chore: Replaced the kuchiki crate with our custom-built HTML tree representation (#186)

Author: Stranger6667

CHANGELOG.md

@@ -2,6 +2,14 @@
 
 ## [Unreleased]
 
+### Internal
+
+- Replaced the `kuchiki` crate with our custom-built HTML tree representation. [#176](https://github.com/Stranger6667/css-inline/issues/176)
+
+### Performance
+
+- 15-30% average performance improvement due switch from `kuchiki` to a custom-built HTML tree representation.
+
 ## [0.8.5] - 2022-11-10
 
 ### Added

README.md

@@ -116,7 +116,7 @@ This attribute also allows you to skip `link` and `style` tags:
 
 ## Standards support & restrictions
 
-`css-inline` is built on top of [kuchiki](https://crates.io/crates/kuchiki) and [cssparser](https://crates.io/crates/cssparser) and relies on their behavior for HTML / CSS parsing and serialization.
+`css-inline` is built on top of [cssparser](https://crates.io/crates/cssparser) and relies on its behavior for CSS parsing.
 Notably:
 
 - Only HTML 5, XHTML is not supported;

bindings/python/CHANGELOG.md

@@ -2,6 +2,10 @@
 
 ## [Unreleased]
 
+### Performance
+
+- 15-30% average performance improvement due switch from `kuchiki` to a custom-built HTML tree representation.
+
 ## [0.8.7] - 2023-01-30
 
 ### Added

bindings/python/README.md

@@ -30,7 +30,7 @@ a Rust compiler to build this package from source. The minimum supported Rust ve
 
