffee.


    In ./Mekanofile:

    bin = `node_modules/.bin` ;
    Concat: `cat $in > $out` ;
    Coffee: `$bin/coffee -cp $in > $out` ;
    Minify: `$bin/uglifyjs < $in > $out` ;
        Coffee => build/**/*.js
        Concat -> dist/concat.js ;
        Minify => dist/*.min.js
        :: all `Update all files` ;

    In your preferred shell:

    $ ls
    Mekanofile    src
    $ mekano update
    Updating...  25.0%   Coffee src/ -> build/foo.js
    Updating...  50.0%   Coffee src/ -> build/bar.js
    Updating...  75.0%   Concat build/foo.js build/bar.js -> dist/concat.js
    Updating... 100.0%   Minify dist/concat.js -> dist/concat.min.js
    $ mekano update
    Everything is up to date.
    $ ls
    Mekanofile    src      build       dist        .mekano
    $ ls dist
    concat.js   concat.min.js
    $ mekano clean
    Removing: dist/concat.min.js
    Removing: dist/concat.js
    Removing: build/foo.js
    Removing: build/bar.js


    mekano is a general-purpose update tool. It examines changes made and updates derived files, called the targets, from the files they are derived, called the prerequisites. Typical cases include (non-exhaustive):

    • compiling a group of C/C++ files to their respective object files;
    • linking a group of object files to a single binary;
    • transpiling a plain JavaScript to a minified JavaScript.

    A description file (called mekanofile, or manifest) contains a description of the relationships between files, and the commands that need to be executed to update the targets and reflect changes in their prerequisites.

    mekano focuses above all on correctness and convenience, then speed. It properly takes account of removed and added files; tracks command-line changes; automatically creates output directories; runs commands concurrently; and attempts to provide simple semantics for dependency definitions. This tool is largely inspired by the UNIX make(1) utility.

    mekano only knows how to update files. It is not well suited for so-called 'tasks' (eg. 'test', 'publish'). Plain scripts are probably a better idea for those (bash, Javascript, Python…). Using npm-scripts or tools like grunt is suggested as well.

    The mekanofile is generally meant to be written by hand, but there is – for now – very little support for update-time decision-making (no 'if', no macros). However, you can use a dedicaced macro or procedural language to generate the mekanofile, like m4, Python, Javascript…

    This specific implementation is made with JavaScript on top of Node.js, but keep in mind it may be usable for any purpose, from C/C++ compilation to web assets build. Node.js just makes it easier to be multiplatform.


    Important: you need node v0.10 or higher to run this program.

    $ npm install mekano

    The tool will be available as node_modules/.bin/mekano. It is not recommended to install it globally, because different projects may need different, incompatible versions (of course, semver is used). Is is recommended, however, to add it to your project's package.json (just use npm's --save option).

    To avoid typing the path every time when installed locally, one decent solution is to create an alias, if your shell supports it. For example:

    $ alias mk=node_modules/.bin/mekano

    If you later forget to npm install a project, your shell will just tell you:

    $ mk update
    bash: no such file or directory: node_modules/.bin/mekano


    mekano <command> [options] [bind=value...] [target_name...]


    • update Update the specified targets. The whole project is updated if no target is specified.
    • watch Keep updating files until a signal is caught. It watches files and updates targets when prerequisites change. If the Mekanofile itself changes, you need to relaunch mekano manually.
    • status Display the modified files and dirty targets. No target is updated. If --silent is specified, return a zero exit value if the targets are up to date; otherwise, return 1.
    • clean Remove the specified and dependent targets. For example, with the above Mekanofile, clean dist/concat.js will remove this file and the minified one. All generated files are removed if no target is specified.
    • aliases Display a list of the defined aliases.
    • print <type> Display the mekanofile interpretation. Types:
      • manifest Output the mekanofile as it had been parsed.
      • dot Output the file graph in the graphviz dot format.
    • help Display mekano own help.

