Assemble static HTML files in parallel using jquery, jsrender and child_process nodejs module, starting from json (and xlsx) files to describe data (products, people... etc), all integrated in a maven like build lifecycle based on conventions against configuration.
Actually we build a medium complexity site of ~4500 pages in ~120secs on a quad-core iMac, with SSD. The same site built on an High-IO EC2 instance takes ~60secs for generation. By default we launch a builder for each core we detect so that they can build pages in parallel.
NOTE: as of 0.5.0, we dropped JSDOM to use cheerio,
and generated a complex site of 11208 pages on an AWS EC2 i2.8xlarge in 24secs.
Unfortunately cheerio does not support pseudo selectors like :eq, :odd, :even,
but you can use the function version like this:
Building web sites often you find your self using a CMS, integrating it with some legacy/es, tuning web servers, application servers, databases, and when it is not enough, go back to the code and squeeze anything you can, and put some kind of caching around, often more than one type, and of course keep those caches in sync. This is a pretty full contact sport.
The thing i always noticed doing that is that all this work often is about producing a file or memory buffer (depending on wich caching type you use) containing the content being actually served to the user.
So the typical runtime flow of a MISS request would be:
user request > build the buffer hocus pocus > cache > response
This is really about having users pull your pages. Why dont'we push the site to the user?
do the hocus pocus/integrations batch > put pages on a CDN near the user < have the user acces things fast
Obviously not all sites may be done like that, but many could, saving you a lot of time.
Install this grunt plugin next to your project'
Then add this line to your project's Gruntfile.js gruntfile:
grunt.loadNpmTasks('grunt-html-builder');The simplest way to get started is:
$ git clone git://github.com/aaaristo/html-builder-sample.git
This will bring you a preconfigured project with most common options, that contains 4 files:
$ find ./ -type f ./Gruntfile.js ./package.json ./README.md ./src/js/page/home.js
Let's skip the Gruntfile.js for the moment. (as usal it contains various grunt tasks configurations)
Looking at the package.json you can see which dependencies will be installed to make the thing work:
{
...
"devDependencies": {
"grunt": "~0.4.0",
"grunt-contrib-watch": "~0.3.1",
"grunt-contrib-clean": "~0.4.0",
"grunt-contrib-connect": "~0.2.0",
"grunt-html-builder": "~0.4.20",
"grunt-contrib-copy": "~0.4.0",
"grunt-s3": "~0.2.0-alpha.1",
"grunt-cloudfront-clear": "0.0.1",
"grunt-contrib-compass": "~0.2.0"
},
...
}Now you have to run npm install from the project directory to install those dependencies.
Once you get the dependencies installed you can finally run grunt and you should see somenthing like:
Running "html-builder-json" task Running "html-builder" task >> Evaluating home.js... >> Launching builders... >> builder[0]: inited >> builder[0]: Generated page dist/index.html Running "copy:client" (copy) task Done, without errors.