 ## Usage
 
-To inline CSS in a HTML document:
+To inline CSS in an HTML document:
 
 ```python
 import css_inline
@@ -118,7 +118,7 @@ inliner.inline("...")
 
 ## Standards support & restrictions
 
-`css-inline` is built on top of [kuchiki](https://crates.io/crates/kuchiki) and [cssparser](https://crates.io/crates/cssparser) and relies on their behavior for HTML / CSS parsing and serialization.
+`css-inline` is built on top of [cssparser](https://crates.io/crates/cssparser) and relies on its behavior for CSS parsing.
 Notably:
 
 - Only HTML 5, XHTML is not supported;

bindings/python/src/lib.rs

@@ -67,12 +67,12 @@ fn parse_url(url: Option<String>) -> PyResult<Option<url::Url>> {
     })
 }
 
-/// CSSInliner(inline_style_tags=True, remove_style_tags=False, base_url=None, load_remote_stylesheets=True, extra_css=None)
+/// CSSInliner(inline_style_tags=True, remove_style_tags=False, base_url=None, load_remote_stylesheets=True, extra_css=None, preallocate_node_capacity=0)
 ///
 /// Customizable CSS inliner.
 #[pyclass]
 #[pyo3(
-    text_signature = "(inline_style_tags=True, remove_style_tags=False, base_url=None, load_remote_stylesheets=True, extra_css=None)"
+    text_signature = "(inline_style_tags=True, remove_style_tags=False, base_url=None, load_remote_stylesheets=True, extra_css=None, preallocate_node_capacity=0)"
 )]
 struct CSSInliner {
     inner: rust_inline::CSSInliner<'static>,
@@ -87,13 +87,16 @@ impl CSSInliner {
         base_url: Option<String>,
         load_remote_stylesheets: Option<bool>,
         extra_css: Option<String>,
+        preallocate_node_capacity: Option<usize>,
     ) -> PyResult<Self> {
         let options = rust_inline::InlineOptions {
             inline_style_tags: inline_style_tags.unwrap_or(true),
             remove_style_tags: remove_style_tags.unwrap_or(false),
             base_url: parse_url(base_url)?,
             load_remote_stylesheets: load_remote_stylesheets.unwrap_or(true),
             extra_css: extra_css.map(Cow::Owned),
+            preallocate_node_capacity: preallocate_node_capacity
+                .unwrap_or(rust_inline::DEFAULT_HTML_TREE_CAPACITY),
         };
         Ok(CSSInliner {
             inner: rust_inline::CSSInliner::new(options),
@@ -131,13 +134,16 @@ fn inline(
     base_url: Option<String>,
     load_remote_stylesheets: Option<bool>,
     extra_css: Option<&str>,
+    preallocate_node_capacity: Option<usize>,
 ) -> PyResult<String> {
     let options = rust_inline::InlineOptions {
         inline_style_tags: inline_style_tags.unwrap_or(true),
         remove_style_tags: remove_style_tags.unwrap_or(false),
         base_url: parse_url(base_url)?,
         load_remote_stylesheets: load_remote_stylesheets.unwrap_or(true),
         extra_css: extra_css.map(Cow::Borrowed),
+        preallocate_node_capacity: preallocate_node_capacity
+            .unwrap_or(rust_inline::DEFAULT_HTML_TREE_CAPACITY),
     };
     let inliner = rust_inline::CSSInliner::new(options);
     Ok(inliner.inline(html).map_err(InlineErrorWrapper)?)
@@ -157,13 +163,16 @@ fn inline_many(
     base_url: Option<String>,
     load_remote_stylesheets: Option<bool>,
     extra_css: Option<&str>,
+    preallocate_node_capacity: Option<usize>,
 ) -> PyResult<Vec<String>> {
     let options = rust_inline::InlineOptions {
         inline_style_tags: inline_style_tags.unwrap_or(true),
         remove_style_tags: remove_style_tags.unwrap_or(false),
         base_url: parse_url(base_url)?,
         load_remote_stylesheets: load_remote_stylesheets.unwrap_or(true),
         extra_css: extra_css.map(Cow::Borrowed),
+        preallocate_node_capacity: preallocate_node_capacity
+            .unwrap_or(rust_inline::DEFAULT_HTML_TREE_CAPACITY),
     };
     let inliner = rust_inline::CSSInliner::new(options);
     inline_many_impl(&inliner, htmls)

bindings/wasm/CHANGELOG.md

@@ -2,6 +2,10 @@
 
 ## [Unreleased]
 
+### Performance
+
+- 15-30% average performance improvement due switch from `kuchiki` to a custom-built HTML tree representation.
+
 ## [0.8.4] - 2022-11-10
 
 ### Added

bindings/wasm/README.md

@@ -74,7 +74,7 @@ This attribute also allows you to skip `link` and `style` tags:
 
 ## Standards support & restrictions
 
-`css-inline` is built on top of [kuchiki](https://crates.io/crates/kuchiki) and [cssparser](https://crates.io/crates/cssparser) and relies on their behavior for HTML / CSS parsing and serialization.
+`css-inline` is built on top of [cssparser](https://crates.io/crates/cssparser) and relies on its behavior for CSS parsing.
 Notably:
 
 - Only HTML 5, XHTML is not supported;

bindings/wasm/src/lib.rs

@@ -67,6 +67,7 @@ struct Options {
     base_url: Option<String>,
     load_remote_stylesheets: bool,
     extra_css: Option<String>,
+    preallocate_node_capacity: Option<usize>,
 }
 
 impl Default for Options {
@@ -77,6 +78,7 @@ impl Default for Options {
             base_url: None,
             load_remote_stylesheets: true,
             extra_css: None,
+            preallocate_node_capacity: None,
         }
     }
 }
@@ -99,6 +101,9 @@ impl TryFrom<Options> for rust_inline::InlineOptions<'_> {
             base_url: parse_url(value.base_url)?,
             load_remote_stylesheets: value.load_remote_stylesheets,
             extra_css: value.extra_css.map(Cow::Owned),
+            preallocate_node_capacity: value
+                .preallocate_node_capacity
+                .unwrap_or(rust_inline::DEFAULT_HTML_TREE_CAPACITY),
         })
     }
 }
@@ -123,6 +128,7 @@ interface InlineOptions {
     base_url?: string,
     load_remote_stylesheets?: boolean,
     extra_css?: string,
+    preallocate_node_capacity?: number,
 }
 
 export function inline(html: string, options?: InlineOptions): string;

css-inline/Cargo.toml

@@ -23,18 +23,19 @@ file = []
 
 [dependencies]
 cssparser = "0.29"
-kuchiki = "0.8"
+html5ever = "0.26.0"
 memchr = "2.4"
 attohttpc = { version = "0", default-features = false, features = ["compress", "tls-rustls"], optional = true }
 url = "2"
 smallvec = "1"
 indexmap = "1.8.1"
 pico-args = { version = "0.3", optional = true }
 rayon = { version = "1.5", optional = true }
+selectors = "0.24.0"
 
 [dev-dependencies]
 assert_cmd = "2.0.6"
-criterion = ">= 0.1"
+criterion = { version = "0.5.1", features = [], default-features = false }
 
 [[bench]]
 name = "inliner"

css-inline/rustfmt.toml

@@ -0,0 +1,2 @@
+imports_granularity = "Crate"
+edition = "2021"