The Glyph Positioning Table

Figure 4a. Vietnamese words with marks.

Other writing systems require sophisticated glyph positioning for correct typographic composition. For instance, Urdu glyphs are calligraphic and connect to one another along a descending, diagonal text line that proceeds from right to left. To properly render Urdu, a text-processing client must modify both the horizontal (X) and vertical (Y) positions of each glyph.

Figure 4b. Urdu layout requires glyph positioning control, as well as contextual substitution

With the GPOS table, a font developer can define a complete set of positioning adjustment features in an OpenType™ font. GPOS data, organized by script and language system, is easy for a text-processing client to use to position glyphs.

Positioning Glyphs with TrueType 1.0

Glyph positioning in TrueType uses only two values, placement and advance, to specify a glyph's position for text layout. If glyphs are positioned with respect to a virtual "pen point" that moves along a line of text, placement describes the glyph's position with respect to the current pen point, and advance describes where to move the pen point to position the next glyph (see Figure 4c). For horizontal text, placement corresponds to the left side bearing, and advance corresponds to the advance width.

Figure 4c. Glyph positioning with TrueType

TrueType specifies placement and advance only in the X direction for horizontal layout and only in the Y direction for vertical layout. For simple Latin text layout, these two values may be adequate to position glyphs correctly. But, for texts that require more sophisticated layout, the values must cover a richer range. Placement and advance may need adjustment vertically, as well as horizontally.

The only positioning adjustment defined in TrueType is pair kerning, which modifies the horizontal spacing between two glyphs. A typical kerning table lists pairs of glyphs and specifies how much space a text-processing client should add or remove between the glyphs to properly display each pair. It does not provide specific information about how to adjust the glyphs in each pair, and cannot adjust contexts of more than two glyphs.

Positioning Glyphs with OpenType

OpenType fonts allow excellent control and flexibility for positioning a single glyph and for positioning multiple glyphs in relation to one another. By using both X and Y values that the GPOS table defines for placement and advance and by using glyph attachment points, a client can more precisely adjust the position of a glyph.

In addition, the GPOS table can reference a Device table to define subtle, device-dependent adjustments to any placement or advance value at any font size and device resolution. For example, a Device table can specify adjustments at 51 pixels per em (ppem) that do not occur at 50 ppem.

X and Y values specified in OpenType fonts for placement operations are always within the typical Cartesian coordinate system (origin at the lower left), regardless of the writing direction. Additionally, all values specified are done so in font unit measurements. This is especially convenient for font designers, since glyphs are drawn in the same coordinate system. However, it's important to note that the meaning of "advance width" changes, depending on the writing direction.

For example, in left-to-right scripts, if the first glyph has an advance width of 100, then the second glyph begins at 100,0. In right-to-left scripts, if the first glyph has an advance width of 100, then the second glyph begins at -100,0. For a top-to-bottom feature, to increase the advance height of a glyph by 100, the YAdvance = 100. For any feature, regardless of writing direction, to lower the dieresis over an 'o' by 10 units, set the YPlacement = -10.

Other GPOS features can define attachment points to combine glyphs and position them with respect to one another. A glyph might have multiple attachment points. The point used will depend on the glyph to be attached. For instance, a base glyph could have attachment points for different diacritical marks.

Base glyph with multiple attachment points.

To reduce the size of the font file, a base glyph may use the same attachment point for all mark glyphs assigned to a particular class. For example, a base glyph could have two attachment points, one above and one below the glyph. Then all marks that attach above glyphs would be attached at the high point, and all marks that attach below glyphs would be attached at the low point. Attachment points are useful in scripts, such as Arabic, that combine numerous glyphs with vowel marks.

Attachment points also are useful for connecting cursive-style glyphs. Glyphs in cursive fonts can be designed to attach or overlap when rendered. Alternatively, the font developer can use OpenType to create a cursive attachment feature and define explicit exit and entry attachment points for each glyph (see Figure 4d).

Figure 4d. Entry and exit points marked on contextual Urdu glyph variations

The GPOS table supports eight types of actions for positioning and attaching glyphs:

• A single adjustment positions one glyph, such as a superscript or subscript.
• A pair adjustment positions two glyphs with respect to one another. Kerning is an example of pair adjustment.
• A cursive attachment describes cursive scripts and other glyphs that are connected with attachment points when rendered.
• A MarkToBase attachment positions combining marks with respect to base glyphs, as when positioning vowels, diacritical marks, or tone marks in Arabic, Hebrew, and Vietnamese.
• A MarkToLigature attachment positions combining marks with respect to ligature glyphs. Because ligatures may have multiple points for attaching marks, the font developer needs to associate each mark with one of the ligature glyph's components.
• A MarkToMark attachment positions one mark relative to another, as when positioning tone marks with respect to vowel diacritical marks in Vietnamese.
• Contextual positioning describes how to position one or more glyphs in context, within an identifiable sequence of specific glyphs, glyph classes, or varied sets of glyphs. One or more positioning operations may be performed on "input" context sequences. Figure 4e illustrates a context for positioning adjustments.
  
• Chaining Contextual positioning describes how to position one or more glyphs in a chained context, within an identifiable sequence of specific glyphs, glyph classes, or varied sets of glyphs. One or more positioning operations may be performed on "input" context sequences.
  