Nice, what happened is that html-builder found a file home.js under the directory src/js/page, containing only this line of javascript:
page({ path: 'index' });The page function tells the builder to produce a page that is described by the object passed as parameter, actually we are telling the builder we want a page called index.html, and the page is created under the dist directory inside of your project. So now you should have a dist/index.html that look like this:
<!DOCTYPE html>
<html>
<head>
<title>Test site</title>
</head>
<body>
</body>
</html>Not so impressive... This is the base html created by the builder, and you can but you can completely adjust as you wish by creating a file src/html/html.html, let's suppose with H5BP http://html5boilerplate.com/:
<!DOCTYPE html>
<!--[if lt IE 7]> <html class="no-js lt-ie9 lt-ie8 lt-ie7"> <![endif]-->
<!--[if IE 7]> <html class="no-js lt-ie9 lt-ie8"> <![endif]-->
<!--[if IE 8]> <html class="no-js lt-ie9"> <![endif]-->
<!--[if gt IE 8]><!--> <html class="no-js"> <!--<![endif]-->
<head>
<meta charset="utf-8">
<meta http-equiv="X-UA-Compatible" content="IE=edge,chrome=1">
<title></title>
<meta name="description" content="">
<meta name="viewport" content="width=device-width">
<!-- Place favicon.ico and apple-touch-icon.png in the root directory -->
<link rel="stylesheet" href="/css/normalize.css">
<link rel="stylesheet" href="/css/main.css">
<script src="/js/vendor/modernizr-2.6.2.min.js"></script>
</head>
<body>
<!--[if lt IE 7]>
<p class="chromeframe">You are using an <strong>outdated</strong> browser. Please <a href="http://browsehappy.com/">upgrade your browser</a> or <a href="http://www.google.com/chromeframe/?redirect=true">activate Google Chrome Frame</a> to improve your experience.</p>
<![endif]-->
<!-- Add your site or application content here -->
<p>Hello world! This is HTML5 Boilerplate.</p>
<script src="//ajax.googleapis.com/ajax/libs/jquery/1.9.1/jquery.min.js"></script>
<script>window.jQuery || document.write('<script src="js/vendor/jquery-1.9.1.min.js"><\/script>')</script>
<script src="/js/plugins.js"></script>
<script src="/js/main.js"></script>
<!-- Google Analytics: change UA-XXXXX-X to be your site's ID. -->
<script>
var _gaq=[['_setAccount','UA-XXXXX-X'],['_trackPageview']];
(function(d,t){var g=d.createElement(t),s=d.getElementsByTagName(t)[0];
g.src='//www.google-analytics.com/ga.js';
s.parentNode.insertBefore(g,s)}(document,'script'));
</script>
</body>
</html>Rerun grunt and try the site with grunt listen (grunt-contrib-connect + grunt-contrib-watch). Ok now we have nice html but we miss some resources, right? So create the src/client/ directory so that you have a root for files that have to be accessed by the browser and place H5BP resources in there like this:
src/client src/client/css src/client/css/main.css src/client/css/normalize.css src/client/img src/client/img/.gitignore src/client/js src/client/js/main.js src/client/js/plugins.js src/client/js/vendor src/client/js/vendor/jquery-1.9.1.min.js src/client/js/vendor/modernizr-2.6.2.min.js
and if you don't have the grunt listen running give it a grunt client to make grunt copy the files and directories under src/client to the dist directory. Now once you have the grunt listen active the are kept in sync by grunt (grunt-contrib-watch + grunt-contrib-copy). The same is true for page generation... Try to edit the html.html file and see the results on your browser by refreshing the page.
Ok, probably you have a more complex site to build than this, so lets take a look at how the builder builds your pages:
-
as seen it uses the src/html/html.html to start (or a default html if it is not provided)
-
it goes throught all the files in src/js/page to understand which pages to build
-
your files tell him how many pages to build by simply calling the page function many times, to experiment try to change the home.js file to read:
_(100).times(function (i) { page({ path: 'index'+i }); });
yes... it dumbly creates 100 pages. But you now know you can use underscore.js in those files by default, and that the path attribute tells the builder which html files to create. There is another thing to note: all those pages from now on are of the same type for the builder, the type is a machine name given by the name of the js file that asks the creation of a page, in this case home type. So yes probably the home.js in almost any site will call the page function 1 time (so revert this dumb thing).
-
Now lets suppose you want to create a page for every people in your company, create a src/js/page/person.js like this:
['John','Jane','Dave','Mike']
.forEach(function (person)
{
page({ path: 'person/'+person.toLowerCase(), title: person });
});so ok you got it... Open html.html and replace the content of the title tag to read:
<title>{{>title}}'s page</title>almost any html is a jsrender template the html.html renders the page object by default, so that you can use it to store the page title / meta description, etc... But this is not ment to create entire pages.
