Unicode 双向算法 (UBA)

The Challenge of Mixed-Direction Text

English reads left-to-right (LTR). Arabic and Hebrew read right-to-left (RTL). When you have both in the same paragraph — a common situation in multilingual documents, URLs in Arabic text, or numbers in Hebrew — the rendering engine needs a precise set of rules to determine the visual order of characters. That set of rules is the Unicode Bidirectional Algorithm (UBA), specified in Unicode Standard Annex #9.

The UBA operates on logical order (the order characters are stored) and produces visual order (the order glyphs are rendered on screen). Most of the time this is invisible to users — text just displays correctly. But when it goes wrong, entire paragraphs can appear mirrored, or security-relevant filenames can be displayed in a different order than they are stored.

Implicit vs. Explicit Directionality

The UBA assigns every character a Bidi category based on its Unicode property. Common categories:

Using these categories, the algorithm assigns embedding levels (even = LTR, odd = RTL) and resolves the visual order automatically. This implicit handling covers the vast majority of cases.

Explicit Directional Formatting Characters

When implicit resolution produces the wrong order, Unicode provides directional formatting characters to override it:

The LRI/RLI/FSI/PDI isolate controls (added in Unicode 6.3) are preferred over the older embedding controls because isolates do not affect the surrounding text's bidi resolution — they are fully contained.

Security: The Bidi Trojan Source Attack

The RLO character (U+202E) can be used maliciously to display a filename or code string in a different order than it is stored. A file named innocent‮fdp.exe can display as innocent.pdf. This "Trojan Source" attack (CVE-2021-42574) affected code editors that rendered bidi formatting in source files. Mitigation: strip or escape U+202A–U+202E and U+2066–U+2069 in user-supplied text displayed in security contexts.

Quick Facts