Figure 4e. Contextual positioning lowered the accent over a vowel glyph that followed an overhanging uppercase glyph

Table Organization

The GPOS table begins with a header that defines offsets to a ScriptList, a FeatureList, and a LookupList (see Figure 4f):

• The ScriptList identifies all the scripts and language systems in the font that use glyph positioning.
• The FeatureList defines all the glyph positioning features required to render these scripts and language systems.
• The LookupList contains all the lookup data needed to implement each glyph positioning feature.

For a detailed discussion of ScriptLists, FeatureLists, and LookupLists, see the chapter, OpenType Common Table Formats. The following discussion summarizes how the GPOS table works.

Figure 4f. High-level organization of GPOS table

The GPOS table is organized so text processing clients can easily locate the features and lookups that apply to a particular script or language system. To access GPOS information, clients should use the following procedure:

1. Locate the current script in the GPOS ScriptList table.
2. If the language system is known, search the script for the correct LangSys table; otherwise, use the script's default language system (DefaultLangSys table).
3. The LangSys table provides index numbers into the GPOS FeatureList table to access a required feature and a number of additional features.
4. Inspect the FeatureTag of each feature, and select the features to apply to an input glyph string.
5. Each feature provides an array of index numbers into the GPOS LookupList table. Lookup data is defined in one or more subtables that contain information about specific glyphs and the kinds of operations to be performed on them.
6. Assemble all lookups from the set of chosen features, and apply the lookups in the order given in the LookupList table.

A lookup uses subtables to define the specific conditions, type, and results of a positioning action used to implement a feature. All subtables in a lookup must be of the same LookupType, as listed in the LookupType Enumeration table:

LookupType Enumeration table for glyph positioning

Each LookupType is defined by one or more subtables, whose format depends on the type of positioning operation and the resulting storage efficiency. When glyph information is best presented in more than one format, a single lookup may define more than one subtable, as long as all the subtables are of the same LookupType. For example, within a given lookup, a glyph index array format may best represent one set of target glyphs, whereas a glyph index range format may be better for another set.

A series of positioning operations on the same glyph or string requires multiple lookups, one for each separate action. The values in the ValueRecords are accumulated in these cases. Each lookup is given a different array number in the LookupList table and is applied in the LookupList order.

During text processing, a client applies a lookup to each glyph in the string before moving to the next lookup. A lookup is finished for a glyph after the client locates the target glyph or glyph context and performs a positioning, if specified. To move to the "next" glyph, the client will typically skip all the glyphs that participated in the lookup operation: glyphs that were positioned as well as any other glyphs that formed a context for the operation.

There is just one exception: the "next" glyph in a sequence may be one of those that formed a context for the operation just performed. For example, in the case of pair positioning operations (i.e., kerning), if the position value record for the second glyph is null, that glyph is treated as the "next" glyph in the sequence.

This rest of this chapter describes the GPOS header and the subtables defined for each LookupType. Several GPOS subtables share other tables: ValueRecords, Anchor tables, and MarkArrays. For easy reference, the shared tables are described at the end of this chapter.

GPOS Header

The GPOS table begins with a header that contains a version number (Version) initially set to 1.0 (0x00010000) and offsets to three tables: ScriptList, FeatureList, and LookupList. For descriptions of these tables, see the chapter, OpenType Common Table Formats. Example 1 at the end of this chapter shows a GPOS Header table definition.

GPOS Header

Lookup Type 1: Single Adjustment Positioning Subtable

A single adjustment positioning subtable (SinglePos) is used to adjust the position of a single glyph, such as a subscript or superscript. In addition, a SinglePos subtable is commonly used to implement lookup data for contextual positioning.

A SinglePos subtable will have one of two formats: one that applies the same adjustment to a series of glyphs, or one that applies a different adjustment for each unique glyph.

Single Adjustment Positioning: Format 1

A SinglePosFormat1 subtable applies the same positioning value or values to each glyph listed in its Coverage table. For instance, when a font uses old-style numerals, this format could be applied to uniformly lower the position of all math operator glyphs.

The Format 1 subtable consists of a format identifier (PosFormat), an offset to a Coverage table that defines the glyphs to be adjusted by the positioning values (Coverage), and the format identifier (ValueFormat) that describes the amount and kinds of data in the ValueRecord.

The ValueRecord specifies one or more positioning values to be applied to all covered glyphs (Value). For example, if all glyphs in the Coverage table require both horizontal and vertical adjustments, the ValueRecord will specify values for both XPlacement and Yplacement.

Example 2 at the end of this chapter shows a SinglePosFormat1 subtable used to adjust the placement of subscript glyphs.

SinglePosFormat1 subtable: Single positioning value

Single Adjustment Positioning: Format 2

A SinglePosFormat2 subtable provides an array of ValueRecords that contains one positioning value for each glyph in the Coverage table. This format is more flexible than Format 1, but it requires more space in the font file.

For example, assume that the Cyrillic script will be used in left-justified text. For all glyphs, Format 2 could define position adjustments for left side bearings to align the left edges of the paragraphs. To achieve this, the Coverage table would list every glyph in the script, and the SinglePosFormat2 subtable would define a ValueRecord for each covered glyph. Correspondingly, each ValueRecord would specify an XPlacement adjustment value for the left side bearing.

Note: All ValueRecords defined in a SinglePos subtable must have the same ValueFormat. In this example, if XPlacement is the only value that a ValueRecord needs to optically align the glyphs, then XPlacement will be the only value specified in the ValueFormat of the subtable.

