Proposed Update Unicode Standard Annex #9

Unicode Bidirectional Algorithm

Summary

This annex describes specifications for the positioning of characters in text containing characters flowing from right to left, such as Arabic or Hebrew.

Status

This is a draft document which may be updated, replaced, or superseded by other documents at any time. Publication does not imply endorsement by the Unicode Consortium. This is not a stable document; it is inappropriate to cite this document as other than a work in progress.

This document has been reviewed by Unicode members and other interested parties, and has been approved for publication by the Unicode Consortium. This is a stable document and may be used as reference material or cited as a normative reference by other specifications.

A Unicode Standard Annex (UAX) forms an integral part of the Unicode Standard, but is published online as a separate document. The Unicode Standard may require conformance to normative content in a Unicode Standard Annex, if so specified in the Conformance chapter of that version of the Unicode Standard. The version number of a UAX document corresponds to the version of the Unicode Standard of which it forms a part.

Please submit corrigenda and other comments with the online reporting form [Feedback]. Related information that is useful in understanding this annex is found in Unicode Standard Annex #41, “Common References for Unicode Standard Annexes.” For the latest version of the Unicode Standard, see [Unicode]. For a list of current Unicode Technical Reports, see [Reports]. For more information about versions of the Unicode Standard, see [Versions]. For any errata which may apply to this annex, see [Errata].

Contents

• 1  Introduction
• 2  Directional Formatting Characters
  • 2.1  Explicit Directional Embeddings
  • 2.2  Explicit Directional Overrides
  • 2.3  Terminating Explicit Directional Embeddings and Overrides
  • 2.4  Explicit Directional Isolates
  • 2.5  Terminating Explicit Directional Isolates
  • 2.6  Implicit Directional Marks
  • 2.7  Markup and Formatting Characters
• 3  Basic Display Algorithm
  • 3.1  Definitions
    • 3.1.1  Basics: BD1, BD2, BD3, BD4, BD5, BD6, BD7
    • 3.1.2  Matching Explicit Directional Formatting Characters: BD8, BD9, BD10, BD11, BD12, BD13
    • 3.1.3  Paired Brackets: BD14, BD15, BD16
    • 3.1.4  Additional Abbreviations
  • 3.2  Bidirectional Character Types
  • 3.3  Resolving Embedding Levels
    • 3.3.1  The Paragraph Level: P1, P2, P3
    • 3.3.2  Explicit Levels and Directions: X1, X2, X3, X4, X5, X5a, X5b, X5c, X6, X6a, X7, X8
    • 3.3.3  Preparations for Implicit Processing: X9, X10
    • 3.3.4  Resolving Weak Types: W1, W2, W3, W4, W5, W6, W7
    • 3.3.5  Resolving Neutral and Isolate Formatting Types: N0, N1, N2
    • 3.3.6  Resolving Implicit Levels: I1, I2
  • 3.4  Reordering Resolved Levels: L1, L2, L3, L4
  • 3.5  Shaping
• 4  Bidirectional Conformance
  • 4.1  Boundary Neutrals
  • 4.2  Explicit Formatting Characters
  • 4.3  Higher-Level Protocols: HL1, HL2, HL3, HL4, HL5, HL6
  • 4.4 Bidirectional Conformance Testing
• 5  Implementation Notes
  • 5.1  Reference Code
  • 5.2  Retaining Explicit Formatting Characters
  • 5.3 Maximizing Compatibility for Paired Brackets on Old Systems
  • 5.3  Joiners
  • 5.4  Vertical Text
  • 5.5  Usage
  • 5.6  Separating Punctuation Marks
  • 5.7 Migrating from 2.0 to 3.0
  • 5.8 Conversion to Plain Text
• 6  Usage
  • 6.1  Joiners
  • 6.2  Vertical Text
  • 6.3  Formatting
  • 6.4  Separating Punctuation Marks
  • 6.5  Conversion to Plain Text
  • 6.6  Maximizing Compatibility for Paired Brackets
• 7  Mirroring
• Migration Issues
  • Section Reorganization
• Acknowledgments
• References
• Modifications

1 Introduction

The Unicode Standard prescribes a memory representation order known as logical order. When text is presented in horizontal lines, most scripts display characters from left to right. However, there are several scripts (such as Arabic or Hebrew) where the natural ordering of horizontal text in display is from right to left. If all of the text has a uniform horizontal direction, then the ordering of the display text is unambiguous.

However, because these right-to-left scripts use digits that are written from left to right, the text is actually bidirectional: a mixture of right-to-left and left-to-right text. In addition to digits, embedded words from English and other scripts are also written from left to right, also producing bidirectional text. Without a clear specification, ambiguities can arise in determining the ordering of the displayed characters when the horizontal direction of the text is not uniform.

This annex describes the algorithm used to determine the directionality for bidirectional Unicode text. The algorithm extends the implicit model currently employed by a number of existing implementations and adds explicit formatting characters for special circumstances. In most cases, there is no need to include additional information with the text to obtain correct display ordering.

However, in the case of bidirectional text, there are circumstances where an implicit bidirectional ordering is not sufficient to produce comprehensible text. To deal with these cases, a minimal set of directional formatting characters is defined to control the ordering of characters when rendered. This allows exact control of the display ordering for legible interchange and ensures that plain text used for simple items like filenames or labels can always be correctly ordered for display.

The directional formatting characters are used only to influence the display ordering of text. In all other respects they should be ignored—they have no effect on the comparison of text or on word breaks, parsing, or numeric analysis.

Each character has an implicit bidirectional type. The bidirectional types left-to-right and right-to-left are called strong types, and characters of those types are called strong directional characters. The bidirectional types associated with numbers are called weak types, and characters of those types are called weak directional characters. With the exception of the directional formatting characters, the remaining bidirectional types and characters are called neutral. The algorithm uses the implicit bidirectional types of the characters in a text to arrive at a reasonable display ordering for text.

When working with bidirectional text, the characters are still interpreted in logical order—only the display is affected. The display ordering of bidirectional text depends on the directional properties of the characters in the text. Note that there are important security issues connected with bidirectional text: for more information, see [UTR36].

2 Directional Formatting Characters

Three types of explicit directional formatting characters are used to modify the standard implicit Unicode Bidirectional Algorithm (UBA). In addition, there are implicit directional formatting characters, the right-to-left and left-to-right marks. The effects of all of these formatting characters are limited to the current paragraph; thus, they are terminated by a paragraph separator. The directional types left-to-right and right-to-left are called strong types, and characters of those types are called strong directional characters. The directional types associated with numbers are called weak types, and characters of those types are called weak directional characters.

These formatting characters all have the property Bidi_Control, and are divided into three groups:

U+200E..U+200F LEFT-TO-RIGHT MARK..RIGHT-TO-LEFT MARK

U+061C         ARABIC LETTER MARK

U+202A..U+202E LEFT-TO-RIGHT EMBEDDING..RIGHT-TO-LEFT OVERRIDE

U+2066..U+2069 LEFT-TO-RIGHT ISOLATE..POP DIRECTIONAL ISOLATE

On web pages, the explicit directional formatting characters (of all types – embedding, override, and isolate) should be replaced by using the dir attribute and the elements BDI and BDO. This does not apply to the implicit directional formatting characters. For more information, see [UTR20].

Although the term embedding is used for some explicit formatting characters, the text within the scope of the embedding formatting characters is not independent of the surrounding text. Characters within an embedding can affect the ordering of characters outside, and vice versa. This is not the case with the isolate formatting characters, however. Characters within an isolate cannot affect the ordering of characters outside it, or vice versa. The effect that an isolate as a whole has on the ordering of the surrounding characters is the same as that of a neutral character, whereas an embedding or override roughly has the effect of a strong character.

Directional isolate characters were introduced in Unicode 6.3 after it became apparent that directional embeddings usually have too strong an effect on their surroundings and are thus unnecessarily difficult to use. The new characters were introduced instead of changing the behavior of the existing ones because doing so might have had an undesirable effect on those existing documents that do rely on the old behavior. Nevertheless, the use of the directional isolates instead of embeddings is encouraged in new documents – once target platforms are known to support them.

2.1 Explicit Directional Embeddings

The following characters signal that a piece of text is to be treated as embedded. For example, an English quotation in the middle of an Arabic sentence could be marked as being embedded left-to-right text. If there were a Hebrew phrase in the middle of the English quotation, that phrase could be marked as being embedded right-to-left text. These codes allow for nested embeddings. Embeddings can be be nested one inside another, and in isolates and overrides.

Abbr.	Code Point	Chart	Name	Description
LRE	U+202A		LEFT-TO-RIGHT EMBEDDING	Treat the following text as embedded left-to-right.
RLE	U+202B		RIGHT-TO-LEFT EMBEDDING	Treat the following text as embedded right-to-left.

