michael/minify-html
michael
/
minify-html
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Use minify-js as JS minifier

This commit is contained in:
Wilson Lin 2022-06-21 17:29:12 +10:00
parent fd60983516
commit c7d0652fbc
18 changed files with 82 additions and 171 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
README.md
View File

@ -13,7 +13,7 @@ A Rust HTML minifier meticulously optimised for speed and effectiveness, with bi
- Advanced minification strategy beats other minifiers while being much faster.
- Uses SIMD searching, direct tries, and lookup tables.
- Handles [invalid HTML](./notes/Parsing.md), with extensive testing and [fuzzing](./fuzz).
- Natively binds to [esbuild](https://github.com/wilsonzlin/esbuild-rs) for super fast JS and CSS minification.
- Uses [minify-js](https://github.com/wilsonzlin/minify-js) for super fast JS minification.
## Performance
@ -55,13 +55,9 @@ minify-html --output /path/to/output.min.html --keep-closing-tags --minify-css /
```toml
[dependencies]
minify-html = { version = "0.8.1", features = ["js-esbuild"] }
minify-html = { version = "0.8.1" }
```
Building with the `js-esbuild` feature requires the Go compiler to be installed as well, to build the [JS and CSS minifier](https://github.com/wilsonzlin/esbuild-rs).
If the `js-esbuild` feature is not enabled, `cfg.minify_js` and `cfg.minify_css` will have no effect.
### Use
Check out the [docs](https://docs.rs/minify-html) for API and usage examples.
@ -190,7 +186,11 @@ All [`Cfg` fields](https://docs.rs/minify-html/latest/minify_html/struct.Cfg.htm
## Minification
Note that some of the minification done can result in HTML that will not pass validation, but remain interpreted and rendered correctly by the browser; essentially, the laxness of the browser is taken advantage of for better minification. These can be turned off via the `Cfg` object.
Note that some of the minification done can result in HTML that will not pass validation, but remain interpreted and rendered correctly by the browser; essentially, the laxness of the browser is taken advantage of for better minification. To prevent this, refer to these configuration options:
- `do_not_minify_doctype`
- `ensure_spec_compliant_unquoted_attribute_values`
- `keep_spaces_between_attributes`
### Whitespace

2
bench/runners/minify-html/Cargo.toml
View File

@ -6,7 +6,7 @@ authors = ["Wilson Lin <code@wilsonl.in>"]
edition = "2018"
[dependencies]
minify-html = { path = "../../../rust/main", features = ["js-esbuild"] }
minify-html = { path = "../../../rust/main" }
serde = { version = "1.0.104", features = ["derive"] }
serde_json = "1.0.44"

2
cli/Cargo.toml
View File

@ -7,5 +7,5 @@ authors = ["Wilson Lin <code@wilsonl.in>"]
edition = "2018"
[dependencies]
minify-html = { path = "../rust/main", features = ["js-esbuild"] }
minify-html = { path = "../rust/main" }
structopt = "0.3"

2
java/Cargo.toml
View File

@ -6,7 +6,7 @@ authors = ["Wilson Lin <code@wilsonl.in>"]
edition = "2018"
[dependencies]
minify-html = { path = "../rust/main", features = ["js-esbuild"] }
minify-html = { path = "../rust/main" }
jni = "0.14.0"
[lib]

4
nodejs/index.d.ts vendored
View File

@ -21,11 +21,11 @@ export function createConfiguration (options: {
/** Keep all comments. */
keep_comments?: boolean;
/**
* If enabled, content in `<script>` tags with a JS or no [MIME type](https://mimesniff.spec.whatwg.org/#javascript-mime-type) will be minified using [esbuild-rs](https://github.com/wilsonzlin/esbuild-rs).
* If enabled, content in `<script>` tags with a JS or no [MIME type](https://mimesniff.spec.whatwg.org/#javascript-mime-type) will be minified using [minify-js](https://github.com/wilsonzlin/minify-js).
*/
minify_js?: boolean;
/**
* If enabled, CSS in `<style>` tags will be minified using [esbuild-rs](https://github.com/wilsonzlin/esbuild-rs).
* If enabled, CSS in `<style>` tags and `style` attributes will be minified.
*/
minify_css?: boolean;
/** Remove all bangs. */