As in Format 1, the Format 2 subtable consists of a format identifier (PosFormat), an offset to a Coverage table that defines the glyphs to be adjusted by the positioning values (Coverage), and the format identifier (ValueFormat) that describes the amount and kinds of data in the ValueRecords. In addition, the Format 2 subtable includes:

• A count of the ValueRecords (ValueCount). One ValueRecord is defined for each glyph in the Coverage table.
• An array of ValueRecords that specify positioning values (Value). Because the array follows the Coverage Index order, the first ValueRecord applies to the first glyph listed in the Coverage table, and so on.

Example 3 at the end of this chapter shows how to adjust the spacing of three dash glyphs with a SinglePosFormat2 subtable.

SinglePosFormat2 subtable: Array of positioning values

Lookup Type 2:
Pair Adjustment Positioning Subtable

A pair adjustment positioning subtable (PairPos) is used to adjust the positions of two glyphs in relation to one another-for instance, to specify kerning data for pairs of glyphs. Compared to a typical kerning table, however, a PairPos subtable offers more flexiblity and precise control over glyph positioning. The PairPos subtable can adjust each glyph in a pair independently in both the X and Y directions, and it can explicitly describe the particular type of adjustment applied to each glyph. In addition, a PairPos subtable can use Device tables to subtly adjust glyph positions at each font size and device resolution.

PairPos subtables can be either of two formats: one that identifies glyphs individually by index (Format 1), or one that identifies glyphs by class (Format 2).

Pair Positioning Adjustment: Format 1

Format 1 uses glyph indices to access positioning data for one or more specific pairs of glyphs. All pairs are specified in the order determined by the layout direction of the text.

Note: For text written from right to left, the right-most glyph will be the first glyph in a pair; conversely, for text written from left to right, the left-most glyph will be first.

A PairPosFormat1 subtable contains a format identifier (PosFormat) and two ValueFormats:

• ValueFormat1 applies to the ValueRecord of the first glyph in each pair. ValueRecords for all first glyphs must use ValueFormat1. If ValueFormat1 is set to zero (0), the corresponding glyph has no ValueRecord and, therefore, should not be repositioned.
• ValueFormat2 applies to the ValueRecord of the second glyph in each pair. ValueRecords for all second glyphs must use ValueFormat2. If ValueFormat2 is set to null, then the second glyph of the pair is the "next" glyph for which a lookup should be performed.

A PairPos subtable also defines an offset to a Coverage table (Coverage) that lists the indices of the first glyphs in each pair. More than one pair can have the same first glyph, but the Coverage table will list that glyph only once.

The subtable also contains an array of offsets to PairSet tables (PairSet) and a count of the defined tables (PairSetCount). The PairSet array contains one offset for each glyph listed in the Coverage table and uses the same order as the Coverage Index.

PairPosFormat1 subtable: Adjustments for glyph pairs

A PairSet table enumerates all the glyph pairs that begin with a covered glyph. An array of PairValueRecords (PairValueRecord) contains one record for each pair and lists the records sorted by the GlyphID of the second glyph in each pair. PairValueCount specifies the number of PairValueRecords in the set.

PairSet table

A PairValueRecord specifies the second glyph in a pair (SecondGlyph) and defines a ValueRecord for each glyph (Value1 and Value2). If ValueFormat1 is set to zero (0) in the PairPos subtable, ValueRecord1 will be empty; similarly, if ValueFormat2 is 0, Value2 will be empty.

Example 4 at the end of this chapter shows a PairPosFormat1 subtable that defines two cases of pair kerning.

PairValueRecord

Pair Positioning Adjustment: Format 2

Format 2 defines a pair as a set of two glyph classes and modifies the positions of all the glyphs in a class. For example, this format is useful in Japanese scripts that apply specific kerning operations to all glyph pairs that contain punctuation glyphs. One class would be defined as all glyphs that may be coupled with punctuation marks, and the other classes would be groups of similar punctuation glyphs.

The PairPos Format2 subtable begins with a format identifier (PosFormat) and an offset to a Coverage table (Coverage), measured from the beginning of the PairPos subtable. The Coverage table lists the indices of the first glyphs that may appear in each glyph pair. More than one pair may begin with the same glyph, but the Coverage table lists the glyph index only once.

A PairPosFormat2 subtable also includes two ValueFormats:

• ValueFormat1 applies to the ValueRecord of the first glyph in each pair. ValueRecords for all first glyphs must use ValueFormat1. If ValueFormat1 is set to zero (0), the corresponding glyph has no ValueRecord and, therefore, should not be repositioned.
• ValueFormat2 applies to the ValueRecord of the second glyph in each pair. ValueRecords for all second glyphs must use ValueFormat2. If ValueFormat2 is set to null, then the second glyph of the pair is the "next" glyph for which a lookup should be performed.

PairPosFormat2 requires that each glyph in all pairs be assigned to a class, which is identified by an integer called a class value. (For details about classes, see the chapter, Common Table Formats.) Pairs are then represented in a two-dimensional array as sequences of two class values. Multiple pairs can be represented in one Format 2 subtable.

