I am trying to transfer an old node-express project over to be able to use es6. I have seen many posts about using gulp with es6. Most of them discuss using a syntax like this:
const gulp = require("gulp");
const babel = require("gulp-babel");
gulp.src('./index.js')
.pipe(
babel({
presets: [
["#babel/env", { modules: false }],
],
})
)
However my existing project's gulpfile does't use gulp.src at all. Instead, it uses gulp-develop-server. The gulpfile looks like this:
const gulp = require("gulp");
const devServer = require("gulp-develop-server");
const spawn = require("child_process").spawn;
const fs = require("fs");
const basedir = ".";
function serverRestart(done) {
// perform some cleanup code here
devServer.restart();
done();
}
function serverStart() {
devServer.listen({
path: basedir + "/index.js",
});
}
function serverWatch() {
serverStart();
gulp.watch(
[
basedir + "/paths/**/*",
// more directories to watch
],
serverRestart
);
}
function reload(done) {
serverWatch();
done();
}
function defaultTask() {
let p;
gulp.watch(["gulpfile.js"], killProcess);
spawnChild();
function killProcess(e) {
if (p && !p.killed) {
devServer.kill();
p.kill("SIGINT");
spawnChild();
}
}
function spawnChild() {
p = spawn("gulp", ["reload"], { stdio: "inherit" });
}
}
process.stdin.resume();
process.on("exit", handleExit.bind(null, { cleanup: true }));
process.on("SIGINT", handleExit.bind(null, { exit: true }));
process.on("uncaughtException", handleExit.bind(null, { exit: true }));
function handleExit(options, err) {
// perform some cleanup code here
if (options.cleanup) {
devServer.kill();
}
if (err) {
console.log(err.stack);
}
if (options.exit) {
process.exit();
}
}
gulp.task("serverRestart", serverRestart);
gulp.task("serverStart", serverStart);
gulp.task("serverWatch", serverWatch);
gulp.task("reload", reload);
gulp.task("default", defaultTask);
The existing flow is important because it executes needed code for setup and cleanup every time I hit save, which runs serverRestart. I've been trying a few different methods based on the other questions which recommended using gulp.src().pipe(), but I havne't had much luck integrating it with the existing pattern which uses gulp-develop-server. I am trying to not have to rewrite the whole gulpfile. Is there a simple way to integrate babel with my existing gulpfile such that I can use es6 in my source code?
There's an example with CoffeeScript in the gulp-develop-server documentation.
Using that as a model, try this:
function serverStart() {
devServer.listen({
path: "./dist/index.js",
});
}
function serverWatch() {
serverStart();
gulp.watch(
[
basedir + "/paths/**/*",
],
serverRestart
);
}
function serverRestart() {
gulp.src('./index.js')
.pipe(
babel({
presets: [
["#babel/env", { modules: false }],
],
})
)
.pipe( gulp.dest( './dist' ) )
.pipe( devServer() );
}
Other suggestions
That being said, your existing Gulp file doesn't actually really use Gulp. That is, everything is defined as a function and it doesn't leverage any of Gulp's useful features, like managing task dependencies. This is because (pre-es6), this was a very simple project. The Gulp tasks in that file are an over-elaborate way to watch files and run a server. The same could be done (with less code) using nodemon.
With the introduction of React and more complicated build processes, Gulp seems to have fallen out of favor with the community (and in my personal experience, Gulp was a time sinkhole anyhow).
If the main change you want to make is to use import, you can simply use a more recent Node version. You'll surely run into the error SyntaxError: Cannot use import statement outside a module. Simply rename the file to .mjs and it will work. This provides a way to incrementally migrate files to import syntax. Other features should automatically work (and are all backwards-compatible, anyhow). Once your project is mostly, or all, compliant, you can add "type": "module" to your package.json file, then rename all of your require-style js files to .cjs, and rename all of your .mjs files to .js, or leave them as .mjs. Read more about the rules of mixing CommonJS and Module imports in the Node.js blog post (note that some things may have changed since that article was written).
Related
After nest build or nest build --webpack dist folder does not contain all required modules and I got Error: Cannot find module '#nestjs/core' when trying to run node main.js.
I could not find any clear instructions on https://docs.nestjs.com/ on how to correctly build app for production, so maybe I missed something?
Out of the box, nest cli does not support including the node_modules dependencies into the dist bundle.
