Codebase list diet-ng / HEAD
HEAD

Tree @HEAD (Download .tar.gz)

Diet-NG
=======

Diet is a generic compile-time template system based on an XML-like structure. The syntax is heavily influenced by [pug](https://pugjs.org/) (formerly "Jade") and [Haml](http://haml.info/) and outputting dynamic HTML is the primary goal. It supports pluggable transformation modules, as well as output modules, so that many other uses are possible.

See the preliminary [Specification](SPEC.md) for a syntax overview.

This repository contains the designated successor implementation of the [`vibe.templ.diet` module](https://vibed.org/api/vibe.templ.diet/) of [vibe.d](https://vibed.org/). The current state is almost stable and feature complete and ready for pre-production testing.

[![DUB link](https://img.shields.io/dub/v/diet-ng.svg)](https://code.dlang.org/packages/diet-ng)
[![Build Status](https://github.com/rejectedsoftware/diet-ng/actions/workflows/ci.yml/badge.svg)](https://github.com/rejectedsoftware/diet-ng/actions/workflows/ci.yml)


Example
-------

	doctype html
	- auto title = "Hello, <World>";
	html
		head
			title #{title} - example page
		body
			h1= title
			h2 Index
			ol.pageindex
				- foreach (i; 0 .. 3)
					li: a(href="##{i}") Point #{i}
			- foreach (i; 0 .. 3)
				h2(id=i) Point #{i}
				p.
					These are the #[em contents] of point #{i}. Multiple
					lines of text are contained in this paragraph.

Generated HTML output:

	<!DOCTYPE html>
	<html>
		<head>
			<title>Hello, &lt;World&gt; - example page</title>
		</head>
		<body>
			<h1>Hello, &lt;World&gt;</h1>
			<h2>Index</h2>
			<ol class="pageindex">
				<li><a href="#0">Point 0</a></li>
				<li><a href="#1">Point 1</a></li>
				<li><a href="#2">Point 2</a></li>
			</ol>
			<h2 id="0">Point 0</h2>
			<p>These are the <em>contents</em> of point 0. Multiple
			lines of text are contained in this paragraph.</p>
			<h2 id="1">Point 1</h2>
			<p>These are the <em>contents</em> of point 1. Multiple
			lines of text are contained in this paragraph.</p>
			<h2 id="2">Point 2</h2>
			<p>These are the <em>contents</em> of point 2. Multiple
			lines of text are contained in this paragraph.</p>
		</body>
	</html>


Implementation goals
--------------------

- Be as fast as possible. This means moving as many operations from run time to
  compile time as possible.
- Avoid any dynamic memory allocations (unless it happens in user code)
- Let the generated code be fully `@safe` (unless embedded user code isn't)
- Be customizable (filters, translation, DOM transformations, output
  generators), without resorting to global library state
- Operate on ranges. HTML output is written to an output range, input ranges
  are supported within string interpolations and filters/translation support
  is supposed to be implementable using ranges (the latter part is not yet
  implemented).


Experimental HTML template caching
----------------------------------

Since compiling complex Diet templates can slow down the overall compilation
process, the library provides an option to cache and re-use results. It is
enabled by defining the version constant `DietUseCache` (
`"versions": ["DietUseCache"]` in dub.json or `versions "DietUseCache"` in
dub.sdl). It is not recommended to use this feature outside of the usual
edit-compile-run development cycle, especially not for release builds.

Once enabled, the template compiler will look for `*_cached_#####.d` files in the
"views/" folder, where the `*` consists of the full path of the Diet template
and `#####` represents a unique hash value that identifies the contents of the
template, as well as included/extended ones. If found, it will simply use the
contents of that file instead of going through the whole compilation process.

At runtime, during initialization, the program will then output the contents of
all newly compiled templates to the "views/" folder. For that reason it is
currently **important that the program is run with the current working directory
set to the package directory!** A drawback of this method is that outdated
cached templates will not be deleted automatically. It is necessary to clear all
`*_cached_*.d` files by hand from time to time.

*Note that hopefully this feature will be obsoleted soon by the [work of Stefan
Koch on DMD's CTFE engine](https://github.com/UplinkCoder/dmd/commits/newCTFE).*

Experimental HTML Live Mode
----------------------------------

Building a diet template at compile-time can be slow, as mentioned above. A
major drawback of this is that during development, any single change to any
diet file requires a complete rebuild of the entire project.

The library now supports a "Live Mode", where any changes to the templates that
are strictly HTML related will be rendered on a page refresh, instead of
requiring a recompilation. This works by replacing output of the HTML portions
of the template with output from a lookup table of strings. Then the strings
are rebuilt whenever the file changes. So for example, adding or removing a
class from an html element, or fixing `herf` to `href` in an anchor element
does not require a recompile.

Changes to code portions of the template (i.e. string interpolations such as
`#{expression}` or `!{expression}`, or any D code escapes such as `- foreach(x;
range)`) will throw an exception, and force you to recompile your project before
continuing. This is because the diet engine can deal with changes to string
data, but cannot recompile your project for you. And obviously, changing code
outside the templates will not change the rendered pages without a recompile.
Even adding new lines or inserting lines where HTML did not exist is supported.

For example:

```pug
- if(cond)
    - auto a = foobar();
```

changed to the following will not require a recompile

```pug
- if(cond)
    a(href="/") Home
    - auto a = foobar();
```

The mode is enabled by defining the version constant `DietUseLive` (
`"versions": ["DietUseLive"]` in dub.json or `versions "DietUseLive"` in
dub.sdl). It is not recommended to use this in production for the same reasons
listed for the caching mode.

To be as efficient as possible, the templates are only parsed on first access,
and re-parsed only when a modification in any template or dependent template is
detected. Note that it still will not be as efficient as the normal mode which
doesn't require any file i/o to render templates.

There are a few limitations to this approach. Like `DietUseCache`, this
REQUIRES the views directory to be accessible to the running executable. In
addition, to keep the code generation simple (and avoid a full D parser),
certain features do not work with Live Mode. Two such features are type
definitions (i.e. structs, unions, or classes), and static functions. There is
no escape mechanism to allow these, so you will have to ensure that they are not
present in your diet templates, or you will get probably very strange compiler
errors.

Any other problems, please report them in github.

This mode and the `DietUseCache` mode can be combined. Just define both
versions in your project's dub configuration.

Examples Directory
------------------

The examples directory contains 2 projects showcasing the features of diet.

* `htmlgenerator` - Uses diet-ng to generate static html files from diet
  templates.
* `htmlserver` - Simple vibe.d project that shows some features of diet
  template parsing. Note that there are multiple configurations that show how
  the caching and live mode work. Please see the README.md file for more details
  in that directory.