- Instead you can instruct the builder to use a layout for your pages so that you can benefit from splitting the page in many regions. For example lets suppose you want a sidebar region to navigate between your peoples, and a content region to represent a people. Now you can create this layout by creating file src/html/layout/sidebar.html like this:
<aside data-region="sidebar"></aside>
<section data-region="content"></section>now that you have your layout you should tell the builder to use it in people's pages so add the layout property to the page object:
['John','Jane','Dave','Mike']
.forEach(function (person)
{
page({ layout: 'sidebar', path: 'person/'+person.toLowerCase(), title: person });
});as you can see the content of the layout is prepended to the body of the page (so that you can leave the js at the end for instance).
- Now that you have regions you can use them to place blocks and templates, lets add a sample peoples block to the page: create a src/html/block/peoples.html like this:
<ul>
<li><a href="/person/john.html">John</a></li>
<li><a href="/person/jane.html">Jane</a></li>
</ul>and tell the builder you want it in the sidebar region (time to indent a bit):
['John','Jane','Dave','Mike']
.forEach(function (person)
{
page
({
layout: 'sidebar',
blocks: { sidebar: 'peoples' },
path: 'person/'+person.toLowerCase(),
title: person
});
});the syntax is:
blocks: { region: 'block' }
or of course (to place multiple blocks in order)
blocks: { region1: ['block1','block2',...], region2: ... }
- so blocks are almost always static pieces of html that you want to reuse in various pages, while templates are somenthing you use to render data, lets add template rendered after the peoples block, create src/html/template/person-menu.html:
<ul>
{{for people}}
<li><a href="">{{>#data}}</a></li>
{{/for}}
</ul>and place it after the block calling the template function:
var people= ['John','Jane','Dave','Mike'];
people.forEach(function (person)
{
page
({
layout: 'sidebar',
blocks: { sidebar: ['peoples', template('person-menu',{ people: people })] },
path: 'person/'+person.toLowerCase(),
title: person
});
});as you see we left the href attribute blank, wouldn't it be nice to have single place where i create urls for a page? Yes, so:
var people= ['John','Jane','Dave','Mike'],
href= function (person)
{
return 'person/'+person.toLowerCase();
};
people.forEach(function (person)
{
page
({
layout: 'sidebar',
blocks: { sidebar: ['peoples', template('person-menu',{ people: people })] },
path: href(person),
href: href, // tell the builder to use this function to build hrefs to this page type
title: person
});
});and now i can use the href converter in the template:
<ul>
{{for people}}
<li><a href="{{:~href('person',#data)}}">{{>#data}}</a></li>
{{/for}}
</ul>and voilà in any template that renders a person i can link it..
You can also use a postBuild hook to modify generator results:
var people= ['John','Jane','Dave','Mike'],
href= function (person)
{
return 'person/'+person.toLowerCase();
};
people.forEach(function (person)
{
page
({
layout: 'sidebar',
blocks: { sidebar: ['peoples', template('person-menu',{ people: people })] },
path: href(person),
href: href, // tell the builder to use this function to build hrefs to this page type
title: person,
postBuild: function ($) // this function is serialized and sent to builder processes
{
$('body').addClass(this.person); // so you have to put any data in the page object to reference it
}
});
});The main idea behind this module is to get the data from some kind of legacy system, transform it and render the data to html pages. So ideally you would have this lifecycle phases:
- download: get the data from all the sources you need and place it in JSON format inside data/json/.json
- transform: transform all the json files that you imported in the most usefull form to generate your pages
- render: generate the html from those data
- postBuild: manipulate and arrange the generated html to fit your needs
- deploy: push your html/css/js on some CDN
How to perform those phases? (generally)
- download: write any kind of code that reads your data sources and writes json files into the data/json/ directory, you can even use asynchronous transformations to perform this task.
- transform: you can place javascript files under src/js/transform/.js to declare transformations
- render: use templates / region blocks and src/js/page/.js to render your markup
- postBuild: declare postBuild hooks and use jquery to modify the generated page markup.
- deploy: use the grunt tasks for your CDN provider to push your dist directory on a CDN.
