rtf-parser-tt 0.6.0

Added

• RtfDocument::from_bytes parses an RTF document from bytes. UTF-8 input is read as is; any
  other input is decoded with the code page the document declares. RtfDocument::from_filepath
  and TryFrom<&mut File> read files the same way, so a file no longer needs to be valid UTF-8.
• RtfDocument::get_text_with returns the plain text according to TextOptions: hidden text
  (off by default), footnotes with their numbers (on by default), and headers and footers (on by
  default). RtfDocument::get_text uses the defaults.
• StyleBlock::part identifies the part of the document a block belongs to: Part::Body,
  Part::NoteReference(n) for a footnote number in the body text, Part::Note(n),
  Part::Header or Part::Footer.
• RtfDocument::info holds the document properties from the \info group: title, subject,
  author, manager, company, operator, category, keywords and comment.
• \'xx escapes are decoded through the code page that \ansicpg, \mac, \pc or \pca
  declares, or that the current font's \fcharset selects: Windows code pages 874 and 1250 to
  1258, DOS code pages 437, 850 and 866, Mac Roman, UTF-8, and the Symbol font.
• The encoding_rs feature, enabled by default, decodes the Japanese (932), Simplified Chinese
  (936), Korean (949) and Traditional Chinese (950) code pages.
• RtfError is the error type of parse_rtf and of the RtfDocument constructors. It implements
  std::error::Error and is Send and Sync, and LexerError and ParserError convert into it.
• Painter::hidden marks hidden text (\v).
• Style exposes the name, character formatting and paragraph properties of a stylesheet entry.
• Footnote numbers (\chftn) appear in the text.

Changed

• RtfDocument::try_from, RtfDocument::from_filepath and parse_rtf return RtfError on
  failure. parse_rtf returns a Result; it previously returned the document directly.
• Public enums and structs are marked #[non_exhaustive]. A match on an enum from this crate
  needs a wildcard arm, and a struct is created from its default value and then modified.
  TextOptions sets each option with a method, for example
  TextOptions::default().hidden_text(true).
• Lexer::scan reports \v, \sect, \page, \cell, \row and 11 other control words with
  their own ControlWord variants. 0.5.x reported them as ControlWord::Unknown.
• Paragraph, section and page breaks (\par, \sect, \page) produce a line break, as \line
  does. Table cells are separated by a tab, and each table row ends with a line break. In 0.5.x,
  paragraphs and table cells ran together.
• A StyleBlock continues across paragraph breaks while the formatting stays the same, so its
  text can contain line breaks between other text. Code that treats a block containing a line
  break as a paragraph break alone must split the text at each line break.
• Characters that follow \uN are skipped according to \ucN (one character by default), as the
  RTF specification requires. 0.5.x kept them in the text: \u8212 ? produced an em dash
  followed by a question mark. A document that writes no replacement character without first
  setting \uc0 loses the character after each \uN.
• Hidden text (\v) is omitted from get_text. get_text_with includes it when
  TextOptions::hidden_text is set.
• Footnotes, headers and footers follow the body text, in the order Word uses when it saves a
  document as plain text. In 0.5.x their text appeared where it occurs in the file, inside the
  body text.
• Picture data (\pict) and document properties (\info) no longer appear in the text.
• Only the space that ends a control word is consumed. A further space, or a tab after a control
  word, is text, and whitespace-only groups such as { } are kept.
• Parsing takes about half the time it took in 0.5.1, or less, on the crate's benchmark
  documents.

Removed

• The recursive_tokenize! and recursive_tokenize_with_init! macros.

Fixed

• parse_rtf returns an error for malformed input instead of panicking.
• Text that directly follows a control word is kept when it begins with punctuation or a
  non-ASCII character, for example a comma after \b0 or Cyrillic text after \i. Letters that
  directly follow a control word are part of the control word, as the RTF specification defines,
  so \iword remains a single control word.
• \uN followed directly by its replacement character, as in \u8212?, produces the character.
• The control symbols \~ (no-break space), \- (soft hyphen) and \_ (non-breaking hyphen)
  produce their characters, and the text after them is kept.
• A control word with a nonzero parameter, such as \b2, turns its formatting on.
• Paragraph indents (\li, \ri, \fi) and the line spacing multiple (\slmult) are applied
  to Paragraph.
• Font names written with hex escapes, or followed by an alternate name (\*\falt), are read. In
  0.5.x these names were empty.
• Color table entries are numbered by position, counting empty entries, so \cfN selects the
  intended color.
• A document followed by NUL bytes after its closing brace parses.

Known issues

• Endnotes are numbered in sequence with footnotes and listed among them in file order. Word
  numbers endnotes separately and lists them after the footnotes.
• Text in the Korean Johab code page (1361) is not decoded.