    General options:

    • -y, --shy Stop an update as soon as an error occurs. By default, the update continues to get a maximum of errors at once.
    • -n, --dry-run Output commands that would be run. No target is updated nor deleted.
    • -f, --file <mekanofile> Specify a different mekanofile. If - is specified, the standard input is used.
    • -r, --robot Output machine-parseable text.
    • -v, --version Output version and exit.

    Binds and target names can be mixed on the command-line, but targets are always evaluated last. Values cannot refer to values declared inside the mekanofile, but the contrary is possible.

    Without the option -f, mekano looks in sequence for the files ./Mekanofile and ./mekanofile. The first found is read.

    The standard output reports the recipes being executed as well as the completion percentage. The -r option makes the output more easily parseable.

    If any of the SIGHUP, SIGTERM, and SIGQUIT signals is received, the tool stops updating but keeps track of updated files so far. However, it does not kill the running sub-processes. If SIGINT (generally Ctrl+C) is received, it follows a "wait and cooperative exit" (WCE) stategy; it waits for processes to complete, and stops only if they ended on SIGINT themselves. In watch mode, SIGINT only stops the update, if any; a second SIGINT may be needed to quit. For compatibility with sloppy signal handlers, return code 130 is also considered a SIGINT.

    At the moment, Mekano cannot update the Mekanofile itself and take account of it in a single run (with a relation such as M4 -> Mekanofile). This is because it won't reload the Mekanofile after its update. You need to run it twice in this case. This will be improved in the future.


    A mekanofile can contain recipes, relations, and binds. Comments start with #, and end with the next new line. Whitespace is never significant is all other cases. Statements shall be terminated with ;, otherwise mekano will not know the boundary between two successive relations.

    unit = { recipe | relation | bind } ;

    The golden rule when writing a mekanofile is that order does not matter. Whatever the ordering of recipes, relations and binds is, the interpretation will always be the same; even if glob patterns are involved. This makes the syntax purely declarative.


    Recipe grammar (in EBNF) is as below:

    recipe = recipe-name, ":", command, ";" ;
    recipe-name = identifier ;
    command = interpolation ;
    identifier = { ? A-Z, a-z, 0-9, '-' or '_' ? } ;
    interpolation = "`", ? any character ?, "`" ;

    There can only be a single command in a recipe. However, multiple processes can be launched using the shell operators like ;, &&, &, | or ||. The pipelines are especially useful to avoid creating temporary files (if your shell supports it). The command can span several lines. Here a simple recipe example:

    Compile: `gcc -c $in -o $out` ;

    Backticks define an interpolation. An interpolation can contain the backtick character when escaped as $`. $$ yields a single dollar sign. Command lines can refer to bound values with $name or $(name). The following values are automatically available during recipe evaluation:

    • in Space-separated shell-quoted list of the input file(s).
    • out Space-separated shell-quoted list of the output file(s).

    Command lines are evaluated by the local shell, typically with sh -c. 'UpperCamel' case is suggested for naming recipes. Recipes can appear anywhere in the mekanofile, either after or before the relations referring to it.


    Relation grammar is as below:

    relation = ref-list, { transformation }, [ alias ], ";" ;
    transformation = recipe-name, ( "=>" | "->" ), ref-list ;
    alias = "::", alias-name, [ alias-description ] ;
    ref-list = { path | path-glob | alias-name } ;
    alias-name = identifier ;
    alias-description = interpolation ;
    path = { ? alphanumeric character with at least a '.' or a '/' ? } ;
    path-glob = { ? same as path, but with at least a '*', '**' or '{,}' operator ? } ;

    A prerequisite or a target may be either a single file path, a globling pattern, or an alias. A path always contains one of / or ., for example ./foo or foo.js instead of just foo; otherwise it is recognized as an alias. Here a simple relation example:

    source/*.c Compile => obj/*.o Link -> ./hello_world
        :: all `Build the hello world program` ;

    It means: "takes all the C files in the source folder, compile them to object files in obj; then link all those into a single binary ./hello_world. This binary can be referred to as the alias all."