A PairPosFormat2 subtable contains offsets to two class definition tables: one that assigns class values to all the first glyphs in all pairs (ClassDef1), and one that assigns class values to all the second glyphs in all pairs (ClassDef2). If both glyphs in a pair use the same class definition, the offset value will be the same for ClassDef1 and ClassDef2. The subtable also specifies the number of glyph classes defined in ClassDef1 (Class1Count) and in ClassDef2 (Class2Count), including Class0.

For each class identified in the ClassDef1 table, a Class1Record enumerates all pairs that contain a particular class as a first component. The Class1Record array stores all Class1Records according to class value.

Note: Class1Records are not tagged with a class value identifier. Instead, the index value of a Class1Record in the array defines the class value represented by the record. For example, the first Class1Record enumerates pairs that begin with a Class 0 glyph, the second Class1Record enumerates pairs that begin with a Class1 glyph, and so on.

PairPosFormat2 subtable: Class pair adjustment

Each Class1Record contains an array of Class2Records (Class2Record), which also are ordered by class value. One Class2Record must be declared for each class in the ClassDef2 table, including Class 0.

Class1Record

A Class2Record consists of two ValueRecords, one for the first glyph in a class pair (Value1) and one for the second glyph (Value2). If the PairPos subtable has a value of zero (0) for ValueFormat1 or ValueFormat2, the corresponding record (ValueRecord1 or ValueRecord2) will be empty.

Example 5 at the end of this chapter demonstrates pair kerning with glyph classes in a PairPosFormat2 subtable.

Class2Record

Lookup Type 3:
Cursive Attachment Positioning Subtable

Some cursive fonts are designed so that adjacent glyphs join when rendered with their default positioning. However, if positioning adjustments are needed to join the glyphs, a cursive attachment positioning (CursivePos) subtable can describe how to connect the glyphs by aligning two anchor points: the designated exit point of a glyph, and the designated entry point of the following glyph.

The subtable has one format: CursivePosFormat1. It begins with a format identifier (PosFormat) and an offset to a Coverage table (Coverage), which lists all the glyphs that define cursive attachment data.

In addition, the subtable contains one EntryExitRecord for each glyph listed in the Coverage table, a count of those records (EntryExitCount), and an array of those records in the same order as the Coverage Index (EntryExitRecord).

CursivePosFormat1 subtable: Cursive attachment

Each EntryExitRecord consists of two offsets: one to an Anchor table that identifies the entry point on the glyph (EntryAnchor), and an offset to an Anchor table that identifies the exit point on the glyph (ExitAnchor). (For a complete description of the Anchor table, see the end of this chapter.)

To position glyphs using the CursivePosFormat1 subtable, a text-processing client aligns the ExitAnchor point of a glyph with the EntryAnchor point of the following glyph. If no corresponding anchor point exists, either the EntryAnchor or ExitAnchor offset may be NULL.

At the end of this chapter, Example 6 describes cursive glyph attachment in the Urdu language.

EntryExitRecord

Lookup Type 4:
MarkToBase Attachment Positioning Subtable

The MarkToBase attachment (MarkBasePos) subtable is used to position combining mark glyphs with respect to base glyphs. For example, the Arabic, Hebrew, and Thai scripts combine vowels, diacritical marks, and tone marks with base glyphs.

In the MarkBasePos subtable, every mark glyph has an anchor point and is associated with a class of marks. Each base glyph then defines an anchor point for each class of marks it uses.

For example, assume two mark classes: all marks positioned above base glyphs (Class 0), and all marks positioned below base glyphs (Class 1). In this case, each base glyph that uses these marks would define two anchor points, one for attaching the mark glyphs listed in Class 0, and one for attaching the mark glyphs listed in Class 1.

To identify the base glyph that combines with a mark, the text-processing client must look backward in the glyph string from the mark to the preceding base glyph. To combine the mark and base glyph, the client aligns their attachment points, positioning the mark with respect to the final pen point (advance) position of the base glyph.

The MarkToBase Attachment subtable has one format: MarkBasePosFormat1. The subtable begins with a format identifier (PosFormat) and offsets to two Coverage tables: one that lists all the mark glyphs referenced in the subtable (MarkCoverage), and one that lists all the base glyphs referenced in the subtable (BaseCoverage).

For each mark glyph in the MarkCoverage table, a record specifies its class and an offset to the Anchor table that describes the mark's attachment point (MarkRecord). A mark class is identified by a specific integer, called a class value. ClassCount specifies the total number of distinct mark classes defined in all the MarkRecords.

The MarkBasePosFormat1 subtable also contains an offset to a MarkArray table, which contains all the MarkRecords stored in an array (MarkRecord) by MarkCoverage Index. A MarkArray table also contains a count of the defined MarkRecords (MarkCount). (For details about MarkArrays and MarkRecords, see the end of this chapter.)

The MarkBasePosFormat1 subtable also contains an offset to a BaseArray table (BaseArray).

MarkBasePosFormat1 subtable: MarkToBase attachment point

The BaseArray table consists of an array (BaseRecord) and count (BaseCount) of BaseRecords. The array stores the BaseRecords in the same order as the BaseCoverage Index. Each base glyph in the BaseCoverage table has a BaseRecord.

BaseArray table

A BaseRecord declares one Anchor table for each mark class (including Class 0) identified in the MarkRecords of the MarkArray. Each Anchor table specifies one attachment point used to attach all the marks in a particular class to the base glyph. A BaseRecord contains an array of offsets to Anchor tables (BaseAnchor). The zero-based array of offsets defines the entire set of attachment points each base glyph uses to attach marks. The offsets to Anchor tables are ordered by mark class.