The effect of right-left line direction, for example, can be accomplished by embedding the text with RLE...PDF. (PDF will be described in section Section 2.3 Terminating Explicit Directional Embeddings and Overrides.)

2.2 Explicit Directional Overrides

The following characters allow the bidirectional character types to be overridden when required for special cases, such as for part numbers. They are to be avoided wherever possible, because of security concerns. For more information, see [UTR36]. These codes allow for nested directional overrides. Directional overrides can be be nested one inside another, and in embeddings and isolates.

Abbr.	Code Point	Chart	Name	Description
LRO	U+202D		LEFT-TO-RIGHT OVERRIDE	Force following characters to be treated as strong left-to-right characters.
RLO	U+202E		RIGHT-TO-LEFT OVERRIDE	Force following characters to be treated as strong right-to-left characters.

The precise meaning of these characters will be made clear in the discussion of the algorithm. The right-to-left override, for example, can be used to force a part number made of mixed English, digits and Hebrew letters to be written from right to left.

2.3 Terminating Explicit Directional Embeddings and Overrides

The following character terminates the scope of the last LRE, RLE, LRO, or RLO whose scope has not yet been terminated and restores the bidirectional state to what it was before that character was encountered.

The precise meaning of this character will be made clear in the discussion of the algorithm.

2.4 Explicit Directional Isolates

The following characters signal that a piece of text is to be treated as directionally isolated from its surroundings. They are very similar to the explicit embedding formatting characters. However, while an embedding roughly has the effect of a strong character on the ordering of the surrounding text, an isolate has the effect of a neutral like U+FFFC OBJECT REPLACEMENT CHARACTER, and is assigned the corresponding display position in the surrounding text. Furthermore, the text inside the isolate has no effect on the ordering of the text outside it, and vice versa.

In addition to allowing the embedding of strongly directional text without unduly affecting the bidirectional order of its surroundings, one of the isolate formatting characters also offers an extra feature: embedding text while inferring its direction heuristically from its constituent characters.

Isolates can be be nested one inside another, and in embeddings and overrides.

Abbr.	Code Point	Chart	Name	Description
LRI	U+2066		LEFT-TO-RIGHT ISOLATE	Treat the following text as isolated and left-to-right.
RLI	U+2067		RIGHT-TO-LEFT ISOLATE	Treat the following text as isolated and right-to-left.
FSI	U+2068		FIRST STRONG ISOLATE	Treat the following text as isolated and in the direction of its first strong directional character that is not inside a nested isolate.

The precise meaning of these characters will be made clear in the discussion of the algorithm.

2.5 Terminating Explicit Directional Isolates

The following character terminates the scope of the last LRI, RLI, or FSI whose scope has not yet been terminated, as well as the scopes of any subsequent LREs, RLEs, LROs, or RLOs whose scopes have not yet been terminated.

Abbr.	Code Point	Chart	Name	Description
PDI	U+2069		POP DIRECTIONAL ISOLATE	End the scope of the last LRI, RLI, or FSI.

The precise meaning of this character will be made clear in the discussion of the algorithm.

2.6 Implicit Directional Marks

These characters are very light-weight formatting. They act exactly like right-to-left or left-to-right characters, except that they do not display or have any other semantic effect. Their use is more convenient than using explicit embeddings or overrides because their scope is much more local.

Abbr.	Code Point	Chart	Name	Description
LRM	U+200E		LEFT-TO-RIGHT MARK	Left-to-right zero-width character
RLM	U+200F		RIGHT-TO-LEFT MARK	Right-to-left zero-width non-Arabic character
ALM	U+061C		ARABIC LETTER MARK	Right-to-left zero-width Arabic character

There is no special mention of the implicit directional marks in the following algorithm. That is because their effect on bidirectional ordering is exactly the same as a corresponding strong directional character; the only difference is that they do not appear in the display.

2.7 Markup and Formatting Characters

The explicit formatting characters introduce state into the plain text, which must be maintained when editing or displaying the text. Processes that are modifying the text without being aware of this state may inadvertently affect the rendering of large portions of the text, for example by removing a PDF.

The Unicode Bidirectional Algorithm is designed so that the use of explicit formatting characters can be equivalently represented by out-of-line information, such as stylesheet information or markup. Conflicts can arise if markup and explicitly formatting characters are both used in the same paragraph. Where available, markup should be used instead of the explicit formatting characters. However, any alternative representation is only to be defined by reference to the behavior of the corresponding explicit formatting characters in this algorithm, to ensure conformance with the Unicode Standard.

HTML 5 provides support for bidi markup as follows:

Whenever plain text is produced from a document containing markup, the equivalent formatting characters should be introduced, so that the correct ordering is not lost. For example, whenever cut and paste results in plain text this transformation should occur.

3 Basic Display Algorithm

The Unicode Bidirectional Algorithm (UBA) takes a stream of text as input and proceeds in four main phases:

• Separation into paragraphs. The rest of the algorithm is applied separately to the text within each paragraph.
• Initialization. A list of bidirectional character types is initialized, with one entry for each character in the original text. The value of each entry is the Bidi_Class property value of the respective character. After this point, the original characters are no longer referenced until the reordering phase. A list of embedding levels, with one level per character, is then initialized. Note that the original characters are referenced in Section 3.3.5, Resolving Neutral and Isolate Formatting Types.
• Resolution of the embedding levels. A series of rules areis applied to the lists of embedding levels and bidirectional character types. Each rule is based operates on the current values of those lists, and can modify those values. Each rule is applied to each of the values in sequence before continuing to the next rule. The original characters and their Bidi_Paired_Bracket and Bidi_Paired_Bracket_Type property values are referenced in the application of certain rules. The result of this phase is a modified list of embedding levels; the list of bidirectional character types is no longer needed.
• Reordering. The text within each paragraph is reordered for display: first, the text in the paragraph is broken into lines, then the resolved embedding levels are used to reorder the text of each line for display.

The algorithm reorders text only within a paragraph; characters in one paragraph have no effect on characters in a different paragraph. Paragraphs are divided by the Paragraph Separator or appropriate Newline Function (for guidelines on the handling of CR, LF, and CRLF, see Section 4.4, Directionality, and Section 5.8, Newline Guidelines of [Unicode]). Paragraphs may also be determined by higher-level protocols: for example, the text in two different cells of a table will be in different paragraphs.

Combining characters always attach to the preceding base character in the memory representation. Even after reordering for display and performing character shaping, the glyph representing a combining character will attach to the glyph representing its base character in memory. Depending on the line orientation and the placement direction of base letterform glyphs, it may, for example, attach to the glyph on the left, or on the right, or above.

This annex uses the numbering conventions for normative definitions and rules in Table 1.

3.1 Definitions

3.1.1 Basics

BD1. The bidirectional characters types are values assigned to each Unicode character, including unassigned characters. The formal property name in the Unicode Character Database [UCD] is Bidi_Class.

BD2. Embedding levels are numbers that indicate how deeply the text is nested, and the default direction of text on that level. The minimum embedding level of text is zero, and the maximum explicit depth is 125, a value referred to as max_depth in the rest of this document.

As rules X1 through X8 will specify, embedding levels are set by explicit formatting characters (embedding, isolate, and override); higher numbers mean the text is more deeply nested. The reason for having a limitation is to provide a precise stack limit for implementations to guarantee the same results. Sixty-one A total of 125 levels is far more than sufficient for ordering, even with mechanically generated formatting; the display becomes rather muddied with more than a small number of embeddings.

BD3. The default direction of the current embedding level (for the character in question) is called the embedding direction. It is L if the embedding level is even, and R if the embedding level is odd.

For example, in a particular piece of text, Level 0 is plain English text. Level 1 is plain Arabic text, possibly embedded within English level 0 text. Level 2 is English text, possibly embedded within Arabic level 1 text, and so on. Unless their direction is overridden, English text and numbers will always be an even level; Arabic text (excluding numbers) will always be an odd level. The exact meaning of the embedding level will become clear when the reordering algorithm is discussed, but the following provides an example of how the algorithm works.

BD4. The paragraph embedding level is the embedding level that determines the default bidirectional orientation of the text in that paragraph.

BD5. The direction of the paragraph embedding level is called the paragraph direction.

• In some contexts the paragraph direction is also known as the base direction.

BD6. The directional override status determines whether the bidirectional type of characters is to be reset. The directional override status is set by using explicit directional formatting characters. This status has three states, as shown in Table 2.

BD7. A level run is a maximal substring of characters that have the same embedding level. It is maximal in that no character immediately before or after the substring has the same level (a level run is also known as a directional run).

