minify-html/README.md

hyperbuild

A fast HTML parser, preprocessor, and minifier, written in C. Designed to be used in C projects, but also runnable on Node.js thanks to Emscripten. Minifier heavily influenced by kangax's html-minifier.

Features

Streaming minification

hyperbuild minifies as it parses, directly streaming processed HTML to the output without having to build a DOM/AST or iterate/traverse around in multiple passes, allowing for super-fast compilation times and near-constant memory usage.

Smart parsing

hyperbuild is aware of strings and comments in JS and CSS sections, and deals with them correctly.

Super low level

hyperbuild is written in C, and exposed to Node.js using Emscripten.

Parsing

Current limitations:

• UTF-8 in, UTF-8 out, no BOM at any time.
• Not aware of exotic Unicode whitespace characters.
• Tested and designed for Linux only.
• Follows HTML5 only.

Errors

Errors marked with a ⌫ can be suppressed using the --suppress option. Use the error name without the HBE_PARSE_ prefix.

HBE_PARSE_MALFORMED_ENTITY ⌫

It's an error if the sequence of characters following an ampersand (&) does not form a valid entity.

Technically, entities must be of one of the following forms:

• &name;, where name is a reference to a valid HTML entity
• &nnnn;, where nnnn is a Unicode code point in base 10
• &#xhhhh;, where hhhh is a Unicode code point in base 16

A malformed entity is an ampersand not followed by a sequence of characters that matches one of the above forms. This includes when the semicolon is missing, and bare ampersands (i.e. followed by whitespace or as the last character).

Note that this is different from HBE_PARSE_INVALID_ENTITY, which is when a well-formed entity references a non-existent entity or Unicode code point.

An ampersand by itself is not necessarily an invalid entity. However, HTML parsers and browsers may have different interpretations of bare ampersands, so it's a good idea to always use the encoded form (&).

When this error is suppressed, malformed entities are outputted untouched.

HBE_PARSE_INVALID_ENTITY ⌫

It's an error if an invalid HTML entity is detected.

If suppressed, invalid entities are outputted untouched.

See entityrefs.c for the list of entity references considered valid by hyperbuild.

Valid entities that reference a Unicode code point must be between 0x0 and 0x10FFFF (inclusive).

HBE_PARSE_NONSTANDARD_TAG ⌫

It's an error if an unknown (non-standard) tag is reached. See tags.c for the list of tags considered valid by hyperbuild.

HBE_PARSE_UCASE_TAG ⌫

It's an error if an opening or closing tag's name has any uppercase characters.

HBE_PARSE_UCASE_ATTR ⌫

It's an error if an attribute's name has any uppercase characters.

HBE_PARSE_UNQUOTED_ATTR ⌫

It's an error if an attribute's value is not quoted with " (U+0022). This means that ` and ' are not valid quote marks.

HBE_PARSE_ILLEGAL_CHILD

It's an error if a tag is declared where it can't be a child of. This is a very simple check, and does not cover the comprehensive HTML rules, which involve backtracking, tree traversal, and lots of conditionals.

This rule is enforced in four parts: whitelistparents.c, blacklistparents.c, whitelistchildren.c, and blacklistchildren.c.

HBE_PARSE_UNCLOSED_TAG

It's an error if a non-void tag is not closed. See voidtags.c for the list of tags considered void by hyperbuild.

This includes tags that close automatically because of siblings (e.g. <li><li>), as it greatly simplifies the complexity of the minifier due to guarantees about the structure.

HBE_PARSE_UNEXPECTED_END and HBE_PARSE_EXPECTED_NOT_FOUND

General syntax errors.

Additional errors

There are additional implicit errors that are considered as generic syntax errors due to the way the parser works:

• Closing void tags; see voidtags.c for the list of tags considered void by hyperbuild.
• Placing whitespace between = and attribute names/values.
• Placing whitespace before the tag name in an opening tag.
• Placing whitespace around the tag name in a closing tag.

Options

--in

Path to a file to process. If omitted, hyperbuild will read from stdin, and imports will be relative to the working directory.

--out

Path to a file to write to; it will be created if it doesn't exist already. If omitted, the output will be streamed to stdout.

--keep

Don't automatically delete the output file if an error occurred. If the output is stdout, or the output is a file but --buffer is provided, this option does nothing.

--buffer

Buffer all output until the process is complete and successful. This won't truncate or write anything to the output until the build process is done, but will use a non-constant amount of memory. This applies even when the output is stdout.

--suppress

Suppress errors specified by this option. hyperbuild will quitely ignore and continue processing when otherwise one of the provided errors would occur.

Omit the HBE_PARSE_ prefix. Separate the error names with commas. Suppressible errors are marked with a ⌫ in the Errors section.

Processing

hyperbuild sits somewhere between Server Side Includes and a templating library, and is designed for simplistic compilation of apps statically rather than dynamic generation of live content.

To achieve this, hyperbuild has special directives that allow special action to be taken when it's processing some HTML code. This includes importing files, getting and setting variables, and escaping text for HTML.

Directives are like functions in any common language: they take some arguments, and return some value. In hyperbuild, all arguments are simple strings, and the return value is directly streamed while processing.

Using directives

There are two methods of getting hyperbuild's attention: using a special tag, and using a special entity.