Note: Anchor tables are not tagged with class value identifiers. Instead, the index value of an Anchor table in the array defines the class value represented by the Anchor table.

Example 7 at the end of this chapter defines mark positioning above and below base glyphs with a MarkBasePosFormat1 subtable.

BaseRecord

Lookup Type 5:
MarkToLigature Attachment Positioning Subtable

The MarkToLigature attachment (MarkLigPos) subtable is used to position combining mark glyphs with respect to ligature base glyphs. With MarkToBase attachment, described previously, a single base glyph defines an attachment point for each class of marks. In contrast, MarkToLigature attachment describes ligature glyphs composed of several components that can each define an attachment point for each class of marks.

As a result, a ligature glyph may have multiple base attachment points for one class of marks. The specific attachment point for a mark is defined by the ligature component that the subtable associates with the mark.

The MarkLigPos subtable can be used to define multiple mark-to-ligature attachments. In the subtable, every mark glyph has an anchor point and is associated with a class of marks. Every ligature glyph specifies a two-dimensional array of data: each component in a ligature defines an array of anchor points, one for each class of marks.

For example, assume two mark classes: all marks positioned above base glyphs (Class 0), and all marks positioned below base glyphs (Class 1). In this case, each component of a base ligature glyph may define two anchor points, one for attaching the mark glyphs listed in Class 0, and one for attaching the mark glyphs listed in Class 1. Alternatively, if the language system does not allow marks on the second component, the first ligature component may define two anchor points, one for each class of marks, and the second ligature component may define no anchor points.

To position a combining mark using a MarkToLigature attachment subtable, the text-processing client must work backward from the mark to the preceding ligature glyph. To correctly access the subtables, the client must keep track of the component associated with the mark. Aligning the attachment points combines the mark and ligature.

The MarkToLigature attachment subtable has one format: MarkLigPosFormat1. The subtable begins with a format identifier (PosFormat) and offsets to two Coverage tables that list all the mark glyphs (MarkCoverage) and Ligature glyphs (LigatureCoverage) referenced in the subtable.

For each glyph in the MarkCoverage table, a MarkRecord specifies its class and an offset to the Anchor table that describes the mark's attachment point. A mark class is identified by a specific integer, called a class value. ClassCount records the total number of distinct mark classes defined in all MarkRecords.

The MarkBasePosFormat1 subtable contains an offset, measured from the beginning of the subtable, to a MarkArray table, which contains all MarkRecords stored in an array (MarkRecord) by MarkCoverage Index. (For details about MarkArrays and MarkRecords, see the end of this chapter.)

The MarkLigPosFormat1 subtable also contains an offset to a LigatureArray table (LigatureArray).

MarkLigPosFormat1 subtable: MarkToLigature attachment

The LigatureArray table contains a count (LigatureCount) and an array of offsets (LigatureAttach) to LigatureAttach tables. The LigatureAttach array lists the offsets to

LigatureAttach tables, one for each ligature glyph listed in the LigatureCoverage table, in the same order as the LigatureCoverage Index.

LigatureArray table

Each LigatureAttach table consists of an array (ComponentRecord) and count (ComponentCount) of the component glyphs in a ligature. The array stores the ComponentRecords in the same order as the components in the ligature. The order of the records also corresponds to the writing direction of the text. For text written left to right, the first component is on the left; for text written right to left, the first component is on the right.

LigatureAttach table

A ComponentRecord, one for each component in the ligature, contains an array of offsets to the Anchor tables that define all the attachment points used to attach marks to the component (LigatureAnchor). For each mark class (including Class 0) identified in the MarkArray records, an Anchor table specifies the point used to attach all the marks in a particular class to the ligature base glyph, relative to the component.

In a ComponentRecord, the zero-based LigatureAnchor array lists offsets to Anchor tables by mark class. If a component does not define an attachment point for a particular class of marks, then the offset to the corresponding Anchor table will be NULL.

Example 8 at the end of this chapter shows a MarkLisPosFormat1 subtable used to attach mark accents to a ligature glyph in the Arabic script.

ComponentRecord

Lookup Type 6:
MarkToMark Attachment Positioning Subtable

The MarkToMark attachment (MarkMarkPos) subtable is identical in form to the MarkToBase attachment subtable, although its function is different. MarkToMark attachment defines the position of one mark relative to another mark as when, for example, positioning tone marks with respect to vowel diacritical marks in Vietnamese.

The attaching mark is Mark1, and the base mark being attached to is Mark2. In the MarkMarkPos subtable, every Mark1 glyph has an anchor attachment point and is associated with a class of marks. Each Mark2 glyph defines an anchor point for each class of marks. For example, assume two Mark1 classes: all marks positioned to the left of Mark2 glyphs (Class 0), and all marks positioned to the right of Mark2 glyphs (Class 1). Each Mark2 glyph that uses these marks defines two anchor points: one for attaching the Mark1 glyphs listed in Class 0, and one for attaching the Mark1 glyphs listed in Class 1.

The Mark2 glyph that combines with a Mark1 glyph is the glyph preceding the Mark1 glyph in glyph string order (skipping glyphs according to LookupFlags). The subtable applies precisely when that Mark2 glyph is covered by Mark2Coverage. To combine the mark glyphs, the Mark1 glyph is moved such that the relevant attachment points coincide. The input context for MarkToBase, MarkToLigature and MarkToMark positioning tables is the mark that is being positioned. If a sequence contains several marks, a lookup may act on it several times, to position them.