As specified below, level runs are important at two different stages of the Bidirectional Algorithm. The first stage occurs after rules X1 through X9 have assigned an explicit embedding level to each character on the basis of the paragraph direction and the explicit directional formatting characters. At this stage, in rule X10, level runs are used to build up the units to which subsequent rules are applied. Those rules further adjust each character’s embedding level on the basis of its implicit bidirectional type and those of other characters in the unit – but not outside it. The level runs resulting from these resolved embedding levels are then used in the actual reordering of the text by rule L2. The following example illustrates level runs at this later stage of the algorithm.

Example

In this and the following examples, case is used to indicate different implicit character types for those unfamiliar with right-to-left letters. Uppercase letters stand for right-to-left characters (such as Arabic or Hebrew), and lowercase letters stand for left-to-right characters (such as English or Russian).

Memory:            car is THE CAR in arabic

Character types:   LLL-LL-RRR-RRR-LL-LLLLLL

Paragraph level:   0

Resolved levels:   000000011111110000000000

Notice that the neutral character (space) between THE and CAR gets the level of the surrounding characters. The level of the neutral characters could be changed by inserting appropriate directional marks around neutral characters, or using explicit directional formatting characters. These marks have no other effects.

	Text			Pairings
	1 2 3 4 5 6 7 8
	a ) b ( c		None
	a ( b ] c		None
	a ( b ) c		2-4
	a ( b [ c ) d ]		2-6
	a ( b ] c ) d		2-6
	a ( b ) c ) d		2-4
	a ( b ( c ) d		4-6
	a ( b ( c ) d )		2-8, 4-6
	a ( b { c } d )		2-8, 4-6
  

3.1.4 Additional Abbreviations

Table 3 lists additional abbreviations used in the examples and internal character types used in the algorithm.

3.2 Bidirectional Character Types

The normative bidirectional character types for each character are specified in the Unicode Character Database [UCD] and are summarized in Table 4. This is a summary only: there are exceptions to the general scope. For example, certain characters such as U+0CBF KANNADA VOWEL SIGN I are given Type L (instead of NSM) to preserve canonical equivalence.

• The term European digits is used to refer to decimal forms common in Europe and elsewhere, and Arabic-Indic digits to refer to the native Arabic forms. (See Section 8.2, Arabic of [Unicode], for more details on naming digits.)
• Unassigned characters are given strong types in the algorithm. This is an explicit exception to the general Unicode conformance requirements with respect to unassigned characters. As characters become assigned in the future, these bidirectional types may change. For assignments to character types, see DerivedBidiClass.txt [DerivedBIDI] in the [UCD].
• Private-use characters can be assigned different values by a conformant implementation.
• For the purpose of the Bidirectional Algorithm, inline objects (such as graphics) are treated as if they are an U+FFFC OBJECT REPLACEMENT CHARACTER.
• As of Unicode 4.0, the Bidirectional Character Types of a few Indic characters were altered so that the Bidirectional Algorithm preserves canonical equivalence. That is, two canonically equivalent strings will result in equivalent ordering after applying the algorithm. This invariant will be maintained in the future.

  Note: The Bidirectional Algorithm does not preserve compatibility equivalence.

3.3 Resolving Embedding Levels

The body of the Bidirectional Algorithm uses bidirectional character types, explicit formatting characters, and bracket pairs to produce a list of resolved levels. This resolution process consists of the following steps:

• Applying rule P1 to split the text into paragraphs, and for each of these:
  • Applying rules P2 and P3 to determine the paragraph level.
  • Applying rule X1 (which employs rules X2–X8) to determine explicit embedding levels and directions.
  • Applying rule X9 to remove many control characters from further consideration.
  • Applying rule X10 to split the paragraph into isolating run sequences and for each of these:
    • Applying rules W1–W7 to resolve weak types.
    • Applying rules N0–N2 to resolve neutral types.
    • Applying rules I1–I2 to resolve implicit embedding levels.

3.3.1 The Paragraph Level

P1. Split the text into separate paragraphs. A paragraph separator is kept with the previous paragraph. Within each paragraph, apply all the other rules of this algorithm.

P2. In each paragraph, find the first character of type L, AL, or R while skipping over any characters between an isolate initiator and its matching PDI or, if it has no matching PDI, the end of the paragraph.

Note that:

• Because paragraph separators delimit text in this algorithm, the character found by this rule will generally be the first strong character after a paragraph separator or at the very beginning of the text.
• The characters between an isolate initiator and its matching PDI are ignored by this rule because a directional isolate is supposed to have the same effect on the ordering of the surrounding text as a neutral character, and the rule ignores neutral characters.
• The characters between an isolate initiator and its matching PDI are ignored by this rule even if the depth limit (as defined in rules X5a through X5c below) prevents the isolate initiator from raising the embedding level. This is meant to make the rule easier to implement.
• Embedding initiators (but not the characters within the embedding) are ignored in this rule. This is because typically they are used to indicate that the embedded text is the opposite direction than the paragraph level.

P3. If a character is found in P2 and it is of type AL or R, then set the paragraph embedding level to one; otherwise, set it to zero.

Whenever a higher-level protocol specifies the paragraph level, rules P2 and P3 may be overridden: see HL1.

3.3.2 Explicit Levels and Directions

All explicit embedding levels are determined from explicit directional formatting characters (embedding, override, and isolate), by applying the explicit level rules X1 through X9. These rules are applied as part of the same This performs a logical pass over the paragraph, applying rules X2–X8 to each characters in turn. The following variables are used during this pass:

As each character is processed, these variables’ values are modified and the character’s explicit embedding level is set as defined by rules X2 through X8 on the basis of the character’s bidirectional type and the variables’ current values.

X1. At the beginning of a paragraph, perform the following steps:

• Set the stack to empty.
• Push onto the stack an entry consisting of the paragraph embedding level, a neutral directional override status, and a false directional isolate status.
• Set the overflow isolate count to zero.
• Set the overflow embedding count to zero.
• Set the valid isolate count to zero.
• Process each character iteratively, applying rules X2 through X8. Only embedding levels from 0 through max_depth are valid in this phase. (Note that in the resolution of levels in rules I1 and I2, the maximum embedding level of max_depth+1 can be reached.)

Explicit Embeddings

X2. With each RLE, perform the following steps:

• Compute the least odd embedding level greater than the embedding level of the last entry on the directional status stack.
• a. If this new level would be valid, and the overflow isolate count and overflow embedding count are both zero, then this RLE is valid. Push an entry consisting of the new embedding level, neutral directional override status, and false directional isolate status onto the directional status stack.
• b. Otherwise, this is an overflow RLE. If the overflow isolate count is zero, increment the overflow embedding count by one. Leave all other variables unchanged.

For example, assuming the overflow counts are both zero, level 0 → 1; levels 1, 2 → 3; levels 3, 4 → 5; ...59, 60 → 61and so on. At max_depth or if either overflow count is non-zero, the level remains the same (overflow RLE).

X3. With each LRE, perform the following steps:

• Compute the least even embedding level greater than the embedding level of the last entry on the directional status stack.
• a. If this new level would be valid, and the overflow isolate count and overflow embedding count are both zero, then this LRE is valid. Push an entry consisting of the new embedding level, neutral directional override status, and false directional isolate status onto the directional status stack.
• b. Otherwise, this is an overflow LRE. If the overflow isolate count is zero, increment the overflow embedding count by one. Leave all other variables unchanged.

For example, assuming the overflow counts are both zero, levels 0, 1 → 2; levels 2, 3 → 4; levels 4, 5 → 6; ...58, 59 → 60and so on. At max_depth or max_depth-1 (which, being even, would have to go to max_depth+1) or if either overflow count is non-zero, the level remains the same (overflow LRE).

Explicit Overrides

An explicit directional override sets the embedding level in the same way the explicit embedding formatting characters do, but also changes the bidirectional character type of affected characters to the override direction.

X4. With each RLO, perform the following steps:

• Compute the least odd embedding level greater than the embedding level of the last entry on the directional status stack.
• a. If this new level would be valid, and the overflow isolate count and overflow embedding count are both zero, then this RLO is valid. Push an entry consisting of the new embedding level, right-to-left directional override status, and false directional isolate status onto the directional status stack.
• b. Otherwise, this is an overflow RLO. If the overflow isolate count is zero, increment the overflow embedding count by one. Leave all other variables unchanged.

X5. With each LRO, perform the following steps:

• Compute the least even embedding level greater than the embedding level of the last entry on the directional status stack.
• a. If this new level would be valid, and the overflow isolate count and overflow embedding count are both zero, then this LRO is valid. Push an entry consisting of the new embedding level, left-to-right directional override status, and false directional isolate status onto the directional status stack.
• b. Otherwise, this is an overflow LRO. If the overflow isolate count is zero, increment the overflow embedding count by one. Leave all other variables unchanged.

