/BuildGraph

DEPRECATED PROJECT IDEA - A visual node-graph-based cascading build system

Primary LanguageCoffeeScript

NOTE: This readme represents an optimistic future for this project, which is basically just an idea.

BuildGraph

Grunt is confusing and overcomplicated. Seriously.

Gulp has a nice piped structure and clean code, but you have to install and setup plugins to get incremental building, and to do watching at all you have to create a separate task for each task you want to allow watching for and manually specify the set of files that will trigger the task. Same goes for cleanup: make a task and manually tell it everything it needs to clean up.

BuildGraph takes Gulp's piped structure, watches files automatically with incremental rebuilds, and lets you define your build tasks without code (except where you need it).

(Anvil also claims incremental rebuilding of changed files and affected files.)

Other build tools have CLIs; BuildGraph can be run as a command-line GUI or a standalone application.

BuildGraph lets you define your tasks by typing out simple conversion commands. You get autocomplete for files and folders, as well as conversions.

Sources, conversions, and destinations get expanded to blocks.

You end up creating a node graph with connections between blocks. Connections transfer streams of files, and are represented as wires. Connections that handle source maps are represented as doubled wires.

Conversions

BuildGraph can infer most conversions from a source and destination.

For example, app/*.coffee -> build/*.js means compile coffee from app/ to build/ whereas app/*.coffee -> build/* means just copy the files to build.

Other conversions you need to type out, like "strip comments" or "alphabetize lines".

See Conversions.MD for a list of possible conversions.

(Conversions are defined by plugins, of course.)

Tasks

Every build system that I know of (Grunt, Gulp, Cake) has support for multiple tasks.

In BuildGraph, they're nicely represented as tabs.

Cleanup

Gulp and grunt require manual cleanup modules. Or you can use convoluted fs code. Or you can just clean up yourself.

BuildGraph knows what files are originative or not and can clean up for itself at the click of a button.

Console

BuildGraph also has a "console" where you can build a temporary graph/task/command to be executed. (with command history)

This is great for one-off transpilation like js->coffee block!)

You can save commands as tasks.

Custom Code

What about my super special code?

BuildGraph has Run Blocks. Run Blocks can be standalone, but they can also be source blocks, destination blocks, or conversion blocks.

Say you want to operate on a list of files from the output of a program

You'd return a list or stream of files (or filenames?), and then pipe the run block into some other block(s).

Say you want to download the latest tarball of a project on Launchpad

You'd make a run block to scrape the webpage to find the tarball URL and then just pipe that to a directory to download and extract it. Okay actually that might just download it and place it in the folder. Worse case senario you'd have to pipe it to a file first, but hopefully you can express that you want extraction easily. It's a little tricky since a URL could return just about anything. Maybe you'd have to start typing out "extract" to explicitly create an extraction block. You'd need to be able to do that sort of thing for plugins that don't have source/destination clues, like "strip comments" etc.

build-graph

build-graph is a cascading build system.

BuildGraph is a visual node-graph based cascading build system.

You include build-graph in your scripts to run it with npm run build, but you edit it with the nice GUI.

build-graph might actually be able to use gulp internally; it does have plugins for incremental rebuilding and there's gulp-watch (I don't know if that's redundant or what)