The MarkToMark attachment subtable has one format: MarkMarkPosFormat1. The subtable begins with a format identifier (PosFormat) and offsets to two Coverage tables: one that lists all the Mark1 glyphs referenced in the subtable (Mark1Coverage), and one that lists all the Mark2 glyphs referenced in the subtable (Mark2Coverage).

For each mark glyph in the Mark1Coverage table, a MarkRecord specifies its class and an offset to the Anchor table that describes the mark's attachment point. A mark class is identified by a specific integer, called a class value. (For details about classes, see the chapter, OpenType Common Table Formats.) ClassCount specifies the total number of distinct mark classes defined in all the MarkRecords.

The MarkMarkPosFormat1 subtable also contains two offsets, measured from the beginning of the subtable, to two arrays:

• The MarkArray table contains all MarkRecords stored by Mark1Coverage Index in an array (MarkRecord). The MarkArray table also contains a count of the number of defined MarkRecords (MarkCount).
• The Mark2Array table consists of an array (Mark2Record) and count (Mark2Count) of Mark2Records.

For details about MarkArrays and MarkRecords, see the end of this chapter.

MarkMarkPosFormat1 subtable: MarkToMark attachment

The Mark2Array, shown next, contains one Mark2Record for each Mark2 glyph listed in the Mark2Coverage table. It stores the records in the same order as the Mark2Coverage Index.

Mark2Array table

Each Mark2Record contains an array of offsets to Anchor tables (Mark2Anchor). The array of zero-based offsets, measured from the beginning of the Mark2Array table, defines the entire set of Mark2 attachment points used to attach Mark1 glyphs to a specific Mark2 glyph. The Anchor tables in the Mark2Anchor array are ordered by Mark1 class value.

A Mark2Record declares one Anchor table for each mark class (including Class 0) identified in the MarkRecords of the MarkArray. Each Anchor table specifies one Mark2 attachment point used to attach all the Mark1 glyphs in a particular class to the Mark2 glyph.

Example 9 at the end of the chapter shows a MarkMarkPosFormat1 subtable for attaching one mark to another in the Arabic script.

Mark2Record

Lookup Type 7:
Contextual Positioning Subtables

A Contextual Positioning (ContextPos) subtable defines the most powerful type of glyph positioning lookup. It describes glyph positioning in context so a text-processing client can adjust the position of one or more glyphs within a certain pattern of glyphs. Each subtable describes one or more "input" glyph sequences and one or more positioning operations to be performed on that sequence.

ContextPos subtables can have one of three formats, which closely mirror the formats used for contextual glyph substitution. One format applies to specific glyph sequences (Format 1), one defines the context in terms of glyph classes (Format 2), and the third format defines the context in terms of sets of glyphs (Format 3).

All three formats of ContextPos subtables specify positioning data in a PosLookupRecord. A description of that record follows.

PosLookupRecord

All contextual positioning subtables specify the positioning data in a PosLookupRecord. Each record contains a SequenceIndex, which indicates where the positioning operation will occur in the glyph sequence. In addition, a LookupListIndex identifies the lookup to be applied at the glyph position specified by the SequenceIndex.

The order in which lookups are applied to the entire glyph sequence, called the "design order," can be significant, so PosLookupRecord data should be defined accordingly.

The contextual substitution subtables defined in Examples 10, 11, and 12 show PosLookupRecords.

PosLookupRecord

Context Positioning Subtable: Format 1

Format 1 defines the context for a glyph positioning operation as a particular sequence of glyphs. For example, a context could be <To>, <xyzabc>, <!?*#@>, or any other glyph sequence.

Within the context, Format 1 identifies particular glyph positions (not glyph indices) as the targets for specific adjustments. When a text-processing client locates a context in a string of text, it makes the adjustment by applying the lookup data defined for a targeted position at that location.

For example, suppose that accent mark glyphs above lowercase x-height vowel glyphs must be lowered when an overhanging capital letter glyph precedes the vowel. When the client locates this context in the text, the subtable identifies the position of the accent mark and a lookup index. A lookup specifies a positioning action that lowers the accent mark over the vowel so that it does not collide with the overhanging capital.

ContextPosFormat1 defines the context in two places. A Coverage table specifies the first glyph in the input sequence, and a PosRule table identifies the remaining glyphs. To describe the context used in the previous example, the Coverage table lists the glyph index of the first component of the sequence (the overhanging capital), and a PosRule table defines indices for the lowercase x-height vowel glyph and the accent mark.

A single ContextPosFormat1 subtable may define more than one context glyph sequence. If different context sequences begin with the same glyph, then the Coverage table should list the glyph only once because all first glyphs in the table must be unique. For example, if three contexts each start with an "s" and two start with a "t," then the Coverage table will list one "s" and one "t."

For each context, a PosRule table lists all the glyphs, in order, that follow the first glyph. The table also contains an array of PosLookupRecords that specify the positioning lookup data for each glyph position (including the first glyph position) in the context.

All the PosRule tables defining contexts that begin with the same first glyph are grouped together and defined in a PosRuleSet table. For example, the PosRule tables that define the three contexts that begin with an "s" are grouped in one PosRuleSet table, and the PosRule tables that define the two contexts that begin with a "t" are grouped in a second PosRuleSet table. Each unique glyph listed in the Coverage table must have a PosRuleSet table that defines all the PosRule tables for a covered glyph.