In the render phase the grunt task forks 1 child for any core it finds on the machine where it is running. Those childs will then request jobs from a queue of pages to generate until all pages are generated, and the queue is emptied. To limit the number of builders you can configure the Gruntfile like this:
grunt.initConfig({
'html-builder': {
cpus: 2
....Any JSON file under src/json/ or data/json/ should contain a JSON array of objects. The src/json/ directory is ment for initial collections / manually managed JSON collections (under SCM), where the data/json/ directory is ment as the target directory for transformations (not under SCM). For example, in a company website, you could have a src/json/persons.json which contains:
[{
name: 'John Doe',
role: 'CEO',
googleplusid: '1323221312'
},
{
name: 'Hipster Hacker',
role: 'CTO',
googleplusid: '31232143444'
},
{
name: 'Mark Hustler',
role: 'CFO',
googleplusid: '53453543534'
}]this is from now on a collection. Collections are there to be transformed or rendered. For example, you may want to integrate those info with what is present in the googleplus account of each person, to generate a more meaningful version of this collection in data/json/persons.json. This is what is, called a tranformation. Or you can use this collection in your page file:
var people= collection('persons'), // the collection function returns the corresponding javascript array of the json file
href= function (person)
{
return 'person/'+person.name.toLowerCase();
};
people.forEach(function (person)
{
page
({
layout: 'sidebar',
blocks: { sidebar: ['peoples', template('person-menu',{ people: _.pluck(people,'name') })] },
path: href(person),
href: href,
title: person.name,
postBuild: function ($)
{
$('body').attr('data-googleplusid',this.person.googleplusid);
}
});
});A transformation is a javascript file under src/js/transform/. For example src/js/transform/google-profile.js:
async(); // tells the framework that this is an asynchronous transformation
var request= prequire('request'), // prequire is a way to require your project modules instead of grunt-html-builder deps
config= grunt.config('google'), // you have access to grunt
profile= function (id,cb)
{
var json= '';
request.get('https://www.googleapis.com/plus/v1/people/'+id+'?key='+config.key,
function (err, res, body)
{
cb(JSON.parse(body));
});
};
var profiles= [], log= grunt.log, _= grunt.util._;
grunt.util.async.forEachSeries(json, // inside any transformation you have a "json" variable, that is the collection you are transforming
function (person,done)
{
if (person.googleplusid)
profile(person.googleplusid,function (p)
{
_.extend(person,{ title: (p.organizations[0].title ? p.organizations[0].title : person.title),
aboutMe: p.aboutMe });
profiles.push(person);
done();
});
});
else
{
_.extend(person,{ foto: person.picture });
profiles.push(person);
done();
}
},
function (err)
{
if (err)
done(err);
else
done(null,profiles);
});Once you have a transformation file you can configure it to run in the Gruntfile.js like this:
grunt.initConfig({
'html-builder': {
json: [
{ collection: 'persons', transform: 'google-profile', dest: 'persons' },
...
{ collection: 'persons', transform: 'simple', dest: 'dummy' },
]
....In this case you have a "complex" transformation based on some asynchronous access to external resources. But you can define simpler trasnformations like src/js/transform/simple.js:
json.forEach(function (person, idx)
{
person.seq= idx;
});
done(json)This is a synchronous transform. The only difference is that you don't call async(), and the done function only accepts one argument with the transformed collection.