X6. For all types besides B, BN, RLE, LRE, RLO, LRO, PDF, RLI, LRI, FSI, and PDI:

• a. Set the current character’s embedding level to the embedding level of the last entry on the directional status stack.
• b. Whenever the directional override status of the last entry on the directional status stack is not neutral, reset the current character type according to the directional override status of the last entry on the directional status stack.

In other words, if the directional override status of the last entry on the directional status stack is neutral, then characters retain their normal types: Arabic characters stay AL, Latin characters stay L, spaces stay WS, and so on. If the directional override status is right-to-left, then characters become R. If the directional override status is left-to-right, then characters become L.

Note that the current embedding level is not changed by this rule.

Terminating Embeddings and Overrides

There is a single code to terminate the scope of the current explicit code, whether an embedding or a directional override. All codes and pushed states are completely popped at the end of paragraphs.

A PDF terminates the scope of the embedding initiator it matches. If it does not match any embedding initiator, it is ignored.

X7. With each PDF, determine the matching embedding or override code. If there was a valid matching code, restore (pop) the last remembered (pushed) embedding level and directional override. perform the following steps:

X8. All explicit directional embeddings, overrides and isolates are completely terminated at the end of each paragraph. Paragraph separators are not included in any embedding, override or isolate, and are thus assigned the paragraph embedding level.

X9. Remove all RLE, LRE, RLO, LRO, PDF, and BN characters.

• Note that an implementation does not have to actually remove the characters; it just has to behave as though the characters were not present for the remainder of the algorithm. Conformance does not require any particular placement of these characters as long as all other characters are ordered correctly.

  See Section 5, Implementation Notes, for information on implementing the algorithm without removing the formatting characters.

• The zero width joiner and non-joiner affect the shaping of the adjacent characters—those that are adjacent in the original backing-store order, even though those characters may end up being rearranged to be non-adjacent by the Bidirectional Algorithm. For more information, see Section 6.1, Joiners.
• Note that FSI, LRI, RLI, and PDI characters are not removed. As indicated by the rules below, they are used, in part, to determine the paragraph’s isolating run sequences, within which they are then treated as neutral characters. Nevertheless, they are of course zero-width characters and, like LRM and RLM, should not be visible in the final output.

X10. Perform the following steps:

• Compute the set of isolating run sequences as specified by BD13, based on the bidirectional types of the characters and the embedding levels assigned by the rules above (X1–X9).
• Determine the start-of-sequence (sos) and end-of-sequence (eos) types, either L or R, for each isolating run sequence. These depend on the higher of the two levels on either side of the sequence boundary:
  • For sos, compare the level of the first character in the sequence with the level of the character preceding it in the paragraph (not counting characters removed by X9), and if there is none, with the paragraph embedding level.
  • For eos, compare the level of the last character in the sequence with the level of the character following it in the paragraph (not counting characters removed by X9), and if there is none or the last character of the sequence is an isolate initiator (lacking a matching PDI), with the paragraph embedding level.
  • If the higher level is odd, the sos or eos is R; otherwise, it is L.
  • Note that these computations must use the embedding levels assigned by the rules above, before any changes are made to them in the steps below.
• Apply rules W1–W7, N0–N2, and I1–I2, in the order in which they appear below, to each of the isolating run sequences, applying one rule to all the characters in the sequence in the order in which they occur in the sequence before applying another rule to any part of the sequence. The order that one isolating run sequence is treated relative to another does not matter. When applying a rule to an isolating run sequence, the last character of each level run in the isolating run sequence is treated as if it were immediately followed by the first character in the next level run in the sequence, if any.

Levels:  0   0   0   1   1   1   2

Runs:   <--- 1 ---> <--- 2 ---> <3>

3.3.4 Resolving Weak Types

Weak types are now resolved one isolating run sequence at a time. At isolating run sequence boundaries where the type of the character on the other side of the boundary is required, the type assigned to sos or eos is used.

First, each nonspacing mark is resolved based on the previous character it follows.

W1. Examine each nonspacing mark (NSM) in the isolating run sequence, and change the type of the NSM to Other Neutral if the previous character is an isolate initiator or PDI, and to the type of the previous character otherwise. If the NSM is at the start of the isolating run sequence, it will get the type of sos. (Note that in an isolating run sequence, an isolate initiator followed by an NSM or any type other than PDI must be an overflow isolate initiator.)

Assume in this example that sos is R:

AL  NSM NSM → AL  AL  AL

sos NSM     → sos R

LRI NSM     → LRI ON

PDI NSM     → PDI ON

The text is next parsed for numbers. This pass will change the directional types European Number Separator, European Number Terminator, and Common Number Separator to be European Number text, Arabic Number text, or Other Neutral text. The text to be scanned may have already had its type altered by directional overrides. If so, then it will not parse as numeric.

W2. Search backward from each instance of a European number until the first strong type (R, L, AL, or sos) is found. If an AL is found, change the type of the European number to Arabic number.

AL EN     → AL AN

AL NI EN  → AL NI AN

sos NI EN → sos NI EN

L NI EN   → L NI EN

R NI EN   → R NI EN
    

W3. Change all ALs to R.

W4. A single European separator between two European numbers changes to a European number. A single common separator between two numbers of the same type changes to that type.

EN ES EN → EN EN EN

EN CS EN → EN EN EN

AN CS AN → AN AN AN

W5. A sequence of European terminators adjacent to European numbers changes to all European numbers.

ET ET EN → EN EN EN

EN ET ET → EN EN EN

AN ET EN → AN EN EN

W6. Otherwise, separators and terminators change to Other Neutral.

AN ET    → AN ON

L  ES EN → L  ON EN

EN CS AN → EN ON AN

ET AN    → ON AN

W7. Search backward from each instance of a European number until the first strong type (R, L, or sos) is found. If an L is found, then change the type of the European number to L.

L  NI EN → L  NI  L

R  NI EN → R  NI  EN

3.3.5 Resolving Neutral and Isolate Formatting Types

In the next phase, neutral and isolate formatting (i.e. NI) characters are resolved one isolating run sequence at a time. The next phase resolves the direction of the neutrals. Its results are that all NIs become either R or L. Generally, NIs take on the direction of the surrounding text. In case of a conflict, they take on the embedding direction. At isolating run sequence boundaries where the type of the character on the other side of the boundary is required, the type assigned to sos or eos is used.

N1. A sequence of NIs takes the direction of the surrounding strong text if the text on both sides has the same direction. European and Arabic numbers act as if they were R in terms of their influence on NIs. The start-of-sequence (sos) and end-of-sequence (eos) types are used at isolating run sequence boundaries.

 L  NI   L  →   L  L   L

 R  NI   R  →   R  R   R

 R  NI  AN  →   R  R  AN

 R  NI  EN  →   R  R  EN

AN  NI   R  →  AN  R   R

AN  NI  AN  →  AN  R  AN

AN  NI  EN  →  AN  R  EN

EN  NI   R  →  EN  R   R

EN  NI  AN  →  EN  R  AN

EN  NI  EN  →  EN  R  EN

N2. Any remaining NIs take the embedding direction.

NI → e

The embedding direction for the given NI character is derived from its embedding level: L if the character is set to an even level, and R if the level is odd. (See BD3.)

Assume in the following example that eos is L and sos is R. Then an application of N1 and N2 yields the following:

L   NI eos → L   L eos

R   NI eos → R   e eos

sos NI L   → sos e L

sos NI R   → sos R R

Examples. A list of numbers separated by neutrals and embedded in a directional run will come out in the run’s order.

Storage:	he said "THE VALUES ARE 123, 456, 789, OK".

Display:	he said "KO ,789 ,456 ,123 ERA SEULAV EHT".

In this case, both the comma and the space between the numbers take on the direction of the surrounding text (uppercase = right-to-left), ignoring the numbers. The commas are not considered part of the number because they are not surrounded on both sides by digits (see Section 3.3.4, Resolving Weak Types). However, if there is a preceding left-to-right sequence, then European numbers will adopt that direction:

Storage:	IT IS A bmw 500, OK.

Display:	.KO ,bmw 500 A SI TI

3.3.6 Resolving Implicit Levels

In the final phase, the embedding level of text may be increased, based on the resolved character type. Right-to-left text will always end up with an odd level, and left-to-right and numeric text will always end up with an even level. In addition, numeric text will always end up with a higher level than the paragraph level. (Note that it is possible for text to end up at levels higher than 61 max_depth+1 as a result of this process.) This results in the following rules:

I1. For all characters with an even (left-to-right) embedding level, those of type R go up one level and those of type AN or EN go up two levels.

I2. For all characters with an odd (right-to-left) embedding level, those of type L, EN or AN go up one level.