To locate a context glyph sequence, the text-processing client searches the Coverage table each time it encounters a new text glyph. If the glyph is covered, the client reads the corresponding PosRuleSet table and examines each PosRule table in the set to determine whether the rest of the context defined there matches the subsequent glyphs in the text. If the context and text string match, the client finds the target glyph position, applies the lookup for that position, and completes the positioning action.

A ContextPosFormat1 subtable contains a format identifier (PosFormat), an offset to a Coverage table (Coverage), a count of the number of PosRuleSets that are defined (PosRuleSetCount), and an array of offsets to the PosRuleSet tables (PosRuleSet). As mentioned, one PosRuleSet table must be defined for each glyph listed in the Coverage table.

In the PosRuleSet array, the PosRuleSet tables are ordered in the Coverage Index order. The first PosRuleSet in the array applies to the first GlyphID listed in the Coverage table, the second PosRuleSet in the array applies to the second GlyphID listed in the Coverage table, and so on.

ContextPosFormat1 subtable: Simple context positioning

A PosRuleSet table consists of an array of offsets to PosRule tables (PosRule), ordered by preference, and a count of the PosRule tables defined in the set (PosRuleCount).

PosRuleSet table: All contexts beginning with the same glyph

A PosRule table consists of a count of the glyphs to be matched in the input context sequence (GlyphCount), including the first glyph in the sequence, and an array of glyph indices that describe the context (Input). The Coverage table specifies the index of the first glyph in the context, and the Input array begins with the second glyph in the context sequence. As a result, the first index position in the array is specified with the number one (1), not zero (0). The Input array lists the indices in the order the corresponding glyphs appear in the text. For text written from right to left, the right-most glyph will be first; conversely, for text written from left to right, the left-most glyph will be first.

A PosRule table also contains a count of the positioning operations to be performed on the input glyph sequence (PosCount) and an array of PosLookupRecords (PosLookupRecord). Each record specifies a position in the input glyph sequence and a LookupList index to the positioning lookup to be applied there. The array should list records in design order, or the order the lookups should be applied to the entire glyph sequence.

Example 10 at the end of this chapter demonstrates glyph kerning in context with a ContextPosFormat1 subtable.

PosRule subtable

Context Positioning Subtable: Format 2

Format 2, more flexible than Format 1, describes class-based context positioning. For this format, a specific integer, called a class value, must be assigned to each glyph in all context glyph sequences. Contexts are then defined as sequences of class values. This subtable may define more than one context.

To clarify the notion of class-based context rules, suppose that certain sequences of three glyphs need special kerning. The glyph sequences consist of an uppercase glyph that overhangs on the right side, a punctuation mark glyph, and then a quote glyph. In this case, the set of uppercase glyphs would constitute one glyph class (Class1), the set of punctuation mark glyphs would constitute a second glyph class (Class 2), and the set of quote mark glyphs would constitute a third glyph class (Class 3). The input context might be specified with a context rule (PosClassRule) that describes "the set of glyph strings that form a sequence of three glyph classes, one glyph from Class 1, followed by one glyph from Class 2, followed by one glyph from Class 3."

Each ContextPosFormat2 subtable contains an offset to a class definition table (ClassDef), which defines the class values of all glyphs in the input contexts that the subtable describes. Generally, a unique ClassDef will be declared in each instance of the ContextPosFormat2 subtable that is included in a font, even though several Format 2 subtables may share ClassDef tables. Classes are exclusive sets; a glyph cannot be in more than one class at a time. The output glyphs that replace the glyphs in the context sequence do not need class values because they are specified elsewhere by GlyphID.

The ContextPosFormat2 subtable also contains a format identifier (PosFormat) and defines an offset to a Coverage table (Coverage). For this format, the Coverage table lists indices for the complete set of glyphs (not glyph classes) that may appear as the first glyph of any class-based context. In other words, the Coverage table contains the list of glyph indices for all the glyphs in all classes that may be first in any of the context class sequences. For example, if the contexts begin with a Class 1 or Class 2 glyph, then the Coverage table will list the indices of all Class 1 and Class 2 glyphs.

A ContextPosFormat2 subtable also defines an array of offsets to the PosClassSet tables (PosClassSet), along with a count (including Class0) of the PosClassSet tables (PosClassSetCnt). In the array, the PosClassSet tables are ordered by ascending class value (from 0 to PosClassSetCnt - 1).

A PosClassSet array contains one offset for each glyph class, including Class 0. PosClassSets are not explicitly tagged with a class value; rather, the index value of the PosClassSet in the PosClassSet array defines the class that a PosClassSet represents.

For example, the first PosClassSet listed in the array contains all the PosClassRules that define contexts beginning with Class 0 glyphs, the second PosClassSet contains all PosClassRules that define contexts beginning with Class 1 glyphs, and so on. If no PosClassRules begin with a particular class (that is, if a PosClassSet contains no PosClassRules), then the offset to that particular PosClassSet in the PosClassSet array will be set to NULL.

ContextPosFormat2 subtable: Class-based context glyph positioning

All the PosClassRules that define contexts beginning with the same class are grouped together and defined in a PosClassSet table. Consequently, the PosClassSet table identifies the class of a context's first component.

