Total Pageviews

Thursday 19 January 2017

PureScript的build工具-pulp

A build tool for PureScript projects.

Installation

Assuming you already have Node set up (and we recommend you also set up NPM to keep your global packages in your home directory), all you need to do to get a working PureScript environment is:
$ npm install -g purescript pulp bower
This installs the PureScript compiler, the Pulp build tool, and the Bower package manager.
Aside: if you're familiar with the JavaScript ecosystem and you're wondering why PureScript uses Bower and not npm, you might be interested to read Why the PureScript community uses Bower. Otherwise, please ignore this and read on.

Getting Started with a Pulp Project

The short version:
$ mkdir purescript-hello
$ cd purescript-hello
$ pulp init
$ pulp run
The structure of your project folder, after running pulp init, will look like this:
  purescript-hello
  - bower.json
  - src/
  - test/
pulp works by convention. It expects all projects to contain a bower.json file, which is needed for package management (which, in PureScript, is handled by Bower). pulp itself currently doesn't use this file for anything other than figuring out where the project root is.
Your project source files go in the src folder. Your test files go in the test folder. Project dependencies will be installed under the Bower standard bower_components folder, and are expected to have the same basic src/test structure. That's all there is to a pulp project.
We employ the purescript- prefix as a convention to identify PureScript projects when they're used as dependencies. You're welcome to call your project anything you like, but without the purescript- prefix it won't be picked up by pulp as a dependency when installed through Bower.

What if I need something a bit more complicated?

If you want to change any of these defaults, you can—pulp offers a number of command line flags to alter its behaviour—but try to avoid using them unless you have a good reason to.
If you get fed up with having to remember long pulp invocations, try using npm as your build tool. pulp's numerous command line flags make it well suited for this.
If that's still not enough, you might try using a more generic build tool, such as webpack with purs-loader, or gulp with gulp-purescript.

Pulp Commands

To get a quick overview of the things pulp can do, you can ask it to give you a list of its available commands:
$ pulp --help
This will print a list of pulp's global command line options, and a list of commands it will accept.
To see the available options for a specific command, you can invoke the command with the --help flag, like this:
$ pulp build --help
This will give you an exhaustive list of ways you can modify the basic behaviour of the command.

Global, Command Specific and Pass-Through Options

Notice that there's a distinction between global command line options and command specific options. Global options must appear before the name of the command, and command specific options must appear after it.
Thus, if you want to run the build command in watch mode (where it will run the command once, then wait and re-run the command whenever you change a source file) you need to put the --watch flag before the command itself, like so:
$ pulp --watch build
On the other hand, if you want to tell the build command to produce optimised code (performing dead code elimination), using the command specific option --optimise, the flag needs to come after the command name:
$ pulp build --optimise

Pass-Through Options

Finally, pulp commands sometimes allows you to pass flags through to the psc compiler. Any options appearing after -- will be passed through to the compiler, or whichever process a pulp command spawns. For instance, if you want to tell psc to skip applying tail call optimisations, you would invoke pulp build like this:
$ pulp build -- --no-tco

Building Projects

At heart, pulp is just a frontend for the PureScript compiler, psc. Its basic function is to compile your project, which you can do by running pulp build. This will simply run psc with all your source files, leaving the compiled JavaScript files in the output folder. These files will all be CommonJS modules, which you can require() using anything which supports CommonJS, such as node.
However, you will usually want to do more with your project than just compile your PureScript code into a jumble of CommonJS modules. pulp provides a number of commands and options for the most common use cases.

Making a JavaScript Bundle

pulp build can also call psc-bundle for you, which is a compiler tool whose job it is to take the output from psc, remove the code which isn't actually being used by your program, and bundle it all up into a single compact JavaScript file.
There are two command line options you can give pulp build to accomplish this, depending on where you want the resulting code. You can use the --optimise flag (or its shorthand alias, -O), which will send the bundled result to standard output, or you can use the --to (or -t) option, passing it a file name, and pulp will store the bundle in a file of that name.
So, you can use either of these methods, which in this example will both have the same effect:
$ pulp build --optimise > hello.js
$ pulp build --to hello.js
Note that using both options (pulp build --optimise --to hello.js) is superfluous. The presence of --to implies the presence of --optimise.