However, there are some community examples of custom webpack configs that include the dependencies in the bundle, e.g. bundled-nest. As described in this issue, it is necessary to include the webpack.IgnorePlugin to whitelist unused dynamic libraries.
bundle-nest has been archived/discontinued:
We've concluded that it is not recommended to bundle NestJS, or actually, NodeJS web servers in general. This is archived for historical reference during the period of time when the community was attempting to tree-shake, and bundle NestJS apps. Refer to #kamilmysliwiec comment for details:
In many real-world scenarios (depending on what libraries are being used), you should not bundle Node.js applications (not only NestJS applications) with all dependencies (external packages located in the node_modules folder). Although this may make your docker images smaller (due to tree-shaking), somewhat reduce the memory consumption, slightly increase the bootstrap time (which is particularly useful in the serverless environments), it won't work in combination with many popular libraries commonly used in the ecosystem. For instance, if you try to build NestJS (or just express) application with MongoDB, you will see the following error in your console:
Error: Cannot find module './drivers/node-mongodb-native/connection' at webpackEmptyContext
Why? Because mongoose depends on mongodb which depends on kerberos (C++) and node-gyp.
Well, about mongo, you can make some exceptions (leave some modules in node_modules), can you? It's not like it's all or nothing. But still, I'm not sure you want to follow this path. I've just succeeded with bundling a nestjs application. It was a proof of concept, I'm not sure if it'll go into production. And it was hard, I might have broken something in the process, but at first glance it works. The most complex part was adminjs. It has rollup and babel as dependencies. And in the app code they unconditionally call watch for some reason (UDP noop in production). Anyways, if you'd like to follow this path you should be ready to debug/inspect your packages' code. And you might need to add workarounds as new packages are added to the project. But it all depends on your dependencies, it may be easier than in my case. For a freshly created nestjs + mysql app it was relatively simple.
The config I ended up with (it overrides the nestjs defaults):
webpack.config.js (webpack-5.58.2, #nestjs/cli-8.1.4):
const path = require('path');
const MakeOptionalPlugin = require('./make-optional-plugin');
module.exports = (defaultOptions, webpack) => {
return {
externals: {}, // make it not exclude `node_modules`
// https://github.com/nestjs/nest-cli/blob/v7.0.1/lib/compiler/defaults/webpack-defaults.ts#L24
resolve: {
...defaultOptions.resolve,
extensions: [...defaultOptions.resolve.extensions, '.json'], // some packages require json files
// https://unpkg.com/browse/babel-plugin-polyfill-corejs3#0.4.0/core-js-compat/data.js
// https://unpkg.com/browse/core-js-compat#3.19.1/data.json
alias: {
// an issue with rollup plugins
// https://github.com/webpack/enhanced-resolve/issues/319
'#rollup/plugin-json': '/app/node_modules/#rollup/plugin-json/dist/index.js',
'#rollup/plugin-replace': '/app/node_modules/#rollup/plugin-replace/dist/rollup-plugin-replace.cjs.js',
'#rollup/plugin-commonjs': '/app/node_modules/#rollup/plugin-commonjs/dist/index.js',
},
},
module: {
...defaultOptions.module,
rules: [
...defaultOptions.module.rules,
// a context dependency
// https://github.com/RobinBuschmann/sequelize-typescript/blob/v2.1.1/src/sequelize/sequelize/sequelize-service.ts#L51
{test: path.resolve('node_modules/sequelize-typescript/dist/sequelize/sequelize/sequelize-service.js'),
use: [
{loader: path.resolve('rewrite-require-loader.js'),
options: {
search: 'fullPath',
context: {
directory: path.resolve('src'),
useSubdirectories: true,
regExp: '/\\.entity\\.ts$/',
transform: ".replace('/app/src', '.').replace(/$/, '.ts')",
},
}},
]},
// adminjs resolves some files using stack (relative to the requiring module)
// and actually it needs them in the filesystem at runtime
// so you need to leave node_modules/#adminjs/upload
// I failed to find a workaround
// it bundles them to `$prj_root/.adminjs` using `rollup`, probably on production too
// https://github.com/SoftwareBrothers/adminjs-upload/blob/v2.0.1/src/features/upload-file/upload-file.feature.ts#L92-L100
