v0.11.0

Updated

• Breaking change: While the public API has not changed, the underlying code
  has been re-written. By default, the parser code is much stricter and follows
  the XML specification more closely. For this release, the parser only has a
  strict mode. Future versions may have options to relax the parsing code and allow
  non-spec behavior.

Fixed

• Fixed parsing of XML document declarations.
  Thanks @Mingun.
  See Issue #9

v0.10.1

Fixed

• Fix incorrect parsing CDATA issue. The pos was not used when scanning for
  the CDATA.
  Thanks @Mingun.
  See Issue #8

v0.10.0

Removed

• Removed deprecated functionality.

v0.9.0

Updated

• Breaking change: Lexer::into_inner() returns a &str.
• Breaking change: Lexer::tokenize() restores the panic behavior in
  previous versions. If the position argument is greater than the input's byte length,
  then a panic occurs. The position argument is also immediately checked to be
  at a UTF-8 character boundary instead of waiting for a panic in a subsequent
  function call.
• Breaking change: Spaces around properties like attribute names, attribute
  values, and processing instructions are trimmed in their representations.
• Breaking change: Minimum supported Rust version is now 1.71.0.
• Renamed Lexer into Reader. Lexer is still exposed as a hidden alias but will
  be removed in a future version.
• Token::ty() is made const.

Removed

• Remove From and from_str implementations from token
  and property types. Due to possible future implementation changes and
  undocumented invariants, the ability to directly construct a type is removed.
• Remove as_str(), as_bytes(), as_str_unchecked(),
  and into_inner() implementations for Attributes and Attribute. Both are opaque
  types where the internal data may change.

Deprecated

• Deprecated from_slice_unchecked() on Lexer. Use core::str::from_utf8_unchecked()
  on the byte slice and then pass into Reader::from_str() if the byte slice is
  guaranteed to be a valid UTF-8 string.

Added

• Most functions are const on the token and property types. Refactored most
  token and property type function implementations to be const compatible.
• Add Reader::parse() as a const alternative function to
  Reader::tokenize(). The function does not mutate the position argument so the
  position must be re-calculated like pos += token.len() in the calling code.
• Add Attributes.parse() as a const function to parse the attributes like
  Reader::parse().
• Add Display implementations to Token and token types.
• Derive PartialOrd, Ord, and Hash for Token and Ty.
• Add len() method to token types.

v0.8.0

Added

• Add Token::as_str() and as_str() to other token and property types.
  as_str() replaces to_str() or as_str_unchecked() as the preferred method
  to get the underlying value.
• Add Token::from_str() and from_str() to other token and property types.
  The method should be used for constructing values in tests.

Deprecated

• Deprecated as_str_unchecked(), to_str() and into_inner() on Token.
• Deprecated as_str_unchecked(), to_str() and into_inner() on token types.
• Deprecated as_str_unchecked() and to_str() on property types.

Removed

• Remove From<&[u8]> and AsRef<[u8]> implementations from token and property
  types.

Changed

• Changed the internal implementation of tokens to use &str instead of
  &[u8]. Micro benchmark performance should be similar to the previous
  implementation, but when using as_str() instead of to_str(), the calling
  code does not have to handle a Result type which should increase overall
  performance.

v0.7.2

Changed

• Make Iterator from Lexer::iter() into an explicit type named Iter.

Added

• Add Lexer::into_inner() to get the underlying bytes passed to the Lexer.
• Derive PartialEq, Eq, and Hash for Lexer and iterators.

v0.7.1

Changed

• Modify IntoIter to be public.

v0.7.0

Changed

• Add Lexer::from_str as the safe instantiation method.
• Refactor Lexer::from_slice to unsafe Lexer::from_slice_unchecked. It is
  assumed the byte slice is UTF-8 characters, and if they are not a valid slice
  of UTF-8 bytes, then the behavior is undefined.
• Dynamically determine Token type instead of storing the Ty. This reduces the
  Token size to 2 words (16 octets on a 64-bit machine).

Removed

• Remove offset field from Token. If required, the offset should be
  calculated when tokens are consumed.
• Remove scan and Scanner from the public API.
• Remove const from Token::ty() method for possible future optimizations.
• Remove public new methods from Token and related types.

Added

• Add unsafe as_str_unchecked to various types in cases where the bytes are
  guaranteed to be valid UTF-8.
• Add as_bytes, to_str, as_str_unchecked, and into_inner to Token.

v0.6.0

Changed

• Change the minimum Rust supported version to 1.56.0 by moving benchmark code
  to own crate.

Removed

• Removed deprecated functionality.

v0.5.0

Added

• Add Lexer type which is the replacement for all evaluators. It borrows the
  input (a slice of bytes) and tokenizes with a borrowed Token. It has the
  benefits of both the existing evaluators (namely borrowing byte slices from the
  input so no allocations are necessary and allowing an Iterator API).
• Add scan function which scans for a token at the beginning of a slice.

Deprecated

• Add deprecations to owned token types and all evaluators. The code will be
  removed in a future version.

Fixed

• Fixed wrong lifetimes in the borrowed token types. The lifetimes should have
  been tied to the borrowed input instead of the token's type.

Changed

• Derive most trait implementations for token and property types.
• Change the minimum Rust supported version to 1.60.0 by changing newer
  syntax code like let...else to older supported syntax.
• Add #[must_use] and const to functions.

Removed

• Remove Token From implementations and other methods which are available
  when using as_bytes().