6
nodejs/native/Cargo.toml
View File

@ -10,13 +10,9 @@ edition = "2018"
name = "minify_html_ffi"
crate-type = ["staticlib"]
[features]
core = ["minify-html"]
js = ["minify-html/js-esbuild"]
[build-dependencies]
cbindgen = "0.14"
[dependencies]
libc = "0.2"
minify-html = { path = "../../rust/main", optional = true }
minify-html = { path = "../../rust/main" }

2
python/Cargo.js.toml
View File

@ -15,7 +15,7 @@ name = "minify_html"
crate-type = ["cdylib"]
[dependencies]
minify-html = { path = "../rust/main", features = ["js-esbuild"] }
minify-html = { path = "../rust/main" }
[dependencies.pyo3]
version = "0.13.0"
features = ["extension-module"]

2
ruby/Cargo.toml
View File

@ -10,5 +10,5 @@ name = "minify_html_ruby_lib"
crate-type = ["cdylib"]
[dependencies]
minify-html = { path = "../rust/main", features = ["js-esbuild"] }
minify-html = { path = "../rust/main" }
rutie = "0.7.0"

40
rust/common/tests/mod.rs
View File

@ -1,6 +1,4 @@
use crate::tests::eval;
#[cfg(feature = "js-esbuild")]
use crate::tests::{eval_with_css_min, eval_with_js_min};
#[test]
@ -460,25 +458,19 @@ fn test_processing_instructions() {
eval(b"av<?xml 1.0 ?>g", b"av<?xml 1.0 ?>g");
}
#[cfg(feature = "js-esbuild")]
#[test]
fn test_js_minification() {
eval_with_js_min(b"<script>let a = 1;</script>", b"<script>let a=1;</script>");
eval_with_js_min(b"<script>let a = 1;</script>", b"<script>let a=1</script>");
eval_with_js_min(
b"<script type=text/javascript>let a = 1;</script>",
b"<script>let a=1;</script>",
);
// `export` statements are not allowed inline.
eval_with_js_min(
b"<script type=module>let a = 1; export a;</script>",
b"<script type=module></script>",
b"<script>let a=1</script>",
);
eval_with_js_min(
br#"
<script>let a = 1;</script>
<script>let b = 2;</script>
"#,
b"<script>let a=1;</script><script>let b=2;</script>",
b"<script>let a=1</script><script>let b=2</script>",
);
eval_with_js_min(
b"<scRIPt type=text/plain> alert(1.00000); </scripT>",
@ -491,39 +483,37 @@ fn test_js_minification() {
let a = 1;
</script>
"#,
b"<script>let a=1;</script>",
b"<script>let a=1</script>",
);
}
#[cfg(feature = "js-esbuild")]
/* TODO Reenable once unintentional script closing tag escaping is implemented in minify-js.
#[test]
fn test_js_minification_unintentional_closing_tag() {
eval_with_js_min(
br#"<script>let a = "</" + "script>";</script>"#,
br#"<script>let a="<\/script>";</script>"#,
);
// TODO Reenable once esbuild handles closing tags case insensitively (evanw/esbuild#1509).
// eval_with_js_min(
// br#"<script>let a = "</S" + "cRiPT>";</script>"#,
// br#"<script>let a="<\/ScRiPT>";</script>"#,
// );
eval_with_js_min(
br#"<script>let a = "</S" + "cRiPT>";</script>"#,
br#"<script>let a="<\/ScRiPT>";</script>"#,
);
eval_with_js_min(
br#"<script>let a = "\u003c/script>";</script>"#,
br#"<script>let a="<\/script>";</script>"#,
);
// TODO Reenable once esbuild handles closing tags case insensitively (evanw/esbuild#1509).
// eval_with_js_min(
// br#"<script>let a = "\u003c/scrIPt>";</script>"#,
// br#"<script>let a="<\/scrIPt>";</script>"#,
// );
eval_with_js_min(
br#"<script>let a = "\u003c/scrIPt>";</script>"#,
br#"<script>let a="<\/scrIPt>";</script>"#,
);
}
*/
#[cfg(feature = "js-esbuild")]
#[test]
fn test_style_element_minification() {
// `<style>` contents.
eval_with_css_min(
b"<style>div { color: yellow }</style>",
b"<style>div{color:#ff0}</style>",
b"<style>div{color:yellow}</style>",
);
}