{test: path.resolve('node_modules/#adminjs/upload/build/features/upload-file/upload-file.feature.js'),
use: [
{loader: path.resolve('rewrite-code-loader.js'),
options: {
replacements: [
{search: /adminjs_1\.default\.bundle\('\.\.\/\.\.\/\.\.\/src\/features\/upload-file\/components\/edit'\)/,
replace: "adminjs_1.default.bundle('/app/node_modules/#adminjs/upload/src/features/upload-file/components/edit')"},
{search: /adminjs_1\.default\.bundle\('\.\.\/\.\.\/\.\.\/src\/features\/upload-file\/components\/list'\)/,
replace: "adminjs_1.default.bundle('/app/node_modules/#adminjs/upload/src/features/upload-file/components/list')"},
{search: /adminjs_1\.default\.bundle\('\.\.\/\.\.\/\.\.\/src\/features\/upload-file\/components\/show'\)/,
replace: "adminjs_1.default.bundle('/app/node_modules/#adminjs/upload/src/features/upload-file/components/show')"},
],
}},
]},
// not sure what babel does here
// I made it return standardizedName
// https://github.com/babel/babel/blob/v7.16.4/packages/babel-core/src/config/files/plugins.ts#L100
{test: path.resolve('node_modules/#babel/core/lib/config/files/plugins.js'),
use: [
{loader: path.resolve('rewrite-code-loader.js'),
options: {
replacements: [
{search: /const standardizedName = [^;]+;/,
replace: match => `${match} return standardizedName;`},
],
}},
]},
// a context dependency
// https://github.com/babel/babel/blob/v7.16.4/packages/babel-core/src/config/files/module-types.ts#L51
{test: path.resolve('node_modules/#babel/core/lib/config/files/module-types.js'),
use: [
{loader: path.resolve('rewrite-require-loader.js'),
options: {
search: 'filepath',
context: {
directory: path.resolve('node_modules/#babel'),
useSubdirectories: true,
regExp: '/(preset-env\\/lib\\/index\\.js|preset-react\\/lib\\/index\\.js|preset-typescript\\/lib\\/index\\.js)$/',
transform: ".replace('./node_modules/#babel', '.')",
},
}},
]},
],
},
plugins: [
...defaultOptions.plugins,
// some optional dependencies, like this:
// https://github.com/nestjs/nest/blob/master/packages/core/nest-application.ts#L45-L52
// `webpack` detects optional dependencies when they are in try/catch
// https://github.com/webpack/webpack/blob/main/lib/dependencies/CommonJsImportsParserPlugin.js#L152
new MakeOptionalPlugin([
'#nestjs/websockets/socket-module',
'#nestjs/microservices/microservices-module',
'class-transformer/storage',
'fastify-swagger',
'pg-native',
]),
],
// to have have module names in the bundle, not some numbers
// although numbers are sometimes useful
// not really needed
optimization: {
moduleIds: 'named',
}
};
};
make-optional-plugin.js:
class MakeOptionalPlugin {
constructor(deps) {
this.deps = deps;
}
apply(compiler) {
compiler.hooks.compilation.tap('HelloCompilationPlugin', compilation => {
compilation.hooks.succeedModule.tap(
'MakeOptionalPlugin', (module) => {
module.dependencies.forEach(d => {
this.deps.forEach(d2 => {
if (d.request == d2)
d.optional = true;
});
});
}
);
});
}
}
module.exports = MakeOptionalPlugin;
rewrite-require-loader.js:
// https://developer.mozilla.org/en-US/docs/Web/JavaScript/Guide/Regular_Expressions#escaping
function escapeRegExp(string) {
return string.replace(/[.*+?^${}()|[\]\\]/g, '\\$&'); // $& means the whole matched string
}
function processFile(source, search, replace) {
const re = `require\\(${escapeRegExp(search)}\\)`;
return source.replace(
new RegExp(re, 'g'),
`require(${replace})`);
}
function processFileContext(source, search, context) {
const re = `require\\(${escapeRegExp(search)}\\)`;
const _d = JSON.stringify(context.directory);
const _us = JSON.stringify(context.useSubdirectories);
const _re = context.regExp;
const _t = context.transform || '';
const r = source.replace(
new RegExp(re, 'g'),
match => `require.context(${_d}, ${_us}, ${_re})(${search}${_t})`);
return r;
}
module.exports = function(source) {
const options = this.getOptions();
return options.context
? processFileContext(source, options.search, options.context)
: processFile(source, options.search, options.replace);
};
rewrite-code-loader.js:
function processFile(source, search, replace) {
return source.replace(search, replace);
}
module.exports = function(source) {
const options = this.getOptions();
return options.replacements.reduce(
(prv, cur) => {
return prv.replace(cur.search, cur.replace);
},
source);
};
The supposed way to build the app is:
$ nest build --webpack
I didn't bother with source maps, since the target is nodejs.