A PosClassSet enumerates all the PosClassRules that begin with a particular glyph class. For instance, PosClassSet0 represents all the PosClassRules that describe contexts starting with Class 0 glyphs, and PosClassSet1 represents all the PosClassRules that define contexts starting with Class 1 glyphs.

Each PosClassSet table consists of a count of the PosClassRules defined in the PosClassSet (PosClassRuleCnt) and an array of offsets to PosClassRule tables (PosClassRule). The PosClassRule tables are ordered by preference in the PosClassRule array of the PosClassSet.

PosClassSet table: All contexts beginning with the same class

For each context, a PosClassRule table contains a count of the glyph classes in a given context (GlyphCount), including the first class in the context sequence. A class array lists the classes, beginning with the second class, that follow the first class in the context. The first class listed indicates the second position in the context sequence.

Note: Text order depends on the writing direction of the text. For text written from right to left, the right-most glyph will be first. Conversely, for text written from left to right, the left-most glyph will be first.

The values specified in the Class array are those defined in the ClassDef table. For example, consider a context consisting of the sequence: Class 2, Class 7, Class 5, Class 0. The Class array will read: Class[0] = 7, Class[1] = 5, and Class[2] = 0. The first class in the sequence, Class 2, is defined by the index into the PosClassSet array of offsets. The total number and sequence of glyph classes listed in the Class array must match the total number and sequence of glyph classes contained in the input context.

A PosClassRule also contains a count of the positioning operations to be performed on the context (PosCount) and an array of PosLookupRecords (PosLookupRecord) that supply the positioning data. For each position in the context that requires a positioning operation, a PosLookupRecord specifies a LookupList index and a position in the input glyph class sequence where the lookup is applied. The PosLookupRecord array lists PosLookupRecords in design order, or the order in which lookups are applied to the entire glyph sequence.

Example 11 at the end of this chapter demonstrates a ContextPosFormat2 subtable that uses glyph classes to modify accent positions in glyph strings.

PosClassRule table: One class context definition

Context Positioning Subtable: Format 3

Format 3, coverage-based context positioning, defines a context rule as a sequence of coverages. Each position in the sequence may specify a different Coverage table for the set of glyphs that matches the context pattern. With Format 3, the glyph sets defined in the different Coverage tables may intersect, unlike Format 2 which specifies fixed class assignments for the lookup (they cannot be changed at each position in the context sequence) and exclusive classes (a glyph cannot be in more than one class at a time).

For example, consider an input context that contains an uppercase glyph (position 0), followed by any narrow uppercase glyph (position 1), and then another uppercase glyph (position 2). This context requires three Coverage tables, one for each position:

• In position 0, the first position, the Coverage table lists the set of all uppercase glyphs.
• In position 1, the second position, the Coverage table lists the set of all narrow uppercase glyphs, which is a subset of the glyphs listed in the Coverage table for position 0.
• In position 2, the Coverage table lists the set of all uppercase glyphs again.

Note: Both position 0 and position 2 can use the same Coverage table.

Unlike Formats 1 and 2, this format defines only one context rule at a time. It consists of a format identifier (PosFormat), a count of the number of glyphs in the sequence to be matched (GlyphCount), and an array of Coverage offsets that describe the input context sequence (Coverage).

Note: The Coverage tables listed in the Coverage array must be listed in text order according to the writing direction. For text written from right to left, the right-most glyph will be first. Conversely, for text written from left to right, the left-most glyph will be first.

The subtable also contains a count of the positioning operations to be performed on the input Coverage sequence (PosCount) and an array of PosLookupRecords (PosLookupRecord) in design order, or the order in which lookups are applied to the entire glyph sequence.

Example 12 at the end of this chapter changes the positions of math sign glyphs in math equations with a ContextPosFormat3 subtable.

ContextPosFormat3 subtable: Coverage-based context glyph positioning

LookupType 8: Chaining Contextual Positioning Subtable

A Chaining Contextual Positioning subtable(ChainContextPos)describes glyph positioning in context with an ability to look back and/or look ahead in the sequence of glyphs. The design of the Chaining Contextual Positioning subtable is parallel to that of the Contextual Positioning subtable, including the availability of three formats.

To specify the context, the coverage table lists the first glyph in the input sequence, and the ChainPosRule subtable defines the rest. Once a covered glyph is found at position i, the client reads the corresponding ChainPosRuleSet table and examines each table to determine if it matches the surrounding glyphs in the text. There is a match if the string <backtrack sequence>+<input sequence>+<lookahead sequence> matches with the glyphs at position i - BacktrackGlyphCount in the text.

If there is a match, then the client finds the target glyphs for positioning and performs the operations. Please note that (just like in the ContextPosFormat1 subtable) these lookups are required to operate within the range of text from the covered glyph to the end of the input sequence. No positioning operations can be defined for the backtracking sequence or the lookahead sequence.

To clarify the ordering of glyph arrays for input, backtrack and lookahead sequences, the following illustration is provided. Input sequence match begins at i where the input sequence match begins. The backtrack sequence is ordered beginning at i - 1 and increases in offset value as one moves away from i. The lookahead sequence begins after the input sequence and increases in logical order.

Chaining Context Positioning Format 1: Simple Chaining Context Glyph Positioning

This Format is identical to Format 1 of Context Positioning lookup except that the PosRule table is replaced with a C