The Mac State Machine Dialog

The mac uses state machines where opentype would use a contextual lookup. The example at right (with appropriate substitutions attached to it) will convert an initial or medial short-s into a long-s (for doing renaissance printing).

PfaEdit currently supports 4 types of state machines: contextual glyph substitution, indic rearrangement, contextual glyph insertion, and contextual kerning. The example at right is of contextual glyph substitution.

State machines have certain similar features. All but kerning be assigned to a feature/setting combination, they all have a set of flags indicating in what circumstances they apply, they all have the glyphs of the font assigned to a set of classes, and finally they all have a series of transitions indicating how each state behaves when given input from the specified class.

On the mac there are four predefined classes, one meaning the end of input, one for all glyphs not specified elsewhere, one for the magic "deleted" glyph (Apple does not always delete a glyph when asked to do so, instead it retains a special mark in the glyph stream indicating that the glyph should be deleted later), and finally a class for the end of a line of text. Any classes after this are user defined. In the example at right there is one user defined class and it contains the letters in the latin alphabet. Any other characters in the font (the digits for example) will automatically be assigned to class 1 ("{Everything Else}").

The state machine itself is shown as a two dimensional matrix of transition, one transition for every state and class. Classes are shown at the top of the matrix, and states down the side. The transition from state 0 when given input from class 0 is shown in the upper left corner of the matrix, and so on.

When the state machine starts up it will be in one of the first two states. It will be in state 0 at the beginning of input, and in state1 when it starts at the beginning of a new line.

The transitions look slightly differently depending on the type of state machine:

Indic	Contextual	Insertion	Kerning

The first line of all transitions shows the next state. The next line contains a set of flags where "M" => the current glyph should be marked (the meaning of this becomes appearant later), "A" => processing should advance to the next glyph. For indic tables there is a third flag shown, "L" => the current glyph is the last glyph of the indic rearrangement, while for kerning tables there is a "P" flag indicating that the current glyph should be pushed onto the kerning stack.

The last two lines have varied meanings depending on the type of the state machine.

• For Indic transitions the two lines should be appended to get the indic rearrangement verb (in the example above that would be "Ax => xA" meaning that the first glyph marked should be moved after the last glyph marked)
• For contextual transitions the two lines specify the names (as 4 character tags) of nested substitutions. The first one is applied to the last glyph marked and the second one to the current glyph (if a line is blank, no substitution is applied to that glyph here).
• For insertion transitions the lines specify the glyphs that are to be inserted. The first line applies to the marked glyph, the second line to current glyph (there may not be room enough to show all the inserted glyphs, in which case the glyph list is truncated). There are additional flags here that are also not shown because of lack of room. They may be viewed by clicking on a transition.
• For kerning transitions the first line shows the first (few) kerning values, the second line shows either "..." if there are more kerning values, or nothing.
  Note: The number of kerning values MUST match the number of glyphs pushed on the kern stack, or your application crashes.

For a complete description of what these state machines can do you should read Apple's documentation

• On the 'morx' and 'mort' tables
• On the 'kern' table (format 1)
• On state machines in general
• The 'feat' (feature) table
• Apple's Font Feature registry

The general idea is that a word processor converts text into a string of glyphs, and then the state machine is applied to that string. The state machine starts in state 0 or 1, it figures out what class the first glyph belongs to, and indexes into the current state by that class. The transition at that location provides a new state to move to (which might be the same as the current one), it will usually advance to the next glyph (so the new state will be indexed by a different glyph) but sometimes you will want to continue processing the current glyph in the new state. Also it will perform a set of operations which differ depending on the type of state machine.

Many of the state machines have the concept of a "marked" glyph. This is important if the desired change for a given glyph depends on what glyphs follow it. In that case when we reach an interesting glyph we "mark" it, and then, after reading a few more glyphs when we know what to do, we apply a transformation to the marked glyph.

Indic	Contextual	Insertion	Kerning
In addition to a "mark"ed glyph, indic transformations also have a the concept of a "last" glyph. There are 16 transformations which may be applied the the glyphs between the "mark"ed and "last" glyphs. Suppose the glyph stream looks like     abcdef And "a" was "mark"ed and "d" was "last" then a transition "AxD => DxA" refers to the marked glyph as "A", the last glyph as "D" and everything in between as "x", so this would change the glyph stream to     dbcaef	This can apply a substitution to the current glyph, a different substitution to the marked glyph, and it may mark the current glyph. The substitutions are indicated by 4 character tags and should be "Nested" substitutions which may be created by the Element->Char Info dialog.	This can insert a string (up to 31 glyphs) before or after the current and the marked glyphs.	Kerning does not have a single "mark"ed glyph, instead it can push up to 8 glyphs onto a stack, and at a later point a set of kerning values may be applied to those glyphs (so the flag which is usually indicated as "M" for mark, is here "P" for push). Once all the glyphs have been pushed, a set of kerning values may be applied. One value for each pushed glyph. This is a number which indicates how much the glyph in question should be kerned. If the glyph stream looks like:     F." and you want to kern "." by -300, and """ by 0 em-units, then you would enter     -300 0 as kerning values.

Editing Transitions

To edit a transition, simply click on it and a dialog will pop up, appropriate to the type of state machine.

Indic	Contextual	Insertion	Kerning

See above for a description of what the fields mean.

If you wish to add a new state, simply edit a transition so that its next state field is the state number desired. PfaEdit will automatically create the state for you.

Classes

If you wish to add a new class press the [New] button under the list of classes. If you wish to edit a class double click on that class. The dialog should change to be like the one at right. The text field contains a list of the names of the glyphs in this class. Pressing the [Set] button will set that field to any selected glyphs in the font view. Pressing the [Select] button will change the selection in the font view to match the glyphs in the text.

-- Up -- TOC --