It's not a config you can just copy-paste, you should figure out what's needed for your project yourself.
One more trick here, but well, you probably won't need it.
UPD adminjs seems to come with prebuilt bundles, so this config may be significantly simpler.
During my Grunt build process, I would like to set the script tag in my index.html which references the Google Maps API dynamically based on process.env.NODE_ENV.
Something like:
let googleMapsUrl;
if (process.env.NODE_ENV === 'development') {
googleMapsUrl = '//maps.googleaps.com/maps/api/js?v=3.exp&libraries=visualization';
} else {
googleMapsUrl = `//maps.googleaps.com/maps/api/js?key=${process.env.GOOGLE_MAPS_API_KEY}v=3.exp&libraries=visualization`;
}
My question is how do I then insert googleMapsUrl into a script tag in index.html like so: <script src=googleMapsUrl></script>
There are many options that you could use. From simple, like grunt-replace to more advanced ones like grunt-processhtml. I will describe the first one as it’s a good option for simple tasks and requires just a simple configuration.
Using grunt-replace
grunt-replace will search for simple variable definitions like ##foo in your source files, and replace those variables with your supplied value or the returning value from a callback.
First install the plugin using npm install grunt-replace --save-dev
Then configure the task as follows:
replace: {
dist: {
options: {
patterns: [
{
match: 'gmaps',
replacement: function() {
var googleMapsUrl;
if(process.env.NODE_ENV === 'development') {
googleMapsUrl = '//maps.googleaps.com/maps/api/js?v=3.exp&libraries=visualization';
} else {
googleMapsUrl = '//maps.googleaps.com/maps/api/js?key=' + process.env.GOOGLE_MAPS_API_KEY + 'v=3.exp&libraries=visualization';
}
return googleMapsUrl;
}
}
]
},
files: [
{
expand: true,
flatten: true,
src: ['path/to/your/source/file.html'], dest: 'destination/folder/'
}
]
}
}
Load the plugin using grunt.loadNpmTasks('grunt-replace') and then add it to your build process.
On your source files just add the defined variable as your script source:
<script src="##gmaps"></script>
The task will replace ##gmaps with the returning value from the defined callback.
Hope it helps.
More about plugin can be found here.
I want to build a quick nodejs script to package a Typescript app as SystemJS modules, a lot like what Angular2 bundles look like.
I tried different configurations but I can't seem to put my finger on it, and haven't found clear enough documentation as of yet.
Note that for this "test", I am not using Gulp or Jspm at all, just systemjs-builder for the time being (and don't plan on using jspm at all either)
Here's what my "project" looks like:
---- Project's Root
-------- index.ts // export * from './modules/index' and eventually more
-------- modules
------------ index.ts // export * from './menu/index'
------------ menu
---------------- menu.component.ts // export class
---------------- menu.service.ts // export class
I want to package this under a single file, where I will have multiple SystemRegister modules that can be consumed in an app thereafter
I tried the following without success:
var Builder = require('systemjs-builder');
// optional constructor options
// sets the baseURL and loads the configuration file
var builder = new Builder('./modules');
builder.bundle('./modules/index.ts', {
/* SystemJS Configuration Here */
baseURL: './modules',
transpiler: 'typescript',
typescriptOptions: {
"module": "system",
"emitDecoratorMetadata": true,
"experimentalDecorators": true
},
defaultExtension: 'ts',
packages: {
'modules': {
defaultExtension: 'ts'
}
}
}, 'infrastructure.js')
.then(function() {
console.log('Build complete');
})
.catch(function(err) {
console.error(err);
})
First of all, the defaultExtension options doesn't seem to work at all
So when I do import {something} from 'filePath'; (without extension), it tries to load filePath, instead of filePath.ts;
Second, if I try adding the .ts extension in my imports (which I don't want to do), it complains that the code is invalid (unexpected token #, unexpected token menuItem and so forth)
Anyone have a good example or some explanations on how this is supposed to work?
