css-inline0.17.0
Published
High-performance library for inlining CSS into HTML 'style' attributes
pip install css-inline
Package Downloads
Authors
Project URLs
Requires Python
>=3.9
Dependencies
css_inline
css_inline is a high-performance library for inlining CSS into HTML 'style' attributes.
This library is designed for scenarios such as preparing HTML emails or embedding HTML into third-party web pages.
For instance, the library transforms HTML like this:
<html>
<head>
<style>h1 { color:blue; }</style>
</head>
<body>
<h1>Big Text</h1>
</body>
</html>
into:
<html>
<head></head>
<body>
<h1 style="color:blue;">Big Text</h1>
</body>
</html>
- Uses reliable components from Mozilla's Servo project
- 10-500x faster than alternatives
- Inlines CSS from
styleandlinktags - Removes
styleandlinktags - Resolves external stylesheets (including local files)
- Optionally caches external stylesheets
- Can process multiple documents in parallel
- Works on Linux, Windows, macOS and in the browser via PyOdide
- Supports HTML5 & CSS3
- Tested on CPython 3.9, 3.10, 3.11, 3.12, 3.13 and PyPy 3.9, 3.10.
Playground
If you'd like to try css-inline, you can check the WebAssembly-powered playground to see the results instantly.
Installation
Install with pip:
pip install css_inline
Pre-compiled wheels are available for most popular platforms. If not available for your platform, a Rust compiler will be needed to build this package from source. Rust version 1.65 or higher is required.
Usage
import css_inline
HTML = """<html>
<head>
<style>h1 { color:blue; }</style>
</head>
<body>
<h1>Big Text</h1>
</body>
</html>"""
inlined = css_inline.inline(HTML)
# HTML becomes this:
#
# <html>
# <head>
# <style>h1 { color:blue; }</style>
# </head>
# <body>
# <h1 style="color:blue;">Big Text</h1>
# </body>
# </html>
Note that css-inline automatically adds missing html and body tags, so the output is a valid HTML document.
Alternatively, you can inline CSS into an HTML fragment, preserving the original structure:
FRAGMENT = """<main>
<h1>Hello</h1>
<section>
<p>who am i</p>
</section>
</main>"""
CSS = """
p {
color: red;
}
h1 {
color: blue;
}
"""
inlined = css_inline.inline_fragment(FRAGMENT, CSS)
# HTML becomes this:
# <main>
# <h1 style="color: blue;">Hello</h1>
# <section>
# <p style="color: red;">who am i</p>
# </section>
# </main>
When there is a need to inline multiple HTML documents simultaneously, css_inline offers inline_many and inline_many_fragments functions.
This feature allows for concurrent processing of several inputs, significantly improving performance when dealing with a large number of documents.
import css_inline
css_inline.inline_many(["<...>", "<...>"])
Under the hood, inline_many, spawns threads at the Rust layer to handle the parallel processing of inputs.
This results in faster execution times compared to employing parallel processing techniques at the Python level.
Note: To fully benefit from inline_many, you should run your application on a multicore machine.
Configuration
For configuration options use the CSSInliner class:
import css_inline
inliner = css_inline.CSSInliner(keep_style_tags=True)
inliner.inline("...")
inline_style_tags. Specifies whether to inline CSS from "style" tags. Default:Truekeep_style_tags. Specifies whether to keep "style" tags after inlining. Default:Falsekeep_link_tags. Specifies whether to keep "link" tags after inlining. Default:Falsekeep_at_rules. Specifies whether to keep "at-rules" (starting with@) after inlining. Default:Falsebase_url. The base URL used to resolve relative URLs. If you'd like to load stylesheets from your filesystem, use thefile://scheme. Default:Noneload_remote_stylesheets. Specifies whether remote stylesheets should be loaded. Default:Truecache. Specifies caching options for external stylesheets (for example,StylesheetCache(size=5)). Default:Noneextra_css. Extra CSS to be inlined. Default:Nonepreallocate_node_capacity. Advanced. Preallocates capacity for HTML nodes during parsing. This can improve performance when you have an estimate of the number of nodes in your HTML document. Default:32
You can also skip CSS inlining for an HTML tag by adding the data-css-inline="ignore" attribute to it:
<head>
<style>h1 { color:blue; }</style>
</head>
<body>
<!-- The tag below won't receive additional styles -->
<h1 data-css-inline="ignore">Big Text</h1>
</body>
The data-css-inline="ignore" attribute also allows you to skip link and style tags:
<head>
<!-- Styles below are ignored -->
<style data-css-inline="ignore">h1 { color:blue; }</style>
</head>
<body>
<h1>Big Text</h1>
</body>
Alternatively, you may keep style from being removed by using the data-css-inline="keep" attribute.