Table 5 summarizes the results of the implicit algorithm.

3.4 Reordering Resolved Levels

The following rules describe the logical process of finding the correct display order. As opposed to resolution phases, these rules act on a per-line basis and are applied after any line wrapping is applied to the paragraph.

Logically there are the following steps:

• The levels of the text are determined according to the previous rules.
• The characters are shaped into glyphs according to their context (taking the embedding levels into account for mirroring).
• The accumulated widths of those glyphs (in logical order) are used to determine line breaks.
• For each line, rules L1–L4 are used to reorder the characters on that line.
• The glyphs corresponding to the characters on the line are displayed in that order.

L1. On each line, reset the embedding level of the following characters to the paragraph embedding level:

1. Segment separators,
2. Paragraph separators,
3. Any sequence of whitespace characters and/or isolate formatting characters (FSI, LRI, RLI, and PDI) preceding a segment separator or paragraph separator, and
4. Any sequence of whitespace characters and/or isolate formatting characters (FSI, LRI, RLI, and PDI) at the end of the line.

• The types of characters used here are the original types, not those modified by the previous phase.
• Because a Paragraph Separator breaks lines, there will be at most one per line, at the end of that line.

In combination with the following rule, this means that trailing whitespace will appear at the visual end of the line (in the paragraph direction). Tabulation will always have a consistent direction within a paragraph.

L2. From the highest level found in the text to the lowest odd level on each line, including intermediate levels not actually present in the text, reverse any contiguous sequence of characters that are at that level or higher.

This rule reverses a progressively larger series of substrings.

The following examples illustrate the reordering, showing the successive steps in application of Rule L2. The original text including any embedding codes for producing the particular levels is shown in the "Storage" row in the example tables. The invisible, zero-width formatting characters LRI, RLI, and PDI are represented with the symbols >, <, and =, respectively. The application of the rules from Section 3.3, Resolving Embedding Levels and of the Rule L1 results in the resolved levels. These are listed in the rows "Before Reordering" and "Resolved Levels" row. (Since these examples only make use of the isolate formatting characters, Rule X9 does not remove any characters. Note that Example 3 would not work if it used embeddings instead because the two right-to-left phrases would have merged into a single right-to-left run, together with the neutral punctuation in between.) Each successive row thereafter shows one pass of reversal from Rule L2, such as "Reverse levels 1-2". At each iteration, the underlining shows the text that has been reversed.

The paragraph embedding level for the first, second, and third examples is 0 (left-to-right direction), and for the second and fourth example is 1 (right-to-left direction).

Example 1 (embedding level = 0)

Example 2 (embedding level = 0)

Example 3 (embedding level = 0)

Example 4 (embedding level = 1)

L3. Combining marks applied to a right-to-left base character will at this point precede their base character. If the rendering engine expects them to follow the base characters in the final display process, then the ordering of the marks and the base character must be reversed.

Many font designers provide default metrics for combining marks that support rendering by simple overhang. Because of the reordering for right-to-left characters, it is common practice to make the glyphs for most combining characters overhang to the left (thus assuming the characters will be applied to left-to-right base characters) and make the glyphs for combining characters in right-to-left scripts overhang to the right (thus assuming that the characters will be applied to right-to-left base characters). With such fonts, the display ordering of the marks and base glyphs may need to be adjusted when combining marks are applied to “unmatching” base characters. See Section 5.13, Rendering Nonspacing Marks of [Unicode], for more information.

L4. A character is depicted by a mirrored glyph if and only if (a) the resolved directionality of that character is R, and (b) the Bidi_Mirrored property value of that character is trueYes.

• The Bidi_Mirrored property is defined by Section 4.7, Bidi Mirrored—Normative of [Unicode]; the property values are specified in [UCD].
• This rule can be overridden in certain cases; see HL6.

For example, U+0028 left parenthesis—which is interpreted in the Unicode Standard as an opening parenthesis—appears as “(” when its resolved level is even, and as the mirrored glyph “)” when its resolved level is odd. Note that for backward compatibility the characters U+FD3E (﴾) ORNATE LEFT PARENTHESIS and U+FD3F (﴿) ORNATE RIGHT PARENTHESIS are not mirrored.

3.5 Shaping

Cursively connected scripts, such as Arabic or Syriac, require the selection of positional character shapes that depend on adjacent characters (see Section 8.2, Arabic of [Unicode]). Shaping is logically applied after the Bidirectional Algorithm is used and is limited to characters within the same level run. (Note that there is no practical difference between limiting shaping to a level run and an isolating run sequence because the isolate initiator and PDI characters are defined to have joining type U, i.e. non-joining. Thus, the characters before and after a directional isolate will not join across the isolate, even if the isolate is empty or overflows the depth limit.) Consider the following example string of Arabic characters, which is represented in memory as characters 1, 2, 3, and 4, and where the first two characters are overridden to be LTR. To show both paragraph directions, the next two are embedded, but with the normal RTL direction.

1	2	3	4
ج 062C JEEM	ع 0639 AIN	ل 0644 LAM	م 0645 MEEM
L	L	R	R

One can use explicit directional formatting characters to achieve this effect in plain text or use markup in HTML, as in the examples below. (The bold text would be for the right-to-left paragraph direction.)

• LRM/RLM LRO JEEM AIN PDF RLO LAM MEEM PDF
• <p dir="ltr"/"rtl">LRO JEEM AIN PDF RLO LAM MEEM PDF</p>
• <p dir="ltr"/"rtl"><bdo dir="ltr">JEEM AIN</bdo>
                <bdo dir="rtl">LAM MEEM</bdo></p>

The resulting shapes will be the following, according to the paragraph direction:

3.5.1 Shaping and Line Breaking

The process of breaking a paragraph into one or more lines that fit within particular bounds is outside the scope of the Bidirectional Algorithm. Where character shaping is involved, the width calculations must be based on the shaped glyphs.

Note that the soft-hyphen (SHY) works in cursively connected scripts as it does in other scripts. That is, it indicates a point where the line could be broken in the middle of a word. If the rendering system breaks at that point, the display—including shaping—should be what is appropriate for the given language. For more information on this and other line breaking issues, see Unicode Standard Annex #14, “Line Breaking Properties” [UAX14].

4 Bidirectional Conformance

A process that claims conformance to this specification shall satisfy the following clauses:

• As is the case for all other Unicode algorithms, this is a logical description—particular implementations can have more efficient mechanisms as long as they produce the same results. See C18 in Chapter 3, Conformance of [Unicode], and the notes following.
• The Bidirectional Algorithm specifies part of the intrinsic semantics of right-to-left characters and is thus required for conformance to the Unicode Standard where any such characters are displayed.

Use of higher-level protocols is discouraged, because it introduces interchange problems and can lead to security problems. For more information, see Unicode Technical Report #36, “Unicode Security Considerations” [UTR36].

4.1 Boundary Neutrals

The goal in marking a formatting or control character as BN is that it have no effect on the rest of the algorithm. (ZWJ and ZWNJ are exceptions; see X9). Because conformance does not require the precise ordering of formatting characters with respect to others, implementations can handle them in different ways as long as they preserve the ordering of the other characters.

4.2 Explicit Formatting Characters

As with any Unicode characters, systems do not have to support any particular explicit directional formatting character (although it is not generally useful to include a terminating character without including the initiator). Generally, conforming systems will fall into four classes:

• No bidirectional formatting. This implies that the system does not visually interpret characters from right-to-left scripts.
• Implicit bidirectionality. The implicit Bidirectional Algorithm and the directional marks ALM, RLM and LRM are supported.
• Non-isolate bidirectionality. The implicit Bidirectional Algorithm, the implicit directional marks, and the explicit non-isolate directional formatting characters are supported: ALM, RLM, LRM, LRE, RLE, LRO, RLO, PDF.
• Full bidirectionality. The implicit Bidirectional Algorithm, the implicit directional marks, and all the explicit directional formatting characters are supported: ALM, RLM, LRM, LRE, RLE, LRO, RLO, PDF, FSI, LRI, RLI, PDI.

4.3 Higher-Level Protocols

The following clauses are the only permissible ways for systems to apply higher-level protocols to the ordering of bidirectional text. Some of the clauses apply to segments of structured text. This refers to the situation where text is interpreted as being structured, whether with explicit markup such as XML or HTML, or internally structured such as in a word processor or spreadsheet. In such a case, a segment is span of text that is distinguished in some way by the structure. 

Clauses HL1 and HL3 are specialized applications of the more general clauses HL4 and HL5. They are provided here explicitly because they directly correspond to common operations.

As an example of the application of HL4, suppose an XML document contains the following fragment. (Note: This is a simplified example for illustration: element names, attribute names, and attribute values could all be involved.)