    During evaluation, multi-transformation relations are internally expanded to multiple single-transformation relations. As such, this statement:

    foo.c Compile -> foo.o Link -> a.out :: all ;

    is equivalent to:

    foo.c Compile -> foo.o ;
    foo.o Link -> a.out ;
    a.out :: all ;


    There are two kind of transformations with mekano:

    • plain transformations noted with a simple arrow ->. The recipe is invoked with all the input files, and is assumed to produce all the output files.

    • generative transformations noted with a fat arrow =>. A plain transformation is instantiated for each file matched in the prerequisite globbing pattern. For example, if we have two files foo.c and bar.c, the relation:

      *.c Compile => *.o *.d ;

      is interpreted as:

      foo.c Compile -> foo.o foo.d ;
      bar.c Compile -> bar.o bar.d ;


    Globbing patterns can appear both as prerequisites and targets, but yield different results. Prerequisite patterns expand from two sources:

    • existing source files matching the pattern;
    • other relations' targets matching the pattern.

    Minimatch is used to match the files. A prerequisite pattern never matches the targets of the same transformation; for example, dist/*.js Minify => dist/*.min.js does not create a cycle.

    Patterns as targets can only appear right of generative transformations. For each prerequisite found, mekano performs a pattern transposition. Currently the system is pretty limited and only the symbols ** and * are accounted for: it copies everything from the left of the first start, and right of the last star.

    For example, for a relation src/**/*.c Compile => obj/**/*.o, if a file src/a/foo.c was found, the target pattern is expanded to obj/a/foo.o. On the other hand, a pattern like src/**/foo/*.c won't work properly because /foo/ will be lost in the process.

    Generative pattern transposition is planned to be hugely improved in the future.


    Value bind grammar is as below:

    bind = value-name, "=", interpolation, ";" ;
    value-name = identifier ;

    A value cannot be unbound or overridden, but can be rebound in inner scopes. Interpolations can refer to existing values with $name or $(name). Example:

    bin = `node_module/.bin` ;
    coffee = `$bin/coffee` ;

    A bound value is available anywhere in the mekanofile, even before the declaration. The order of declaration does not matter; but you cannot have circular references.

    File update

    Once the mekanofile has been read, mekano executes the steps below.

    • Build a graph of all the files according with the relations.
    • Topologically sort the files considering their dependencies.
    • Determine the imprint of each file involved. The imprint is a MurmurHash accounting for a file prerequisites and timestamp.
    • Determine which transformations are needed to be run by comparing imprints to the update log generated by the previous run of mekano, if any.
    • Remove orphans: old generated files that are not in the graph anymore.
    • Invoke recipes in order to update files. When possible, recipes are called asynchronously to make the update faster.

    Any output directory containing targets is automatically created by mekano during the update.

    The update log is located in .mekano/log.json and contains the imprint of each generated file. Never delete the log file! To do a whole rebuild, use the clean command instead. If you delete the log, bad things will happen, because mekano will consider all the generated files so far as sources. This means, for example, that minified files (foo.min.js) will be minified again (foo.min.min.js).

    When you are using a version control tool, like git(1): if you are adding built files to the repository (and not just the sources), then you shall add the log file as well. Otherwise, people checking out the repo. will get trouble because mekano won't know which file is generated and which is not.

    Known limitations

    It is generally planned to improve upon those limitations.

    • no logic, no 'if', limited semantics;
    • pattern transposition for generative transformations is very basic;
    • no dependency-only prerequisites, no parsing of preprocessors dependency files (eg. gcc -MM);
    • no recipe-specific or transformation-specific binds;
    • a bug appears, for the command `watch' only, when a file is renamed;
    • generative transformations accept only one pattern prerequisite;
    • may be too slow for medium or large projects.

    See also the ROADMAP.


    Why not reusing the make syntax?

    The classic make syntax "targets: prerequisites" is not employed because:

    • it may not be very clear how to express transformation chains (something like foo: bar: glo?);
    • inference is done the other way around than make (it infers targets based on prerequisites; make does the opposite with rules like %.o: %c).

    Shout out

    To the masters from which mekano is inspired:


    npm i mekano

    DownloadsWeekly Downloads






    Last publish


    • undashes