Thank you
here you have an example: angular typescript skeleton
build task looks like this:
const path = require('path');
const Builder = require('jspm').Builder;
const builder = new Builder();
const packageJson = require(path.join(config.projectDir, 'package.json'));
return beginBuild()
.then(buildSFX)
.catch((err) => console.log('Build Failed', err));
function beginBuild() {
builder.reset();
return builder.loadConfig(path.join(config.projectDir, packageJson.jspm.configFile))
}
function buildSFX() {
const appName = packageJson.name;
const distFileName = `${appName}.min.js`;
const outFile = path.join(config.distDir, distFileName);
const moduleName = 'app';
const buildConfig = {
format: 'global',
minify: true,
sourceMaps: true
};
return builder.buildStatic(moduleName, outFile, buildConfig);
}
and jspm conf looks like this:
System.config({
defaultJSExtensions: true,
transpiler: "typescript",
typescriptOptions: {
"tsconfig": "src/tsconfig.json"
},
paths: {
"github:*": "vendor/jspm_packages/github/*",
"npm:*": "vendor/jspm_packages/npm/*",
"app": "src/index"
}
/// ...
}
Why do you want to bundle typescript? Bundling is a method used for optimizing the delivery of source code to the browser. The browser doesn't know typescript, it only knows javascript (unless you do on the fly transpiling).
I am using gulp with browserify and factor-bundle.
I have the following code:
b = browserify({
entries: [ 'a.js', 'b.js'],
plugin: [ [ 'factor-bundle', { outputs: [ 'build/a.js', 'build/b.js' ] } ] ]
})
.bundle()
.pipe(source('bundle.js'))
.pipe(buffer())
.pipe(gulp.dest('/build/common'));
I want to pipe some actions (like uglify, bundle-collapser or other job) on the parial bundles ('build/a.js' and 'build/b.js'). I tried to use the method described on the factor-bundle's page:
b.plugin('factor-bundle', { outputs: [ write('x'), write('y') ] });
function write (name) {
return concat(function (body) {
console.log('// ----- ' + name + ' -----');
console.log(body.toString('utf8'));
});
}
But I don't understand the write() method and don't know how to perform uglification and how to gulp.dest the result.
Any idea? explanation?
The write() method returns a writable stream that allows you to pipe bundles
generated by the factor-bundle plugin through further downstream transformations.
For instance, your write() method may look something like this:
var path = require('path');
var file = require('gulp-file');
var sourcemaps = require('gulp-sourcemaps');
function write (filepath) {
return concat(function (content) {
// create new vinyl file from content and use the basename of the
// filepath in scope as its basename.
return file(path.basename(filepath), content, { src: true })
// uglify content
.pipe(uglify())
// write content to build directory
.pipe(gulp.dest('./build/scripts'))
});
}
And you would use it like this:
browserify({
entries: [ 'a.js', 'b.js'],
plugin: [ [ 'factor-bundle', { outputs: [ write('a.js'), write('b.js') ] } ] ]
})
.bundle()
.pipe(write('common.js'))
// Could have use these instead, but it wouldn't be as DRY.
// .pipe(source('common.js'))
// .pipe(uglify())
// .pipe(gulp.dest('./build/scripts'))
Using the factor-bundle plugin affects the output of browserify after
.bundle() is called. Normally, it would generate bundles as readable streams
mapping to each of your entry files, then you would be able to apply further
transformations to them.
Instead you will get a single readable stream that contains a bundle with the
shared common modules from the supplied entry files, which I have called
common.js on the example above. Then you need to handle the transfomations
of the readable streams that map to each entry file separately.
In the example above I have added writable streams to the outputs array, arranged
in the same order as my entry files, which receive their respective bundle as
readable stream and apply further transformations to them
You could also leverage the factor.pipeline event:
var b = browserify({ ... });
b.on('factor.pipeline', function (id, pipeline) {
pipeline.get('wrap').push(write(id));
});
b.plugin(factor);
return b.bundle().pipe(write('common.js'));
I think it is worth noting that applying further downstream work to the outputs
is completely detached from the pipeline. So if you were using gulp and returned
the stream from browserify, the task would have completed prematurely because
it would still be performing operations on the entry files. I haven't run into
issues with this yet.
Hope this helps.
This is a bit old, but it might be usefull to someone else.