This is useful if you want to keep @media queries for responsive emails in separate style tags.
Such tags will be kept in the resulting HTML even if the keep_style_tags option is set to false.
<head>
<!-- Styles below are not removed -->
<style data-css-inline="keep">h1 { color:blue; }</style>
</head>
<body>
<h1>Big Text</h1>
</body>
Another possibility is to set keep_at_rules option to true. At-rules cannot be inlined into HTML therefore they
get removed by default. This is useful if you want to keep at-rules, e.g. @media queries for responsive emails in
separate style tags but inline any styles which can be inlined.
Such tags will be kept in the resulting HTML even if the keep_style_tags option is explicitly set to false.
<head>
<!-- With keep_at_rules=true "color:blue" will get inlined into <h1> but @media will be kept in <style> -->
<style>h1 { color: blue; } @media (max-width: 600px) { h1 { font-size: 18px; } }</style>
</head>
<body>
<h1>Big Text</h1>
</body>
If you'd like to load stylesheets from your filesystem, use the file:// scheme:
import css_inline
# styles/email is relative to the current directory
inliner = css_inline.CSSInliner(base_url="file://styles/email/")
inliner.inline("...")
You can also cache external stylesheets to avoid excessive network requests:
import css_inline
inliner = css_inline.CSSInliner(
cache=css_inline.StylesheetCache(size=5)
)
inliner.inline("...")
Caching is disabled by default.
XHTML compatibility
If you'd like to work around some XHTML compatibility issues like closing empty tags (<hr> vs. <hr/>), you can use the following snippet that involves lxml:
import css_inline
from lxml import html, etree
document = "..." # Your HTML document
inlined = css_inline.inline(document)
tree = html.fromstring(inlined)
inlined = etree.tostring(tree).decode(encoding="utf-8")
Performance
css-inline is powered by efficient tooling from Mozilla's Servo project and significantly outperforms other Python alternatives in terms of speed.
Most of the time it achieves over a 10x speed advantage compared to the next fastest alternative.
Here is the performance comparison:
| Size | css_inline 0.17.0 | premailer 3.10.0 | toronado 0.1.0 | inlinestyler 0.2.5 | pynliner 0.8.0 | |
|---|---|---|---|---|---|---|
| Basic | 230 B | 4.27 µs | 107.68 µs (25.17x) | 603.05 µs (140.97x) | 1.02 ms (238.87x) | 1.13ms (264.25x) |
| Realistic-1 | 8.58 KB | 85.54 µs | 1.20 ms (14.13x) | 13.79 ms (161.32x) | 26.37 ms (308.27x) | 44.94 ms (525.39x) |
| Realistic-2 | 4.3 KB | 58.26 µs | 1.74 ms (29.88x) | ERROR | 17.71 ms (303.98x) | ERROR |
| GitHub page | 1.81 MB | 139.74 ms | 15.02 s (107.49x) | ERROR | ERROR | ERROR |
The "Basic" case was obtained by benchmarking the example from the Usage section.
Note that the toronado, inlinestyler, and pynliner libraries both encountered errors when used to inline CSS in the last scenario.
The benchmarking code is available in the benches/bench.py file. The benchmarks were conducted using the stable rustc 1.87, Python 3.12.5 on Ryzen 9 9950X.
Comparison with other libraries
Besides performance, css-inline differs from other Python libraries for CSS inlining.
- Generally supports more CSS features than other libraries (for example,
toronadoandpynlinerdo not support pseudo-elements); - It has fewer configuration options and is not as flexible as
premailer; - Works on fewer platforms than LXML-based libraries (
premailer,inlinestyler,toronado, and optionallypynliner); - Does not have debug logs yet;
- Supports only HTML 5.
Further reading
If you want to know how this library was created & how it works internally, you could take a look at these articles:
License
This project is licensed under the terms of the MIT license.