Running Your PureScript Project

If you're developing a Node project using PureScript, you can tell pulp to run it after compiling using the pulp run command. This command will first run pulp build for you, if necessary, then launch your compiled code using node. If you have used any pass-through command line options, these will be passed to the node process.
So, to run the hello world project you get from pulp init, you would simply:
$ pulp run
If you want to pass command line arguments to your application, pulp lets you do that too:
$ pulp run file1.txt file2.txt file3.txt
Beware, though, that these are processed as pass-through options by pulp, so it will first consume any options it recognises as pulp run's command specific options; these will not be passed through to your application.
If you want to run your application using something other than node, pulp lets you do that too, with the --runtime option. For instance, if you've written an application which runs on PhantomJS, you might launch it like this:
$ pulp run --runtime phantomjs

Running Test Suites

pulp has a command pulp test, which works much like pulp run, except it will also compile the code you've placed in your test folder, and instead of running the main function in your Main module, it will use Test.Main. This module should be located in your test folder.
pulp doesn't care what test framework you've chosen, as long as there's a main function in your Test.Main module to be run. If the process exits with a non-zero return code, that means your test suite failed, as far as pulp is concerned, and it will itself exit with an error.
In short, to run your tests:
$ pulp test
To continuously run your tests when you change the source code:
$ pulp --watch test

Running Commands Before and After an Action

It's sometimes useful to kick off a command before or after an action, particularly in combination with the --watch option above. To do this, you can use --before, or --then and --else for successful or failing actions respectively:
$ pulp --watch --before clear build       # Clears the screen before builds.
$ pulp --watch --then 'say Done' build    # On OS X, announces 'Done' after a successful build.
$ pulp --watch --else 'say Failed' build  # Announces 'Failed' if a build failed.

# A more long-winded example combining the three:
$ pulp --watch --before clear --then "say $(basename `pwd`) succeeded." --else 'say $(basename `pwd`) failed.' build

CommonJS Aware Builds

Often, you'll want to go outside PureScript and leverage some of the enormous body of JavaScript code available on NPM. This is such a common use case that pulp provides a command for it: pulp browserify. As the name suggests, this uses Browserify to bundle up your PureScript code with Node style CommonJS dependencies.
For instance, the majority of web UI libraries for PureScript these days depend on either virtual-dom or React as a CommonJS dependency. Here is how you would add React to your project and build a JS bundle with React included (assuming your PureScript code requires it):
$ npm install react
$ pulp browserify --to hello.js
Essentially, pulp browserify --to works exactly like pulp build --to, except it also resolves CommonJS dependencies and includes them in the bundle. The resulting JS file can now be loaded directly into the browser, and everything you need to run your application should be included.
If you omit the --to option, the bundle is piped to standard output. This would thus have the same effect as the example above:
$ pulp browserify > hello.js

Optimising Code Size

pulp browserify will pull code in at the module level by default, so every file required from your entry point will appear in the bundle. The PureScript compiler, as we know, is able to perform dead code elimination on your compiled PureScript code, and we can leverage this in pulp browserify using the --optimise flag.
$ pulp browserify --optimise --to hello.js
Note that, unlike pulp build, --to doesn't automatically imply --optimise. In fact, if you omit --optimise, pulp browserify will not only omit the dead code elimination step, it will also run Browserify as an incremental build, which means it will run considerably faster. You should use --optimise only when you're building production code—when you're developing, you'll probably prefer the much faster compile times provided by Browserify's incremental mode.

Reimporting Browserified Bundles

While browserified bundles are intended to be consumed directly by browsers, you may sometimes prefer to access the bundle from some external code. While it's generally preferable to consume CommonJS modules directly, there are use cases where you might want to provide a single JS file ready to be required by a consumer without needing to deal with installing and resolving dependencies. Browserify provides the --standalone mechanism for that, and pulp browserify supports it:
$ pulp browserify --standalone myBundle --to myBundle.js
This makes a bundle which comes wrapped in a UMD header (meaning it supports both CommonJS and AMD, and will install itself in the global namespace under the name you provided if neither is present), and the exports it provides will be the same as those you export in your Main module.
So, given the example above produces a bundle where a PureScript function Main.main exists, you can access it from JavaScript via CommonJS like this:
var myBundle = require("./myBundle");
myBundle.main();

Building Documentation

PureScript has an inline syntax for documentation, which can be extracted into Markdown files using the psc-docs command. pulp provides the pulp docs command to make this process easy:
$ pulp docs [--with-dependencies]
This extracts the documentation from your source files, and places it in the generated-docs folder under your project's root folder. By default, dependencies are not included, but this can be enabled with the --with-dependencies flag.
You can also extract documentation from your tests, if you like:
$ pulp docs --with-tests