The answer above from #Christian helped me, but i had to solve the issue of the task completion. I did it by adding a counter for opened streams, and calling the task callback once they are all closed.
gulp.task('build:js:compile', function(cb) {
const apps = getAllJavascriptFilesPaths(); // this returns an array of full path to the js files i want to bundle
const dist = 'dist'; // the output path
const files = [];
const streams = [];
let openedStreams = 0;
// We use browserify factor-bundle to get the shared code in a separated js file, and not in all apps files
// The write function here handles the post processing of each browserified app by returning a writable stream
// We check the number of opened streams, and call the callback once they are all closed, to be sure the task is
// complete
function write(filepath) {
openedStreams++;
return concat(function (content) {
// create new vinyl file from content and use the basename of the
// filepath in scope as its basename.
return file(path.basename(filepath), content, { src: true })
.pipe(uglify())
.pipe(gulp.dest(dist))
.on('finish', function () {
openedStreams--;
if (openedStreams == 0) {
cb();
}
});
});
}
apps.forEach(function (file) {
files.push(file);
streams.push(write(file)));
});
browserify(files)
.plugin(factor, { outputs: streams })
.transform("babelify", {presets: 'babel-preset-env'})
.bundle()
.pipe(write('common.js'));
});
I'm trying to set the current Git SHA in my project's Grunt configuration, but when I try to access it from another task it isn't available, What am I missing?
grunt.registerTask('sha', function () {
var done = this.async();
grunt.util.spawn({
cmd: 'git',
args: ['rev-parse', '--short', 'HEAD']
}, function (err, res) {
if (err) {
grunt.fail.fatal(err);
} else {
grunt.config.set('git', {sha: res.stdout});
if (grunt.option('debug') || grunt.option('verbose')) {
console.log("[sha]:", res.stdout);
}
}
done();
});
});
After running the task, I expect the config to be available in another task configuration:
requirejs: {
dist: {
...
out: '<%= app.dist %>/scripts/module_name.<%= git.sha %>.js'
...
}
}
So... What's the problem?
The problem is that Require JS is writing to the file public/scripts/module_name..js, the SHA is not available in the configuration (when the filename should be public/scripts/module_name.d34dc0d3.js).
UPDATE:
The problem is that I'm running requirejs tasks with grunt-concurrent, so the Grunt configuration is not available for requirejs.
grunt.registerTask('build', [
...
'getsha',
'concurrent:dist',
...
]);
And the concurrent task, looks like:
concurrent: {
dist: [
...
'requirejs',
...
]
}
Since grunt-concurrent will spawn tasks in child processes, they do not have access to the context of the parent process. Which is why doing grunt.config.set() within the parent context is not available in the config of the child context.
Some of the solutions to make the change available in the child context are:
Write the data to the file system
Write the data to a temporary file with grunt.file.write('./tmp/gitsha', res.stdout) and then have the task being ran in a child process read the temporary file:
out: (function() {
var out = grunt.config('app.dist') + '/scripts/module_name.';
if (grunt.file.exists('./tmp/gitsha')) {
out += grunt.file.read('./tmp/gitsha');
} else {
out += 'unknown';
}
return out + '.js';
}())
Use a socket
This is a very convoluted solution but a solution nonetheless. See the net node docs: http://nodejs.org/api/net.html#net_net_createserver_options_connectionlistener on creating a server on the parent process then have the child process connect to the socket for the data.
Or check out https://github.com/shama/blackbox for a library that makes this method a bit simpler.
Fork the parent process instead of spawn/exec
Another method is to use fork: http://nodejs.org/api/child_process.html#child_process_child_process_fork_modulepath_args_options instead of grunt-concurrent. Fork lets you send messages to child processes with child.send('gitsha') and receive them in the child with process.on('message', function(gitsha) {})
This method also can get very convoluted.
Use a proxy task
Have your sha task set the config as you're currently doing:
grunt.registerTask('sha', function() {
grunt.config.set('git', { sha: '1234' });
});
Change your concurrent config to call a proxy task with the sha:
grunt.initConfig({
concurrent: {
dist: [
'proxy:requirejs:<%= git.sha %>'
]
}
});
Then create a proxy task that runs a task with setting the passed value first:
grunt.registerTask('proxy', function(task, gitsha) {
grunt.config.set('git', { sha: gitsha });
grunt.task.run(task);
});
The above can be simplified to set values specifically on requirejs but just shown here as a generic example that can be applied with any task.