8
rust/main/Cargo.toml
View File

@ -15,13 +15,9 @@ include = ["/src/**/*", "/Cargo.toml", "/LICENSE", "/README.md"]
[badges]
maintenance = { status = "actively-developed" }
[features]
default = []
js-esbuild = ["crossbeam", "esbuild-rs"]
[dependencies]
aho-corasick = "0.7"
crossbeam = { version = "0.7", optional = true }
esbuild-rs = { version = "0.13.8", optional = true }
css-minify = "0.2.2"
minify-js = "0.1.0"
lazy_static = "1.4"
memchr = "2"

7
rust/main/src/cfg/mod.rs
View File

@ -14,13 +14,10 @@ pub struct Cfg {
pub keep_spaces_between_attributes: bool,
/// Keep all comments.
pub keep_comments: bool,
/// If enabled, CSS in `<style>` tags are minified using
/// [esbuild-rs](https://github.com/wilsonzlin/esbuild-rs). The `js-esbuild` feature must be
/// enabled; otherwise, this value has no effect.
/// If enabled, CSS in `<style>` tags and `style` attributes are minified.
pub minify_css: bool,
/// If enabled, JavaScript in `<script>` tags are minified using
/// [esbuild-rs](https://github.com/wilsonzlin/esbuild-rs). The `js-esbuild` feature must be
/// enabled; otherwise, this value has no effect.
/// [minify-js](https://github.com/wilsonzlin/minify-js).
///
/// Only `<script>` tags with a valid or no
/// [MIME type](https://mimesniff.spec.whatwg.org/#javascript-mime-type) is considered to

44
rust/main/src/minify/attr.rs
View File

@ -1,10 +1,8 @@
use aho_corasick::{AhoCorasickBuilder, MatchKind};
use lazy_static::lazy_static;
use std::str::from_utf8_unchecked;
#[cfg(feature = "js-esbuild")]
use {
crate::minify::css::MINIFY_CSS_TRANSFORM_OPTIONS, crate::minify::esbuild::minify_using_esbuild,
};
use aho_corasick::{AhoCorasickBuilder, MatchKind};
use css_minify::optimizations::{Level, Minifier};
use lazy_static::lazy_static;
use crate::common::gen::attrs::ATTRS;
use crate::common::gen::codepoints::DIGIT;
@ -312,27 +310,25 @@ pub fn minify_attr(
};
};
#[cfg(feature = "js-esbuild")]
if name == b"style" && cfg.minify_css {
let mut value_raw_wrapped = Vec::with_capacity(value_raw.len() + 3);
let mut value_raw_wrapped = String::with_capacity(value_raw.len() + 3);
// TODO This isn't safe for invalid input e.g. `a}/*`.
value_raw_wrapped.extend_from_slice(b"x{");
value_raw_wrapped.extend_from_slice(&value_raw);
value_raw_wrapped.push(b'}');
let mut value_raw_wrapped_min = Vec::with_capacity(value_raw_wrapped.len());
minify_using_esbuild(
&mut value_raw_wrapped_min,
&value_raw_wrapped,
&MINIFY_CSS_TRANSFORM_OPTIONS.clone(),
);
// TODO If input was invalid, wrapper syntax may not exist anymore.
if value_raw_wrapped_min.starts_with(b"x{") {
value_raw_wrapped_min.drain(0..2);
value_raw_wrapped.push_str("x{");
value_raw_wrapped.push_str(unsafe { from_utf8_unchecked(&value_raw) });
value_raw_wrapped.push('}');
let result = Minifier::default().minify(&value_raw_wrapped, Level::Three);
// TODO Collect error as warning.
if let Ok(min) = result {
let mut value_raw_wrapped_min = min.into_bytes();
// TODO If input was invalid, wrapper syntax may not exist anymore.
if value_raw_wrapped_min.starts_with(b"x{") {
value_raw_wrapped_min.drain(0..2);
};
if value_raw_wrapped_min.ends_with(b"}") {
value_raw_wrapped_min.pop();
};
value_raw = value_raw_wrapped_min;
};
if value_raw_wrapped_min.ends_with(b"}") {
value_raw_wrapped_min.pop();
};
value_raw = value_raw_wrapped_min;
}
// Make lowercase before checking against default value or JAVASCRIPT_MIME_TYPES.

