Dart implementation of the Unicode Bidirectional Algorithm (UBA) for converting logical-order text to visual-order Unicode character sequences.
This is the Otzaria-maintained fork of the original package. It is used by Otzaria's text/PDF printing pipeline to order mixed Hebrew, Latin, numerals, punctuation and combining marks when the output renderer needs visual order.
The underlying resolver began with Unicode UAX #9 revision 17. Recent improvements handle supplementary Unicode scalars, UTF-16 source mapping, Hebrew combining-mark order and paired brackets. It is not yet a complete implementation of modern UAX #9: directional isolates (LRI/RLI/FSI/PDI) and all isolating-run-sequence rules are not fully supported. Use a conformant text engine for unrestricted Unicode inputs.
Regular Flutter Text widgets already implement bidirectional text layout.
Do not pre-reorder strings with this package before displaying them in
those widgets: visual reordering is intended for output engines that explicitly
need visual-order glyph/text sequences.
Original upstream demo (not hosted by the Otzaria fork)
To use the fork improvements documented here, add this Git dependency to
your pubspec.yaml:
dependencies:
bidi:
git:
url: https://github.com/Otzaria/bidi.git
ref: 4531defe12ae543fc0034e3690e828653c76f1d6The published bidi: ^2.0.13 dependency does not include these fork improvements.
Then, in your code import and use the package:
import 'package:bidi/bidi.dart' as bidi;
final visualCodePoints = bidi.logicalToVisual('אבג Otzaria 2026');
final visualText = String.fromCharCodes(visualCodePoints);
print(visualText);This package is also written in Rust (apart from this package, which is written in Dart). If you are looking for a pure Rust implementation of this package, look at https://crates.io/crates/bidi/.
For applications that need to associate a visual character with a source
position (for example text selection or PDF hit testing), use
BidiString.fromLogical(input).paragraphs. Each Paragraph.indices entry
refers to an original UTF-16 offset relative to that paragraph; the
aligned Paragraph.sourceLengths entries describe consumed UTF-16 spans.
A paragraph's separator is appended to bidiText but is not part of its
indices or sourceLengths lists.
Alternatively logicalToVisual2(input, indexes, lengths) appends the
global UTF-16 offsets and lengths, including paragraph separators, to
caller-provided lists. A supplementary Unicode scalar occupies two UTF-16
code units in the source, but one Unicode code point in the visual output.
The published bidi package on pub.dev
is currently associated with the original pwa.ir publisher. Merging
changes into this fork does not publish a new package version. A release
must be arranged by an authorized package uploader/publisher. The release
workflow triggers only on version tags such as v2.0.14 and requires the
package owner to configure trusted publishing for Otzaria/bidi on pub.dev.