ARABICenglishARABIC<e1 type='ab'>ARABICenglish<e2 type='cd'>english

This can be analyzed as being five different segments:

1. ARABICenglishARABIC
2. <e1 type='ab'>
3. ARABICenglish
4. <e2 type='cd'>
5. english

To make the XML file readable as source text, the display in an editor could order these elements all in a uniform direction (for example, all left-to-right) and apply the Bidirectional Algorithm to each field separately. It could also choose to order the element names, attribute names, and attribute values uniformly in the same direction (for example, all left-to-right). For final display, the markup could be ignored, allowing all of the text (segments a, c, and e) to be reordered together.

4.4 Bidirectional Conformance Testing

The BidiTest.txt file in the Unicode Character Database [UCD] provides a conformance test for UBA implementations. It is designed to be reasonably compact, and yet provide a thorough test of all cases up to a given limit (currently 4). The format is described in detail in the header of the file.

The Unicode Character Database [UCD] includes two files that provide conformance tests for implementations of the Bidirectional Algorithm [Tests9]. One of the test files, BidiTest.txt, comprises exhaustive test sequences of bidirectional types up to a given length, currently 4. The other test file, BidiCharacterTest.txt, contains test sequences of explicit code points, including, for example, bracket pairs. The format of each test file is described in the header of that file.

5 Implementation Notes

5.1 Reference Code

There are two versions of BIDI reference code available. Both have been tested to produce identical results. One version is written in Java, and the other is written in C++. The Java version is designed to closely follow the steps of the algorithm as described below. The C++ code is designed to show one of the optimization methods that can be applied to the algorithm, using a state table for one phase.

One of the most effective optimizations is to first test for right-to-left characters and not invoke the Bidirectional Algorithm unless they are present.

There are two directories containing source code for reference implementations at [Code9]. Implementers are encouraged to use this resource to test their implementations. There is an online demo of bidi code at http://unicode.org/cldr/utility/bidi.jsp, which shows the results, plus the levels and the rules invoked for each character.

A reference implementation of the Bidirectional Algorithm written in Java is available. The source code can be downloaded from [Code9]. Implementers are encouraged to use this resource to test their implementations. An online demo is also available at [Demo9], which shows the results of the Bidirectional Algorithm, as well as the embedding levels and the rules invoked for each character.

The reference code is designed to follow the steps of the algorithm without applying any optimizations. An example of an effective optimization is to first test for right-to-left characters and invoke the Bidirectional Algorithm only if they are present. Another example of optimization is in matching bracket pairs. The bidirectional bracket pairs (the characters with Bidi_Paired_Bracket_Type property values Open and Close) constitute a subset of the characters with bidirectional type ON. Conversely, the characters with a bidirectional type distinct from ON have the Bidi_Paired_Bracket_Type property value None. Therefore, lookup of Bidi_Paired_Bracket_Type property values for the identification of bracket pairs can be optimized by restricting the processing to characters whose bidirectional type is ON.

5.2 Retaining Explicit Formatting Characters

Some implementations may wish to retain the explicit directional embedding and override formatting characters when running the algorithm. The following provides a summary of how this may be done. Note that this summary is an informative implementation guideline; it should provide the same results as the explicit algorithm above, but in case of any deviation the explicit algorithm is the normative statement for conformance.

• In rule X9, instead of removing the embedding and override formatting characters, assign the embedding level to each formatting character, and turn it into BN.
• In rule X10, assign L or R to the last of a sequence of adjacent BNs according to the eor / sor, and set the level to the higher of the two levels when determining the sos and eos for an isolating run sequence, skip over any BNs when looking for the character preceding the isolating run sequence's first character and following its last character.
• In rule W1, search backward from each NSM to the first character in the isolating run sequence whose bidirectional type is not BN, and set the NSM to ON if it is an isolate initiator or PDI, and to its type otherwise. If the NSM is the first non-BN character, change the NSM to the type of sos.
• In rule W4, scan past BN types that are adjacent to ES or CS.
• In rule W5, change all appropriate sequences of ET and BN, not just ET.
• In rule W6, change all BN types adjacent to ET, ES, or CS to ON as well.
• In rule W7, scan past BN.
• In rules N0–N2, treat BNs that adjoining neutrals the same as those neutrals.
• In rules I1 and I2, ignore BN.
• In rule L1, include the embedding and override formatting characters and BN together with whitespace characters and isolate formatting characters in the sequences whose level gets reset before a separator or line break. Resolve any LRE, RLE, LRO, RLO, PDF, or BN to the level of the preceding character if there is one, and otherwise to the base level.

Implementations that display visible representations of formatting characters will want to adjust this process to position the formatting characters optimally for editing.

6 Usage

6.1 Joiners

As described under X9, the zero width joiner and non-joiner affect the shaping of the adjacent characters—those that are adjacent in the original backing-store order—even though those characters may end up being rearranged to be non-adjacent by the Bidirectional Algorithm. To determine the joining behavior of a particular character after applying the Bidirectional Algorithm, there are two main strategies:

• When shaping, an implementation can refer back to the original backing store to see if there were adjacent ZWNJ or ZWJ characters.
• Alternatively, the implementation can replace ZWJ and ZWNJ by an out-of-band character property associated with those adjacent characters, so that the information does not interfere with the Bidirectional Algorithm and the information is preserved across rearrangement of those characters. Once the Bidirectional Algorithm has been applied, that out-of-band information can then be used for proper shaping.

6.2 Vertical Text

In the case of vertical line orientation, the Bidirectional Algorithm is still used to determine the levels of the text. However, these levels are not used to reorder the text, because the characters are usually ordered uniformly from top to bottom. Instead, the levels are used to determine the rotation of the text. Sometimes vertical lines follow a vertical baseline in which each character is oriented as normal (with no rotation), with characters ordered from top to bottom whether they are Hebrew, numbers, or Latin. When setting text using the Arabic script in vertical lines, it is more common to employ a horizontal baseline that is rotated by 90° counterclockwise so that the characters are ordered from top to bottom. Latin text and numbers may be rotated 90° clockwise so that the characters are also ordered from top to bottom.

The Bidirectional Algorithm is used when some characters are ordered from bottom to top. For example, this happens with a mixture of Arabic and Latin glyphs when all the glyphs are rotated uniformly 90° clockwise. The Unicode Standard does not specify whether text is presented horizontally or vertically, or whether text is rotated. That is left up to higher-level protocols.

6.3 Formatting

Because of the implicit character types and the heuristics for resolving neutral and numeric directional behavior, the implicit bidirectional ordering will generally produce the correct display without any further work. However, problematic cases may occur when a right-to-left paragraph begins with left-to-right characters, or there are nested segments of different-direction text, or there are weak characters on directional boundaries. In these cases, embeddings or directional marks may be required to get the right display. Part numbers may also require directional overrides.

The most common problematic case is that of neutrals on the boundary of an embedded language. This can be addressed by setting the level of the embedded text correctly. For example, with all the text at level 0 the following occurs:

Memory:  he said "I NEED WATER!", and expired.

Display: he said "RETAW DEEN I!", and expired.

If the exclamation mark is to be part of the Arabic quotation, then the user can select the text I NEED WATER! and explicitly mark it as embedded Arabic, which produces the following result:

Memory:  he said "RLII NEED WATER!PDI", and expired.

Display: he said "!RETAW DEEN I", and expired.

However, an often simpler and better method of doing this is to place a right directional mark (RLM) after the exclamation mark. Because the exclamation mark is now not on a directional boundary, this produces the correct result. This is the best approach when manually editing text or programmatically generating text meant to be edited, or dealing with an application that simply does not support explicit formatting characters.

Memory:  he said "I NEED WATER!RLM", and expired.

Display: he said "!RETAW DEEN I", and expired.

This latter approach is preferred because it does not make use of the explicit formatting characters, which can easily get out of sync if not fully supported by editors and other string manipulation. Nevertheless, the explicit formatting characters are generally needed only for more complex (and rare)absolutely necessary in cases such as double embeddings,where text of one direction contains text of the opposite direction which itself contains text of the original direction. Such cases are not as rare as one might think, because Latin-script brand names, technical terms, and abbreviations are often written in their original Latin characters when used in non-Latin-script text, including right-to-left text, as in the following:

Memory:  DID YOU SAY ‘LREhe said "I NEED WATER!RLM", and expired.PDF’?

Display: ?‘he said "!RETAW DEEN I", and expired.’ YAS UOY DID

Memory:  it is called "RLIAN INTRODUCTION TO javaPDI" - $19.95 in hardcover.

Display: it is called "java OT NOITCUDORTNI NA" - $19.95 in hardcover.