Directive tags

<hb-dir arg1="val1" arg2="val2">valarg</hb-dir>

• Replace dir with a hyperbuild directive name
• The value for the argument value is provided via the inner content of the tag
• All other arguments are provided via attributes
• Directive entities inside argument values, and nested directive tags, will be processed

Directive entities

&hb-dir(arg1=val1, arg2=val2);

• Replace dir with a hyperbuild directive name
• Arguments are provided in name-value pairs between parentheses, separated by commas
• All characters between the = and next , or ) count as the argument's value, including whitespace characters
• To use commas, right parentheses, or ampersands in argument values, use HTML entities (,, ), and &)
• Directive entities inside argument values will be processed

Available directives

import

Read, parse, process, and minify another file, and stream the result.

Argument	Format	Required	Description
path	Relative or absolute file system path	Y	The path to the file. If it starts with a slash, it is interpreted as an absolute path; otherwise, it's a path relative to the directory of the importee, or the working directory if the input is stdin.

Minification

Theory

Whitespace

Beginning and end

Between text and tags

Contiguous

Whole text

Content

Specific tags

Tags not in one of the categories below are specific tags.

Formatting tags

<strong> moat </strong>

Content tags

<p>Some <strong>content</strong></p>

Content-first tags

<li>Anthony</li>

<li>
  <div>
  </div>
</li>

Layout tags

Overview

Type	Content
Formatting tags	Text nodes
Content tags	Formatting tags, text nodes
Layout tags	Layout tags, formatting tags

Options

For options that have a list of tags as their values, the tags should be separated by a comma.

An * (asterisk, U+002A) can be used to represent the complete set of possible tags. It essentially fully enables or disables the option.

For brevity, hyperbuild has built-in sets of tags that can be used in place of declaring all their members; they begin with a $ sign:

Name	Tags	Source
$content	address, audio, button, canvas, caption, figcaption, h1, h2, h3, h4, h5, h6, legend, meter, object, option, p, summary, textarea, video	contenttags.c
$contentfirst	dd, details, dt, iframe, label, li, noscript, output, progress, slot, td, template, th	contentfirsttags.c
$formatting	a, abbr, b, bdi, bdo, cite, data, del, dfn, em, i, ins, kbd, mark, q, rp, rt, rtc, ruby, s, samp, small, span, strong, sub, sup, time, u, var, wbr	formattingtags.c
$layout	blockquote, body, colgroup, datalist, dialog, div, dl, fieldset, figure, footer, form, head, header, hgroup, html, main, map, menu, nav, ol, optgroup, picture, section, select, table, tbody, tfoot, thead, tr, ul	layouttags.c
$specific	area, base, br, code, col, embed, hr, img, input, param, pre, script, source, track	specifictags.c
$heading	hgroup, h1, h2, h3, h4, h5, h6	headingtags.c
$media	audio, video	mediatags.c
$sectioning	article, aside, nav, section	sectioningtags.c
$void	area, base, br, col, embed, hr, img, input, keygen, link, meta, param, source, track, wbr	voidtags.c
$wss	pre, code	wsstags.c

--MXcollapseWhitespace $wss

Reduce a sequence of whitespace characters in text nodes to a single space (U+0020), unless they are a child of the tags specified by this option.

Before	After
<p>↵ ··The·quick·brown·fox↵ ··jumps·over·the·lazy↵ ··dog.↵ </p>	<p>·The·quick·brown·fox·jumps·over·the·lazy·dog.·</p>

--MXdestroyWholeWhitespace $wss,$content,$formatting

Remove any text nodes that only consist of whitespace characters, unless they are a child of the tags specified by this option.

Especially useful when using display: inline-block so that whitespace between elements (e.g. indentation) does not alter layout and styling.

Before	After
<div>↵ ··<h1></h1>↵ ··<ul></ul>↵ ··A·quick·<strong>brown</strong>·<em>fox</em>.↵ </div>	<div><h1></h1><ul></ul>↵ ··A·quick·<strong>brown</strong><em>fox</em>.↵ </div>

--MXtrimWhitespace $wss,$formatting

Remove any whitespace from the start and end of a tag, if the first and/or last node is a text node, unless the tag is one of the tags specified by this option.

Useful when combined with whitespace collapsing.

Other whitespace between text nodes and tags are not removed, as it is not recommended to mix non-formatting tags with raw text.

Basically, a tag should only either contain text and formatting tags, or only non-formatting tags.

Before	After
<p>↵ ··Hey,·I·<em>just</em>·found↵ ··out·about·this·<strong>cool</strong>·website!↵ ··<div></div>↵ </p>	<p>Hey,·I·<em>just</em>·found↵ ··out·about·this·<strong>cool</strong>·website!↵ ··<div></div></p>

--MXtrimClassAttribute

Don't trim and collapse whitespace in class attribute values.

Before	After
<div class=" hi lo a b c d e f g "></div>	<div class="hi lo a b c d e f g"></div>

--MXdecEnt

Don't decode any valid entities into their UTF-8 values.

--MXcondComments

Don't minify the contents of conditional comments, including downlevel-revealed conditional comments.

--MXattrQuotes

Don't remove quotes around attribute values when possible.

--MXcomments

Don't remove any comments, except conditional comments.

--MXoptTags

Don't remove optional starting or ending tags.

--MXtagWS

Don't remove spaces between attributes when possible.

Non-options

Explicitly important

The following removal of attributes and tags as minification strategies are not available in hyperbuild, as they should not have been declared in the first place.

If they exist, it is assumed there is a special reason for being so.

• Remove empty attributes
• Remove empty elements
• Remove redundant attributes
• Remove type attribute on <script> tags
• Remove type attribute on <style> and <link> tags