xmlparser

Added

• Token::span, Tokenizer::stream and allow cloning of Tokenizer. Thanks to @krtab.

Changed

• Optimize is_xml_char function. Makes parsing 5-10% faster. Thanks to @Simon-Martens.

Fixed

• Do no use recursive calls during parsing. Could lead to stack overflow on some input.
• Revert Do not expand predefined references in Stream::consume_reference.
• Tests on Rust 1.61. Thanks to @krtab.

Fixed

• Do not expand predefined references in Stream::consume_reference. Thanks to @Jesse-Bakker.

Changed

• Documentation fixes by @kneasle.

Fixed

• DtdEnd token parsing when ] and > are separated by a whitespace.

Fixed

• Allow processing instruction before DTD.

Fixed

• Allow comments before DTD.

Changed

• Moved to Rust 2018.
• Completely new Error enum.
• New error messages.
• 10-20% faster parsing.
• Use Tokenizer::from_fragment instead of Tokenizer::enable_fragment_mode.

Removed

• TokenType.

Changed

• ]]> is no longer allowed inside a Text node.
• Only XML characters are allowed now. Otherwise, StreamError::NonXmlChar will occur.
• Disallow - at the end of a comment. <!--a---> is an error now.
• A missing space between attributes is an error now.
• StreamError::InvalidQuote and StreamError::InvalidSpace signature changed.

Added

• no_std support thanks to hugwijst.

Changed

• StreamError::InvalidString doesn't store an actual string now.

Changed

• 10-15% faster parsing.
• Merge ByteStream and Stream.
• StreamError::InvalidChar signature changed.
• StreamError::InvalidChar was split into InvalidChar and InvalidCharMultiple.

Fixed

• Check for NameStartChar during qualified name parsing.

  E.g. <-p> is an invalid tag name from now.

• Qualified name with multiple : is an error now.

• ]> is a valid text/CharData now. Previously it was parsed as DoctypeEnd.

Removed

• StreamError::InvalidAttributeValue. StreamError::InvalidChar will be emitted instead.

Added

• span field to all Token variants, which contains a whole token span in bytes.
• Stream::try_consume_byte.

Changed

• All Token variants are structs now and not tuples.

• StrSpan contains an actual string span an not only region now.

  So we can use a non-panic and zero-cost StrSpan::as_str instead of StrSpan::to_str, that was performing slicing each time.

• Split Stream into ByteStream and Stream.

• Stream::skip_spaces will parse only ASCII whitespace now.

• Rename StrSpan::to_str into StrSpan::as_str.

• Rename Reference::EntityRef into Reference::Entity.

• Rename Reference::CharRef into Reference::Char.

• StrSpan::from_substr and StrSpan::slice_region are private now.

Removed

• Token::Whitespaces. Will be parsed as Token::Text.
• Stream::curr_char.
• Stream::is_curr_byte_eq.
• Stream::consume_either.
• Stream::skip_ascii_spaces. Use Stream::skip_spaces instead.
• StrSpan::trim.
• StrSpan::len.
• StrSpan::full_len.
• StrSpan::as_bytes.

Fixed

• Declaration attributes with mixed quotes parsing.

Changed

• Changed the crate category in the Cargo.toml

Added

• Error::pos().

Changed

• Rename Stream::gen_error_pos into Stream::gen_text_pos.
• Rename Stream::gen_error_pos_from into Stream::gen_text_pos_from.
• Stream::gen_text_pos speed up.

Fixed

• TextPos is Unicode aware now.
• XML declaration parsing when file has a BOM.

Changed

• < inside an attribute value is an error now.
• Token::Declaration represents standalone as bool now.
• XML declaration must be defined only once now.
• XML declaration must start at 0 position.
• DTD must be defined only once now.

Added

• Stream::curr_byte_unchecked.

Fixed

• UTF-8 BOM processing.

Changed

• Reference::EntityRef contains &str and not StrSpan now.
• Rename Stream::try_consume_char_reference into try_consume_reference. And it will return Reference and not char now.
• Rename Tokenizer::set_fragment_mode into enable_fragment_mode.
• Rename ErrorPos into TextPos.

Fixed

• TextPos calculation via Stream::gen_error_pos.

Removed

• TextUnescape and XmlSpace because useless.

Added

• StreamError::InvalidChar.
• StreamError::InvalidSpace.
• StreamError::InvalidString.

Changed

• Stream::consume_reference will return only InvalidReference error from now.
• Error::InvalidTokenWithCause merged into Error::InvalidToken.
• Stream::gen_error_pos_from does not require mut self from now.
• StreamError::InvalidChar requires Vec<u8> and not String from now.
• ErrorPos uses u32 and not usize from now.

Removed

• failure dependency.
• log dependency.

Added

• An ability to parse an XML fragment.

Changed

• Relicense from MIT to MIT/Apache-2.0.

Removed

• FromSpan trait.
• from_str and from_span methods are removed. Use the From trait instead.

Changed

• Use failure instead of error-chain.
• Minimum Rust version is 1.18.
• New error messages.
• TokenType is properly public now.

Removed

• ChainedError

Added

• Qualified name parsing.

Changed

• Breaking. Token::ElementStart and Token::Attribute contains prefix and local part of the qualified name now.

Added

• Stream::skip_ascii_spaces.
• Small performance optimizations.

Changed

• log 0.3 -> 0.4