Unreleased

4.0.0

Update html5ever to 0.27
feature: change lifetimes on UrlRelative (breaking change)
chore: use the Display trait (and ToString) instead of an inherent to_string method (breaking change)
Security fixes are not backported to the 2.0 branch any more. Only the v3 and v4 branches are supported.

3.3.0

chore: bump MSRV to 1.60.0
docs: fix incorrect XSS example
feature: add UrlRelative::RewriteWithRoot

3.2.1

chore: bump MSRV to 1.51.0 serde-rs/serde#2255
chore: fix broken links in documentation

3.2.0

chore: bump MSRV to 1.49.0
chore: update to html5ever 0.26
chore: switch from lazy_static to once_cell
feature: add Builder::empty() constructor

3.1.4

chore: use #[non_exhaustive] instead of hidden variant for UrlRelative policy
chore: remove unnecessary dependency on third-party matches! macro

3.1.3

fix: incorrect FF/CR handling in clean_text
fix: split class name attribute by all ASCII whitespace, not just SP 0x20
docs: fix incorrect English in is_html documentation

3.1.2

fix: unexpected namespace switches can allow XSS via svg/mathml parsing

3.1.1

fix: Crash on invalid URLs in some configurations (issue #136)

3.1.0

feature: Whitelist generic attribute prefix
chore: Update html5ever to 0.25
feature: support adding attribute values to elements

3.0.0

Add clean_text function.
Updated to rust-url 2.0.
Updated to the 2018 edition.

2.1.4

fix: split class name attribute by all ASCII whitespace, not just SP 0x20 (backported from 3.1.3)

2.1.3

fix: unexpected namespace switches can allow XSS via svg/mathml parsing (backported from 3.1.2)

2.1.2

Fix a memory leak caused by certain node types.

2.1.1

Update dependencies

2.1.0

Bump minimum supported Rust version to 1.30.
Fix a potential DoS attack from pathologically nested input.

2.0.0

Add "script" and "style" to the default set of clean content tags
Make all iterator-accepting APIs use IntoIterator and Borrow, so that you can pass slices directly to them.

1.2.0

Recognize action, formaction and ping as URL attributes for scheme and origin filtering
Add Builder::url_filter_map which allows URLs, both relative and absolute, to be pre-filtered

1.1.0

Add Builder::clean_content_tags which allows elements to be removed entirely instead of just having the tags removed

1.0.1

Update dependencies

1.0.0

Breaking change: The Ammonia struct is now called Builder and uses that pattern for better forward compatibility
Breaking change: The Builder::clean() method now returns a Document struct instead of a String. You can use the Document::to_string method to obtain a String.
Breaking change: keep_cleaned_elements has changed from being an off-by-default option to the only supported behavior
Breaking change: Using a tag with allowed_classes means that the class attribute is banned from tag_attributes (it used to be required)
Breaking change: The default set of allowed elements and attributes was expanded
Added support for reading the input from a stream
Added UrlRelative::Custom, allowing you to write your own relative URL resolver
Changed UrlRelative::RewriteWithBase take a custom URL. This made the url crate a public dependency.
Added id_prefix, which can be used to avoid element id collisions with the rest of the page
Added property getters to Builder, to see what everything is currently set to
Added property modifiers, to change the existing whitelist (instead of completely replacing it)

0.7.0

Add allowed_classes, allowing the user to set only specific items that can go in the class attribute

0.6.1

Fix a bug in the traversal code

0.6.0

Resolve relative URLs with a given base (off by default, you need to specify that base URL)
Add rel="noreferrer noopener" to links, as a security measure
Avoid closing void tags, such as turning <br> into <br></br>
Bump the html5ever version
Switch to using docs.rs to host docs

0.5.0

Bump html5ever to 0.18 (this updates serde from 0.9 to 1.0)

0.4.0

Upgrade to html5ever 0.17

0.3.0

Add an option to keep elements that had attributes removed

0.2.0

Removed the strip option. Not a security problem, but it was wrong and looked stupid. I'm not going to reintroduce this until html5ever allows me to preserve the original text enough to have non-stripped tags come out exactly like they go in.
Treat the data attribute of object as a URL. In non-default configurations, this could have been a leak.
Update to the newest html5ever.