Thus, when text is programmatically generated by inserting data into a template, and is not intended for later manual editing, and a particular insert happens to be of the opposite direction to the template's text, it is easiest to wrap the insert in explicit formatting characters (or their markup equivalent) declaring its direction, without analyzing whether it is really necessary to do so, or if the job could be done just with stateless directional marks.

Furthermore, in this common scenario, it is highly recommended to use directional isolate formatting characters as opposed to directional embedding formatting characters (once targeted display platforms are known to support isolates). This is because embeddings affect the surrounding text similarly to a strong character, whereas directional isolates have the effect of a neutral. The embeddings' stronger effect is often difficult to anticipate and is rarely useful. To demonstrate, here is the example above with embeddings instead of isolates:

Memory:  it is called "RLEAN INTRODUCTION TO javaPDF" - $19.95 in hardcover.

Display: it is called "$19.95 - "java OT NOITCUDORTNI NA in hardcover.

This, of course, is not the intended display, and is due to the number “sticking” to the preceding RTL embedding (along with all the neutral characters in between), just as it would “stick” to a preceding RTL character.

Directional isolates also offer a solution to the very common case where the direction of the text to be programmatically inserted is not known. Instead of analyzing the characters of the text to be inserted in order to decide whether to use an LRE or RLE (or LRI or RLI - or nothing at all), the software can take the easy way out and always wrap each unknown-direction insert in an FSI and PDI. Thus, an FSI instead of an RLI in the example above would produce the same display. FSI's first-strong heuristic is not infallible, but it will work most of the time even on mixed-script text.

Although wrapping inserts in isolates is a useful technique, it is best not to wrap text that is known to contain no opposite-direction characters that are not already wrapped in an isolate. Unnecessary layers of wrapping not only add bulk and complexity; they can also wind up exceeding the depth limit and rendering ineffective the innermost isolates, which can make the text display incorrectly. One very common case of an insert that does not need wrapping is one known to be localized to the context locale, e.g. a translated message with all its inserted values either themselves localized, or wrapped in an isolate.

6.4 Separating Punctuation Marks

A common problem case is where the text really represents a sequence of items with separating punctuation marks, often programmatically concatenated. These separators are often strings of neutral characters. For example, a web page might have the following at the bottom:

advertising programs - business solutions - privacy policy - help - about

This might be built up on the server by concatenating a variable number of strings with " - " as a separator, for example. If all of the text is translated into Arabic or Hebrew and the overall page direction is set to be RTL, then the right result occurs, such as the following:

TUOBA - PLEH - YCILOP YCAVIRP - SNOITULOS SSENISUB - SMARGORP GNISITREVDA

However, suppose that in the translation, there remain some LTR characters. This is not uncommon for company names, product names, technical terms, and so on. If one of the separators is bounded on both sides by LTR characters, then the result will be badly jumbled. For example, suppose that "programs" in the first term and "business" in the second were left in English. Then the result would be

TUOBA - PLEH - YCILOP YCAVIRP - SNOITULOS programs - business GNISITREVDA

The result is a jumble, with the apparent first term being "advertising business" and the second being "programs solutions". The simplest solution for this problem is to include an RLM character in each separator string. That will cause each separator to adopt a right-to-left direction, and produce the correct output:

TUOBA - PLEH - YCILOP YCAVIRP - SNOITULOS business - programs GNISITREVDA

The explicit formatting characters (LRE, RLE, and PDF or LRI, RLI, FSI, and PDI) can be used to achieve the same effect; web pages would use spans with the attributes dir="ltr" or dir="rtl". Each separate field would be embedded, excluding the separators. In general, LRM and RLM are preferred to the explicit formatting characters because their effects are more local in scope, and are more robust than the dir attributes when text is copied. (Ideally programs would convert dir attributes to the corresponding explicit formatting characters when converting to plain text, but that is not generally supported.)

6.5 Conversion to Plain Text

For consistent appearance, when bidirectional text subject to a higher-level protocol is to be converted to Unicode plain text, formatting characters should be inserted to ensure that the display order resulting from the application of the Unicode Bidirectional Algorithm matches that specified by the higher-level protocol. The same principle should be followed whenever text using a higher-level protocol is converted to marked-up text that is unaware of the higher-level protocol. For example, if a higher-level protocol sets the paragraph direction to 1 (R) based on the number of L versus R/AL characters, when converted to plain text the paragraph would be embedded in a bracketing pair of RLE..PDF formatting characters. If the same text were converted to HTML4.0 the attribute dir = "rtl" would be added to the paragraph element.

7 Mirroring

The mirrored property is important to ensure that the correct characters are used for the desired semantic. This is of particular importance where the name of a character does not indicate the intended semantic, such as with U+0028 “(” LEFT PARENTHESIS. While the name indicates that it is a left parenthesis, the character really expresses an open parenthesis—the leading character in a parenthetical phrase, not the trailing one.

Some of the characters that do not have the Bidi_Mirrored property may be rendered with mirrored glyphs, according to a higher level protocol that adds mirroring: see Section 4.3, Higher-Level Protocols, especially HL6. Except in such cases, mirroring must be done according to rule L4, to ensure that the correct character code is used to express the intended semantic of the character, and to avoid interoperability and security problems.

Implementing rule L4 calls for mirrored glyphs. These glyphs may not be exact graphical mirror images. For example, clearly an italic parenthesis is not an exact mirror image of another— “(” is not the mirror image of “)”. Instead, mirror glyphs are those acceptable as mirrors within the normal parameters of the font in which they are represented.

In implementation, sometimes pairs of characters are acceptable mirrors for one another—for example, U+0028 “(” LEFT PARENTHESIS and U+0029 “)” RIGHT PARENTHESIS or U+22E0 “⋠” DOES NOT PRECEDE OR EQUAL and U+22E1 “⋡” DOES NOT SUCCEED OR EQUAL. Other characters such as U+2231 “∱” CLOCKWISE INTEGRAL do not have corresponding characters that can be used for acceptable mirrors. The informative BidiMirroring.txt data file [Data9], lists the paired characters with acceptable mirror glyphs. The formal property name for this data in the Unicode Character Database [UCD] is Bidi_Mirroring_Glyph. A comment in the file indicates where the pairs are “best fit”: they should be acceptable in rendering, although ideally the mirrored glyphs may have somewhat different shapes.

Migration Issues

There are two major enhancements in the Unicode 6.3 version of the UBA:

• Directional isolates
• Bracket Pairs

Implementations of the new directional isolates should see very few compatibility issues; the UBA has been carefully modified to minimize differences for older text written without them. There are a few edge cases near the limit of the number of levels where there are some differences, but those are not likely to be encountered in practice.

With bracket pairs, there may be more changes. The problem is that without knowing (or having good UI access to) the directional marks or embeddings, people have constructed text with the correct visual appearance but incorrect underlying structure (eg …[…[…, appearing as …[…]…). The new algorithm catches cases like these, because such malformed sequences of brackets are not matched.

However, there are some cases where older implementations without rule N0 produced the desired appearance, and newer implementations will not. The user feedback on implementations was sufficiently positive that the decision was made to add N0.

There are also incompatibilities from some implementation's failing to updating correctly to previous versions of Unicode, notably in the mishandling solidus such that "T 1/2" (T is an Arabic character) appears incorrectly as "2/1 T".

To mitigate compatibility problems, it is strongly recommended that implementations take the following steps:

• Add appropriate directional formatting characters on both any parentheses that are resolved with rule N0 so that they appear properly on older systems. This can be done with directional marks (RLM or LRM) on both sides of each parenthesis. For forward compatibility, text authored on older systems should use semantically correct brackets (with directional formatting characters as necessary) to ensure correct display on systems with implementations after Unicode 6.3.
• Add the appropriate explicit embedding around any sequence of numbers + solidus + numbers.

Section Reorganization

In Unicode 6.3, there was significant reorganization of the text. The following table shows the new and old section numbers.

Acknowledgments

Mark Davis created the initial version of this annex and maintains the text. Aharon Lanin and Andrew Glass made substantial additions to Revision 29 (Unicode 6.3.0).