Launching a REPL

The psci interactive shell for PureScript is fantastically useful, but setting it up can be a bit of a chore, especially with a large number of dependencies. That's where pulp psci comes in.
pulp psci will generate a .psci file for your project automatically whenever you invoke it, and launch psci for you directly. It's as simple as:
$ pulp psci

Launching a Development Server

While technically out of scope for a build tool like pulp, a common need when developing client side web apps is a tightly integrated development web server, which takes care of compilation for you on the fly. This is what the purs-loader project is for: it provides a PureScript loader for Webpack, which works with Webpack's development server and makes recompilation seamless: whenever you make a change to your source files, you just switch to your browser and hit the refresh button, and the server will compile and deliver your assets on the fly. No need to wait for the PureScript compiler to finish before switching to the browser.
pulp provides the pulp server command to quickly set up a Webpack development server for your project. It only provides the most basic functionality: it will serve static assets from your project root, and it will serve your compiled JS bundle from /app.js.

A Quick Example

To see how this works, let's set up a project for serving the default hello world app through pulp server.
$ mkdir hello-server
$ cd hello-server
$ pulp init
We need an index.html file to load our compiled PureScript code. Place this in your new hello-server folder:
<!doctype html>
<html>
  <body>
    <h1>Hello sailor!</h1>
    <script src="/app.js"></script>
  </body>
</html>
Now, start the server:
$ pulp server
It will tell you that it's launched a web server at http://localhost:1337/, and after a little while it will tell you that it's finished compiling (bundle is now VALID). If you browse to http://localhost:1337/, you should, in addition to the "Hello sailor!" header on the webpage, see that your PureScript code has printed the text "Hello sailor!" to the console.

I Need More

As mentioned, this is a very bare bones development server. You're likely to quickly need more features if you plan on doing any kind of serious web development. At this point, you'll need to set up your own Webpack configuration using purs-loader. Due to the way Webpack works, it's not really useful to extend pulp server with further configuration options. It's intended as a starting point only.

Dependency Management

pulp is not a package manager, only a build tool. The PureScript community has standardised on Bower as its package manager.
Bower expects the presence of a bower.json file in your project root, in which it records your project’s dependencies and other metadata. pulp init will have installed a basic bower.json file for you along with the project skeleton, but you should read the Bower documentation on the file format and make sure you’ve configured it to your satisfaction before you publish your package. In particular, mind that you’ve added a license field.

Dependency Management Cheat Sheet

This document isn't going to explain how Bower works, or go into details about PureScript dependency management. However, a tl;dr is often enough to get you started and productive without having to dive into yet another package management system. It's going to be especially easy if you're already used to npm. So, here we go.

Installing Dependencies

To install the purescript-profunctor package into your project:
$ bower install purescript-profunctor
To also record this as a dependency in the bower.json file:
$ bower install --save purescript-profunctor
To install every dependency which has been recorded in bower.json as needed by your project:
$ bower install

Housekeeping

To remove an installed package:
$ bower uninstall purescript-profunctor
To remove it from bower.json as well:
$ bower uninstall --save purescript-profunctor
To list all packages installed in your project:
$ bower ls
To update all installed packages to the most recent version allowed by bower.json:
$ bower update

Releasing Packages

Bower packages are installed directly from Git repositories, and versioning follows Git tags. Imagine you've created a new PureScript library for working with zygohistomorphic prepromorphisms (because who doesn't need zygohistomorphic prepromorphisms), called purescript-zygo.
Note that there is a convention of prefixing PureScript package names with purescript-. Please stick with that unless you have an especially good reason not to, as pulp and many other tools expect installed dependencies to follow this convention.
You would start by tagging an initial version:
$ cd /path/to/purescript-zygo
$ pulp version 0.1.0
This runs a few checks to ensure that your package is properly set up for publishing, and if they pass, creates a Git tag v0.1.0.
Once you've tagged a version, all you need to do to make a new release is push that tag to GitHub, register your package in the Bower registry, and upload your package's documentation to Pursuit. Pulp is able to do all of this for you:
$ pulp publish
For subsequent releases, the process is the same: pulp version <newversion> followed by pulp publish. When tagging a new version, pulp version also allows you to supply an argument of the form patch, minor, or major, in addition to specific versions. If you run pulp version patch, for example, Pulp will look through your Git tags to find the version number for the latest release, and then generate the new verision number by bumping the patch component. The minor and major arguments respectively perform minor and major version bumps in the same way.

from https://github.com/bodil/pulp

No comments:

Post a Comment