48
rust/main/src/minify/css.rs
View File

@ -1,41 +1,19 @@
#[cfg(feature = "js-esbuild")]
use {
crate::minify::esbuild::minify_using_esbuild,
esbuild_rs::{
Charset, LegalComments, Loader, SourceMap, TransformOptions, TransformOptionsBuilder,
},
lazy_static::lazy_static,
std::sync::Arc,
};
use std::str::from_utf8_unchecked;
use crate::cfg::Cfg;
use crate::common::whitespace::trimmed;
use css_minify::optimizations::{Level, Minifier};
#[cfg(feature = "js-esbuild")]
lazy_static! {
pub static ref MINIFY_CSS_TRANSFORM_OPTIONS: Arc<TransformOptions> = {
let mut builder = TransformOptionsBuilder::new();
builder.charset = Charset::UTF8;
builder.legal_comments = LegalComments::None;
builder.loader = Loader::CSS;
builder.minify_identifiers = true;
builder.minify_syntax = true;
builder.minify_whitespace = true;
builder.source_map = SourceMap::None;
builder.build()
};
}
#[cfg(not(feature = "js-esbuild"))]
pub fn minify_css(_cfg: &Cfg, out: &mut Vec<u8>, code: &[u8]) {
pub fn minify_css(cfg: &Cfg, out: &mut Vec<u8>, code: &[u8]) {
if cfg.minify_css {
let result = Minifier::default().minify(unsafe { from_utf8_unchecked(code) }, Level::Three);
// TODO Collect error as warning.
if let Ok(min) = result {
if min.len() < code.len() {
out.extend_from_slice(min.as_bytes());
return;
};
};
}
out.extend_from_slice(trimmed(code));
}
#[cfg(feature = "js-esbuild")]
pub fn minify_css(cfg: &Cfg, out: &mut Vec<u8>, code: &[u8]) {
if !cfg.minify_css {
out.extend_from_slice(trimmed(code));
} else {
minify_using_esbuild(out, code, &MINIFY_CSS_TRANSFORM_OPTIONS.clone());
}
}

18
rust/main/src/minify/esbuild.rs
View File

@ -1,18 +0,0 @@
#[cfg(feature = "js-esbuild")]
use {crossbeam::sync::WaitGroup, esbuild_rs::TransformOptions};
#[cfg(feature = "js-esbuild")]
// TODO The use of WG is ugly and we don't want to be multi-threaded; wait for Rust port esbuild-transform-rs.
pub fn minify_using_esbuild(out: &mut Vec<u8>, code: &[u8], transform_options: &TransformOptions) {
let wg = WaitGroup::new();
unsafe {
let wg = wg.clone();
// esbuild now officially handles escaping `</script` and `</style`.
esbuild_rs::transform_direct_unmanaged(code, transform_options, move |result| {
let min_code = result.code.as_str().trim().as_bytes();
out.extend_from_slice(min_code);
drop(wg);
});
};
wg.wait();
}

