Change Log

Unreleased

2.2.1

Correctly parse some previously-failing mis-nested combinations of a and formatting elements. (Fixes #95.)

2.2

When no character set is specified or detected, fall back to Windows-1252 with replacement characters on iOS 8+ and Mac OS X 10.10+.
- Windows-1252 does not specify a character for every position. Previously, falling back to Windows-1252 would fail upon encountering any characters at unused positions, at which point HTMLReader would fall further back to ISO 8859-1 (which has no unused positions). As of iOS 8 and Mac OS X 10.10, NSString exposes a lossy decoding API that allows HTMLReader to fall back to Windows-1252 and replace unused positions with U+FFFD REPLACEMENT CHARACTER.

2.1.8

Change HTMLReaderTests/html5lib submodule to use https.
- To update your local checkout, after pulling at least this version of the HTMLReader repository, run git submodule sync.

2.1.7

Publicize functions HTMLStringEncodingForLabel() and HTMLInvalidStringEncoding().
- HTMLStringEncodingForLabel() is useful in any context where HTML-style encoding names can occur, such as a form element's accept-charset attribute. When looking up an encoding name fails, HTMLInvalidStringEncoding() is returned.

2.1.6

Fix "header missing from umbrella header" build error when using Swift Package Manager.

2.1.5

Fix general sibling combinator (~) in selectors.
- This would only work when both sides were the same, e.g. span ~ span, which is not at all how the combinator is specified.

2.1.4

Fix -Wshorten-64-to-32 warnings new to Xcode 11 beta.

2.1.3

Fix crash when given data that starts with a UTF-8 BOM but is not valid UTF-8.

2.1.2

Avoid symbol collision for parseNumber with Tapjoy SDK by marking it and other helper functions in HTMLSelector.m as static. (Fixes #85.)
Replace subscripting to quiet -Wunguarded-availability warnings (despite ARClite making the method available at the deployment target).

2.1.1

Update project for and fix build failures new to Xcode 9.3.

2.1

Add a -[HTMLDocument parsedStringEncoding] property.
- Some parts of the HTML spec make use of a document's string encoding, such as when submitting a form.
Import updated character encoding labels from the WHATWG.

2.0.6

Update project settings for Xcode 9 GM and fix warnings.

2.0.5

Fix bump-version script to save CFBundleVersion as a string (not a number) in Info.plist. (Fixes #74.)

2.0.4

Fix a compiler warning new to Xcode 8.3.

2.0.3

Move tests folder to support Swift 3's Package Manager.

2.0.2

Fix retain cycle on documents created with -[HTMLDocument initWithData:contentTypeHeader:].

2.0.1

Fix buffer overflow when parsing named entities.
- This would happen when attempting to parse the first semicolonless named entity AElig.

2.0

Fix HTMLElement's subscripting abilities not getting bridged into Swift (issue #59 revisited).
- This is a breaking change because Swift code used to see HTMLElement.objectForKeyedSubscript(_:) and now sees HTMLElement.subscript(_:).
Update project and playground for Xcode 8 and Swift 3.

1.0.1

Pass updated html5lib-tests.
Update return type of -[HTMLNode textComponents] to an array of NSString.
Add a nonempty Package.swift as now required by Swift Package Manager.

1.0 – 2016-07-02

Rearrange source folder tree to match Swift Package Manager convention.
Update html5lib-tests submodule to fix cloning.

0.9.6 – 2016-04-02

Fix Objective-C generics (and their import into Swift) by spelling things correctly. (Fixes #59.) (Fixes #60.)
Revert back to quoted #import to fix installation by copying files over.

0.9.5 – 2016-03-15

Fix incorrect parsing of selector groups when a selector included a pseudo-class.

0.9.4 – 2016-02-02

Fix nullability attributions and uses of nullable values.
- Fixed by dlkinney in #49.
Add -[HTMLDocument bodyElement] for convenient access to the <body> element.
- Added by zoul in #57.
Add -addChild: and -removeChild: methods to HTMLNode for convenient access to the most common node manipulations.
- Added by zoul and nolanw in #57.

0.9.3 – 2015-11-08

Add tvos deployment target to podspec.

0.9.2 – 2015-10-25

Make HTMLTextNode publicly accessible so that instances are usable when enumerating a node's descendants.
Add -[HTMLNode textComponents] for convenient access to a node's direct text contents.

0.9.1 – 2015-10-23

Export public headers when building static library.
Add watchos deployment target to podspec.

0.9 – 2015-09-20

Add selector groups (e.g. p, span to find all paragraphs and spans). Works in :not() too.

0.8.2 – 2015-09-03

Fix a crash when a document provides an unhelpful <meta charset=>.

0.8.1 – 2015-07-29

HTMLReader no longer crashes when the Content-Type header has the wrong string encoding. Instead, it will pretend that the Content-Type had said nothing at all.

0.8 – 2015-06-27

The public API now has nullability annotations.
The public API now uses generics where available.
Some default values have changed when initializing nodes via init, in order to conform to the (now explicit) nullability annotations:
- HTMLComment defaults to a data of "" (the empty string). Previously its default data was nil.
- HTMLDocumentType defaults to a name of "html". Previously its default name was nil.
- HTMLElement defaults to a tagName of "" (the empty string). Previously its default tagName was nil.
Nullability annotations for parameters are checked using NSParameterAssert. Some methods which previously returned nil when passed a nil parameter will now raise an assertion error. If you get assertion errors after upgrading where you previously did not get assertion errors, this may be why.
HTMLNode's -nodesMatchingSelector:, -firstNodeMatchingSelector, -nodesMatchingParsedSelector:, and -firstNodeMatchingParsedSelector: methods now always throw an NSInvalidArgumentException if the selector fails to parse. Previously they would raise an assertion, but otherwise fail in other (more poorly-defined) ways.
Use angle bracket imports (#import <HTMLReader/…>) throughout public headers, like a proper framework.

0.7.1 – 2015-04-03

Selectors can now include escaped characters, allowing for e.g. matching elements like <some-ns:some-tag> using a selector like some-ns\:some-tag. Supported escapes are \XXXXXX for 1-6 hex digits X, and \c for any other character c.

0.7 – 2015-03-16

Rename namespace properties to htmlNamespace to support compilation as Objective-C++.

0.6.2 – 2015-03-15

Update build settings to allow HTMLReader.framework use in Swift on OS X.

0.6.1 – 2015-02-06

Remove private header HTMLTextNode.h from the built framework.

0.6 – 2015-02-06

A new document initializer, -[HTMLDocument initWithData:contentTypeHeader:], detects the character encoding as a browser would. This is the ideal initializer for HTML documents fetched over the network, or any other time when the character encoding isn't previously known. Note that HTML does not blindly use the encoding specified by the HTTP Content-Type header, so this initializer is still superior to parsing the Content-Type yourself.