Thanks to the following people for their contributions to the Bidirectional Algorithm or for their feedback on earlier versions of this annex: Ahmed Talaat (أحمد طلعت), Alaa Ghoneim (علاء غنيم), Ahmed Talaat (أحمد طلعت), Asmus Freytag, Avery Bishop, Ayman Aldahleh (أيمن الدحلة), Behdad Esfahbod (بهداد اسفهبد), Doug Felt, Dwayne Robinson, Eric Mader, Ernest Cline, Gidi Shalom-Bendor (גידי שלום-בן דור), Gilead Almosnino (גלעד אלמוסנינו), Isai Scheinberg, Israel Gidali (ישראל גידלי), Joe Becker, John McConnell, Jonathan Kew, Jonathan Rosenne (יונתן רוזן), Khaled Sherif (خالد شريف), Kamal Mansour (كمال منصور), Kenneth Whistler, Khaled Sherif (خالد شريف), Laurențiu Iancu, Maha Hassan (مها حسن), Markus Scherer, Martin Dürst, Mati Allouche (מתתיהו אלוש), Michel Suignard, Mike Ksar (ميشيل قصار), Murray Sargent, Paul Nelson, Peter Constable, Rick McGowan, Robert Steen, Roozbeh Pournader (روزبه پورنادر), Steve Atkin, and Thomas Milo (تُومَاسْ مِيلُو).

References

For references for this annex, see Unicode Standard Annex #41, “Common References for Unicode Standard Annexes.”

Modifications

The following summarizes modifications from previous revisions of this annex.

Revision 28

• Draft 14
• Merge sections 5.3, 6.6 into Migration.
• Draft 13
• Added support for canonical equivalents in BD16.
• Changed logic in N0 to not check forwards for context in the case of enclosed text opposite the embedding direction.
• Draft 12
• Increased the max_depth to 125 in BD2.
• Removed CSS table from Section 2.7 Markup and Formatting Characters, fixed FSI.
• Minor rewording in Section 6.6  Maximizing Compatibility for Paired Brackets
• Added subsection of Migration: Section Reorganization
• Draft 11
• Moved 5.3 to a new Migration section, removing text specific to Unicode 3.0, and adding a table of old and new section numbers.
• Added Section 2.7 Markup and Formatting Characters.
• Previous Drafts
• Major extension of the algorithm to allow for the implementation of directional isolates and the introduction of new isolate-related values to the Bidi_Class property.
• Adds BD8, BD9, BD10, BD11, BD12, BD13, BD14, BD15, and BD16, Sections 2.4 and 2.5, and Rules X5a, X5b, X5c and X6a.
• Extensively revises Section 3.3.2, Explicit Levels and Directions and its existing X rules to formalize the algorithm for matching a PDF with the embedding or override initiator whose scope it terminates.
• Moves Rules X9 and X10 into a separate new Section 3.3.3, Preparations for Implicit Processing.
• Modifies Rule X10 to make the isolating run sequence the unit to which subsequent rules are applied.
• Invalidates an RLE or RLO nested inside the scope of an LRE or LRO nested inside embedding level 60 (where that LRE or LRO was already invalid). This was required to allow nesting embeddings and overrides in isolates and vice versa.
• Modifies Rule W1 to change an NSM preceded by an isolate initiator or PDI into ON.
• Adds Rule N0 and makes other changes to Section 3.3.5, Resolving Neutral and Isolate Formatting Types to resolve bracket pairs to the same level.
• Adds the new ARABIC LETTER MARK (U+061C) character to Section 2.6, Implicit Directional Marks and Table 4 Bidirectional Character Types.
• Changes the examples after L2 to use isolates instead of embeddings.
• Moves usage-related sub-sections of Section 5, Implementation Notes to a new Section 6, Usage, and renumbers Mirroring to 7.
• Renames old Section 5.5, Usage to 6.3, Formatting and adds to it suggestions for and examples of directional isolate usage.
• Adds Sections 5.4 and 6.6 with guidelines on maximizing compatibility for bracket pairs.
• Proposed Update for Unicode 6.3.0.

Revision 27

• Reissued for Unicode 6.2.0.

Revision 26 being a proposed update, only changes between versions 23 and 27 are noted here.

Revision 25

• Reissued for Unicode 6.1.0.

Revision 24 being a proposed update, only changes between versions 23 and 25 are noted here.

Revision 23

• Reissued for Unicode 6.0.0.
• Added anchors on tables.
• Added text to clarify HL1, and clarified statement in P3.
• Added links on rules.
• Added section heading for 5.3 Migrating from 2.0 to 3.0
• Moved text from the end of 4.3 Higher-Level Protocols to a new section 5.8 Conversion to Plain Text
• Rephrased the relationship between clauses HL1 and HL3 and HL4 and HL5

Revision 22 being a proposed update, only changes between versions 21 and 23 are noted here.

Revision 21

• Reissued for Unicode 5.2.0.
• Added Section 4.4 Bidirectional Conformance Testing.
• Added BN to Rule X6 (removing certain characters).
• Clarified examples in Rule N1 (affecting characters next to EN or AN characters).
• Added to HL6 the clause: Those with a resolved directionality of L and whose bidi classbidirectional type is R or AL.
• Clarified the text at the start of 3  Basic Display Algorithm.
• Added Bidi_Class and Bidi_Mirroring_Glyph property names.
• Added clarifications to 3.3.2  Explicit Levels and Directions, to X6, and to N2.
• Fixed typos in 3.4  Reordering Resolved Levels.
• Added links on items in Table 4, and clarified BN there.
• Removed a note in N1.

Revision 20 being a proposed update, only changes between versions 19 and 21 are noted here.

Revision 19

• Updated for Version 5.1.0.
• Clarified BD6.
• Made some examples more explicit.
• Added the common problem case of separators in Section 6.3  Formatting.
• Added notes on security concerns with RLO and LRO, and the use of dir="ltr" or "rtl" with web pages.
• Fixed example under N2.
• Fixed example in Section 3.3.4  Resolving Neutral Types
• Made last part of Section 6.3  Formatting into a new Section 6.4  Separating Punctuation Marks, changed the term "Separators" and added a note on stateful formatting characters.

Revision 18 being a proposed update, only changes between versions 19 and 17 are noted here.

Revision 17

• This revision incorporates successive changes. The latest changes were on 2006-02-24.
• Modified L4 and HL6, in conjunction with proposed property change to Bidi_Mirrored (PRI #80)
• Added note on U+ FD3E (﴾) ORNATE LEFT PARENTHESIS and U+ FD3F (﴿) ORNATE RIGHT PARENTHESIS.
• Used new format for conformance clause numbering.
• Added caution on use of higher-level protocols, after UAX9-C2.
• Some wording changes in 7  Mirroring, for consistency with new L4 and HL6.
• Moved text to Shaping and Line Breaking, and added note on SHY in 3.4 there
• Removed two notes indicating that the conformance clauses override clause C13 of Unicode 4.0.
• Changed some references to Unicode4.0

Revision 16 being a proposed update, only changes between versions 17 and 15 are noted here.

Revision 15:

• Minor editing
• Fixed section Number for Mirroring
• Changed “Tracking Number” to Revision
• Added note on U+0CBF KANNADA VOWEL SIGN I
• Added note after N1, and clarified example after N2.
• Fixed references to sections of the Unicode Standard

Revision 14:

• Aliased directional run and level run
• Pointed to DerivedBidiClass.txt for unassigned character assignments.

Revision 13:

• 4. Bidirectional Conformance: added explicit clauses.
• 4.3. Higher-Level Protocols:
  • Added clarifying text, and renumbered options.
  • Removed option regarding number shaping (because it was irrelevant to bidirectional ordering).
  • Broadened the ability to override on the basis of context, and clarified number handling.
  • Made clear that bidiUBA could be applied to segments
• 1. Introduction: added note that the changes in 4. Bidirectional Conformance override clause C13 of Unicode 4.0 [Unicode], and tighten the conformance requirements from what they had been previously.
• Minor editing for clarification.

Revision 11:

• Updated for Unicode 4.0.
• Added note on canonical equivalence
• Added Joiners section on ZWJ and ZWNJ
• Clarified L2 and examples following.
• Added a section on the interaction of shaping and bidirectional reordering.
• Moved lists for unassigned characters into UCD.html (also now explicit in DerivedBidiClass.txt)
• Updated references for Newline Guidelines (because the UAX is incorporated into the 4.0 book)
• The first two sections were rearranged, with Reference Code going into Implementation Notes, and Mirroring in its own section at the end.
  • This is not highlighted in the proposed text.
• Sections were renumbered and the table of contents is more detailed.
  • This is not highlighted in the proposed text.
• Misc. editing.

Revision 10:

• Updated for Unicode 3.2.
• Updated UAX boilerplate in the status section.

Revision 9:

• Clarified the language of P2
• Corrected the implementation note on “Retaining Explicit Formatting Characters” in Implementation Notes
• Minor editing

Copyright © 2000-2013 Unicode, Inc. All Rights Reserved. The Unicode Consortium makes no expressed or implied warranty of any kind, and assumes no liability for errors or omissions. No liability is assumed for incidental and consequential damages in connection with or arising out of the use of the information or programs contained or accompanying this technical report. The Unicode Terms of Use apply.

Unicode and the Unicode logo are trademarks of Unicode, Inc., and are registered in some jurisdictions.