46
rust/main/src/minify/js.rs
View File

@ -1,38 +1,18 @@
#[cfg(feature = "js-esbuild")]
use {
crate::minify::esbuild::minify_using_esbuild,
esbuild_rs::{Charset, LegalComments, SourceMap, TransformOptions, TransformOptionsBuilder},
lazy_static::lazy_static,
std::sync::Arc,
};
use crate::cfg::Cfg;
use crate::common::whitespace::trimmed;
use minify_js::minify as minifier;
#[cfg(feature = "js-esbuild")]
lazy_static! {
static ref TRANSFORM_OPTIONS: Arc<TransformOptions> = {
let mut builder = TransformOptionsBuilder::new();
builder.charset = Charset::UTF8;
builder.legal_comments = LegalComments::None;
builder.minify_identifiers = true;
builder.minify_syntax = true;
builder.minify_whitespace = true;
builder.source_map = SourceMap::None;
builder.build()
};
}
#[cfg(not(feature = "js-esbuild"))]
pub fn minify_js(_cfg: &Cfg, out: &mut Vec<u8>, code: &[u8]) {
pub fn minify_js(cfg: &Cfg, out: &mut Vec<u8>, code: &[u8]) {
if cfg.minify_js {
let source = code.to_vec();
// TODO Write to `out` directly, but only if we can guarantee that the length will never exceed the input.
let mut output = Vec::new();
let result = minifier(source, &mut output);
// TODO Collect error as warning.
if !result.is_err() && output.len() < code.len() {
out.extend_from_slice(output.as_slice());
return;
};
}
out.extend_from_slice(trimmed(code));
}
#[cfg(feature = "js-esbuild")]
pub fn minify_js(cfg: &Cfg, out: &mut Vec<u8>, code: &[u8]) {
if !cfg.minify_js {
out.extend_from_slice(trimmed(code));
} else {
minify_using_esbuild(out, code, &TRANSFORM_OPTIONS.clone());
}
}

1
rust/main/src/minify/mod.rs
View File

@ -5,7 +5,6 @@ pub mod content;
pub mod css;
pub mod doctype;
pub mod element;
pub mod esbuild;
pub mod instruction;
pub mod js;
#[cfg(test)]

5
rust/main/src/tests/mod.rs
View File

@ -8,14 +8,12 @@ pub fn eval_with_cfg(src: &'static [u8], expected: &'static [u8], cfg: &Cfg) {
assert_eq!(from_utf8(&min).unwrap(), from_utf8(expected).unwrap(),);
}
#[cfg(feature = "js-esbuild")]
pub fn eval_with_js_min(src: &'static [u8], expected: &'static [u8]) -> () {
let mut cfg = Cfg::new();
cfg.minify_js = true;
eval_with_cfg(src, expected, &cfg);
}
#[cfg(feature = "js-esbuild")]
pub fn eval_with_css_min(src: &'static [u8], expected: &'static [u8]) -> () {
let mut cfg = Cfg::new();
cfg.minify_css = true;
@ -146,12 +144,11 @@ fn test_viewport_attr_minification() {
);
}
#[cfg(feature = "js-esbuild")]
#[test]
fn test_style_attr_minification() {
eval_with_css_min(
br#"<div style="color: yellow;"></div>"#,
br#"<div style=color:#ff0></div>"#,
br#"<div style=color:yellow></div>"#,
);
// `style` attributes are removed if fully minified away.
eval_with_css_min(br#"<div style=" /* */ "></div>"#, br#"<div></div>"#);

2
version
View File

@ -92,7 +92,7 @@ if (
}
cmd("git", "pull");
cmd("bash", "./prebuild.sh");
cmd("cargo", "test", "--features", "js-esbuild", { workingDir: RUST_MAIN_DIR });
cmd("cargo", "test", { workingDir: RUST_MAIN_DIR });
cmd("cargo", "test", "--features", "js-esbuild", {
workingDir: RUST_ONEPASS_DIR,
});