Updated support for bidirectional text:
Implemented correct order of directional runs for paragraphs whose base direction is RTL.
Numbers now create an implicit LTR embedding level.
The direction of weak directional characters and neutral characters is now determined by the embedding level of surrounding characters, or if that does not match, by the base paragraph direction.
For the legacy plain text interface, base paragraph direction is now set according to the first strong character.
Shaping uses the base paragraph direction when no strongly directional characters are present.
Added new Data.Text.ParagraphLayout.Rich interface to support rich text.
Supports multiple fonts in the same paragraph. Where line height varies, text will be aligned by its top edge.
Supports nested boxes, each with optional spacing applied to its left and right edge.
Requires user to set base text direction (LTR or RTL).
Moved existing plain text interface into Data.Text.ParagraphLayout.Plain
submodule and marked it as deprecated.
Marked Data.Text.ParagraphLayout.ParagraphConstruction as deprecated.
Input text spans can now have arbitrary user data attached to them.
All related data types (Paragraph, Span, Fragment, SpanLayout,
ParagraphLayout) have been extended with a type variable.
Future-proofed ParagraphOptions and SpanOptions by hiding their
constructors. Use defaultParagraphOptions and defaultSpanOptions instead.
Added line numbers to laid out fragments.
Added information about ancestor boxes of laid out fragments. The plain text interface implicitly generates one box per span, while the rich text interface gives the user direct control over the box hierarchy.
Increased font size for the "shaped runs" output so that it better fits a window with 640 pixels of width.
Fixed compatibility with base < 4.15.
Added partial support for bidirectional text (LTR and RTL in the same paragraph).
The paragraph direction is assumed to be LTR. This will be configurable in a future interface.
Only strong directional characters are used to determine text direction.
The direction of weak directional characters and neutral characters is determined by the nearest preceding strong directional character, or if none is found, the nearest following strong directional character.
Explicit bidirectional formatting characters are ignored.
Exposed functions to allow verification of correct input text slicing:
paragraphSpanBounds, paragraphSpanTexts, and paragraphText.
Added support for forced (hard) line breaks in the input text.
Now also trimming white space at the beginning of lines.
Internally, language tags will be cut at the first invalid character before being passed to ICU.
Tests will generate output in "shaped runs" format which can be passed to a text rendering application.
Text shaping using HarfBuzz.
Breaking text on line and character boundaries using the International Components for Unicode (ICU) library.
Trimming white space at the end of lines.
Calculation of containing block bounds and fragment bounds.
Support for plain text (single font and line height) in one horizontal direction (LTR or RTL).
Input can be divided into spans. This division will be preserved in the output.
Each span can use a different language for text shaping and line breaking.