You can use also xlsx files as a source by placing them in the data/excel/ directory, and transforming them by:
grunt.initConfig({
'html-builder': {
json: [
{ excel: 'persons', transform: 'google-profile', dest: 'persons' },
]
....and the modified version of an ipothetical google-profile.js:
async(); // tells the framework that this is an asynchronous transformation
var request= prequire('request'), // prequire is a way to require your project modules instead of grunt-html-builder deps
config= grunt.config('google'), // you have access to grunt
profile= function (id,cb)
{
var json= '';
request.get('https://www.googleapis.com/plus/v1/people/'+id+'?key='+config.key,
function (err, res, body)
{
cb(JSON.parse(body));
});
};
var workbook= json, profiles= [], log= grunt.log, _= grunt.util._,
persons= workbook.sheet('Persons')
.toJSON(['name', // xlsx columns to object attributes (position / attribute name)
'role'],
function (p) { /*alter p as you whish*/ if (valid) return p; else return undefined; });
grunt.util.async.forEachSeries(persons, // inside any transformation you have a "json" variable, that is the collection you are transforming
function (person,done)
{
if (person.googleplusid)
profile(person.googleplusid,function (p)
{
_.extend(person,{ title: (p.organizations[0].title ? p.organizations[0].title : person.title),
aboutMe: p.aboutMe });
profiles.push(person);
done();
});
});
else
{
_.extend(person,{ foto: person.picture });
profiles.push(person);
done();
}
},
function (err)
{
if (err)
done(err);
else
done(null,profiles);
});When transoforming large amounts of data you may find convenint to use index or mindex functions. Suppose you are transforming a long list of persons and you have a "knows" array that contains ids for persons that the current person knows, but you want the complete object in place instead of only the id: You may:
json.forEach(function (person)
{
person.knows.forEach(function (id,idx)
{
person.knows[idx]= _.findWhere(json,{ id: id });
});
});
done(json);While this will work, it will not be very fast... Because you are scanning the json array N times for each person, to transform. A better approach would be:
var _person= index(json,'id');
json.forEach(function (person)
{
person.knows.forEach(function (id,idx)
{
person.knows[idx]= _person(id);
});
});
done(json);This time it will be really faster, because the index function will create an index (_person) looping 1 time on the json array, when you call * _person* it will access an object by key returning you the result much faster.
The index function is ment for unique indexes (like ids), while the mindex function is ment for non-unique indexes, like:
var _persons= mindex(json,'name');
json.forEach(function (person)
{
person.manyJohn= _persons('John');
});
done(json);where you can have multiple results for one index key, so the result is an array.
You can enable multi-language support by defining the languages propoerty in the Gruntfile configuration, like this:
grunt.initConfig({
'html-builder': {
languages: ['en','it','es','fr']
....this means that the default language for the site is English (the first), and you have some alternative language. From now on the builder will create 4 html files for every path you define in a page file, for example:
index.html it/index.html es/index.html fr/index.html
As you can see the default language has no prefix whether all other languages are prefixed with the prefix you specified in the languages array. We encourage you to use standards. Ho you switch between languages is up to your implementation: you just need to link the current path with the right prefix.
Once you configured the languages array, you can use the t converter in your templates, to translate data or strings:
<p>This is a translated string: {{t:'Hello world'}}</p>
<p>This is a translated object property: {{t:role}}</p>
In order to get your data translated you should define special object properties keyed by language like this:
[{
name: 'John Doe',
role: {
en: 'Sales',
it: 'Vendite',
es: 'Venta',
fr: 'Ventes'
}
}]while to get strings translated you should define a special collection src/json/i18n.json, like this:
{
"Hello world": {
it: 'Ciao mondo',
es: 'Hola mundo',
fr: 'Bonjour tout le monde'
}
}That is actually JSON object where the default language string is the key, and the value is an object keyed by alternative languages.
You can specify different url aliases for different languages in your page.js just configure your href function properly
href= function (entity)
{
return {
it: entity.alias.it,
en: entity.alias.en
}
};
path:
{
it: href(entity).it,
en: href(entity).en
},** ... doc in progress ...**
Things to document:
- html html
- html layouts / regions
- html blocks
- html templates (jsrender)
- client files (images,css,js...)
- filtering pages
- s3 / cloudfront deploy
- RDF / jsonld support
- 2013-05-15 v0.4.20 First documented BETA release
Copyright (c) 2013 Andrea Amerigo Aristodemo Gariboldi
Licensed under the Apache2 license.
