Design Principles

    $f->{'loca'}{'glyphs'}[$glyph]{'yMax'}

Or, if we are feeling very lazy and don't mind waiting:

    $f->{'loca'}{'glyphs'}[$f->{'cmap'}->ms_lookup(0x41)]{'yMax'}

The disadvantage of this method is that it behoves module users to behave themselves. Thus it does not hold your hand and ensure that if you make a change to a table, that the table is marked as dirty, or that other tables are updated accordingly.

It is up to the application developer to understand the implications of the changes they make to a font, and to take the necessary action to ensure that the data they get out is what they want. Thus, you could go and change the "yMax" value on a glyph and output a new font with this change, but it is up to you to ensure that the font's bounding box details in the "head" table are correct, and even that your changing "yMax" is well motivated.

To help with using the system, each module (or table) will not only describe the methods it supports, which are relatively few, but also the instance variables it supports, which are many. Most of the variables directly reflect table attributes as specified in the OpenType specification, available from Microsoft (<http://www.microsoft.com/typography>), Adobe and Apple. A list of the names used is also given in each module, but not necessarily with any further description. After all, this code is not a TrueType manual as well!

Conventions

Firstly we consider the documentation conventions regarding instance variables. Each instance variable is marked indicating whether it is a (P)rivate variable which users of the module are not expected to read and certainly not write to or a (R)ead only variable which users may well want to read but not write to.

read and read_dat

    $f = Font::TTF::Font->open($filename) || die "Unable to read $filename";

This will open an existing font and read its directory header. Notice that at this point, none of the tables in the font have been read. (Actually, the "head" and "maxp" tables are read at this point too since they contain the commonly required parameters of):

    $f->{'head'}{'unitsPerEm'}
    $f->{'maxp'}{'numGlyphs'}

In order to be able to access information from a table, it is first necessary to "read" it. Consider trying to find the advance width of a space character (U+0020). The following code should do it:

    $f = Font::TTF::Font->open($ARGV[0]);
    $snum = $f->{'cmap'}->ms_lookup(0x0020);
    $sadv = $f->{'hmtx'}{'advance'}[$snum];
    print $sadv;

This would result in the value zero being printed, which is far from correct. But why? The first line would correctly read the font directory. The second line would, incidently, correctly locate the space character in the Windows cmap (assuming a non symbol encoded font). The third line would not succeed in its task since the "hmtx" table has not been filled in from the font file. To achieve what we want we would first need to cause it to be read:

    $f->{'hmtx'}->read;
    $sadv = $f->{'hmtx'}{'advance'}[$snum];

Or for those who are too lazy to write multiple lines, "read" returns the object it reads. Thus we could write:

    $sadv = $f->{'hmtx'}->read->{'advance'}[$snum];

Why, if we always have to read tables before accessing information from them, did we not have to do this for the "cmap" table? The answer lies in the method call. It senses that the table hasn't been read and reads it for us. This will generally happen with all method calls, it is only when we do direct data access that we have to take the responsibility to read the table first.

Reading a table does not necessarily result in all the data being placed into internal data structures. In the case of a simple table "read" is sufficient. In fact, the normal case is that "read_dat" reads the data from the file into an instance variable called ' dat' (including the space) and not into the data structures.

This is true except for the "glyph" class which represents a single glyph. Here the process is reversed. Reading a "glyph" reads the data for the glyph into the ' dat' instance variable and sets various header attributes for the glyph ("xMin", "numContours", etc.). The data is converted out of the variable into data structures via the "read_dat" method.

The aim, therefore, is that "read" should do the natural thing (read into data structures for those tables and elements for which it is helpful -- all except "glyph" at present) and "read_dat" should do the unnatural thing: read just the binary data for normal tables and convert binary data to data structures for "glyph"s.

In summary, therefore, use "read" unless you want to hack around with the internals of glyphs in which case see Font::TTF::Glyph for more details.

update

       +--loca
       |
 glyf--+--maxp
       |
       +---+--head
           |
 hmtx------+--hhea

 cmap-----OS/2

 name--

 post--
There is an important universal dependency which it is up to the user to
keep up to date. This is C<maxp/numOfGlyphs> which is used to iterate over all
the glyphs. Note that the glyphs themselves are not held in the C<glyph> table
but in the C<loca> table, so adding glyphs, etc. automatically involves keeping
the C<loca> table up to date.

Creating fonts

The following information is required:

    $f->{'loca'}{'glyphs'}
    $f->{'head'}{'upem'}
    $f->{'maxp'}{'numGlyphs'}   (doesn't come from $f->{'loca'}{'glyphs'})
    $f->{'hmtx'}{'advance'}
    $f->{'post'}['format'}
    $f->{'post'}{'VAL'}
    $f->{'cmap'}
    $f->{'name'}

Pretty much everything else is calculated for you. Details of what is needed for a glyph may be found in Font::TTF::Glyph. Once we have all the information we need (and there is lots more that you could add) then we simply

    $f->dirty;          # mark all tables dirty
    $f->update;         # update the font