PIKApp/pdb
Cassowary 3bbdd873ef Updated with upstream update 2023-10-30 15:55:30 -07:00
..
groups Updated with upstream update 2023-10-30 15:55:30 -07:00
README Initial checkin of Pika from heckimp 2023-09-25 15:35:21 -07:00
README_NEW_PDB_PROC Initial checkin of Pika from heckimp 2023-09-25 15:35:21 -07:00
app.pl Updated with upstream update 2023-10-30 15:55:30 -07:00
enumcode.pl Initial checkin of Pika from heckimp 2023-09-25 15:35:21 -07:00
enumgen.pl Updated with upstream update 2023-10-30 15:55:30 -07:00
enums-external.pl Initial checkin of Pika from heckimp 2023-09-25 15:35:21 -07:00
enums.pl Initial checkin of Pika from heckimp 2023-09-25 15:35:21 -07:00
groups.pl Updated with upstream update 2023-10-30 15:55:30 -07:00
lib.pl Updated with upstream update 2023-10-30 15:55:30 -07:00
meson-pdbgen.sh Initial checkin of Pika from heckimp 2023-09-25 15:35:21 -07:00
meson.build Updated with upstream update 2023-10-30 15:55:30 -07:00
pdb.pl Updated with upstream update 2023-10-30 15:55:30 -07:00
pdbgen.pl Initial checkin of Pika from heckimp 2023-09-25 15:35:21 -07:00
stddefs.pdb Initial checkin of Pika from heckimp 2023-09-25 15:35:21 -07:00
util.pl Initial checkin of Pika from heckimp 2023-09-25 15:35:21 -07:00

README

Some mostly unfinished docs are here.

-Yosh

This document describes the tool PDBGEN.

If you added or modified .pdb files do not run this tool manually but
run make instead! It will call pdbgen.pl then to generate the files
into the right output directories.

PDBGEN
------------------

What is this?
PDBGEN is a tool to automate much of the drudge work of making PDB interfaces
to PIKA internals. Right now, it generates PDB description records,
argument marshallers (with sanity checking) for the app side, as well
as libpika wrappers for C plugins. It's written so that extending it
to provide support for CORBA and other languages suited to static
autogeneration.

Invoking PDBGEN from the command line:
1. Change into the ./pdb directory.
2. $ ./pdbgen.pl DIRNAME
where DIRNAME is either "lib" or "app", depending on which set of files
you want to generate. The files are written into $destdir/app or $destdir/libpika.
$destdir is the environment variable destdir. If it's not set,
then it's the ./pdb directory. Make sure the directories
$destdir/app and $destdir/libpika already exist and you have write permissions.
Otherwise the code generator will fail and exit.
It's up to you to diff the file you changed. When you're happy with
the generated file, copy it into the actual ./app/ or ./libpika/ directory
where it finally gets built.

Anatomy of a PDB descriptor:
PDB descriptors are Perl code. You define a subroutine, which corresponds
to the PDB function you want to create. You then fill certain special
variables to fully describe all the information pdbgen needs to generate
code. Since it's perl, you can do practically whatever perl lets you
do to help you do this. However, at the simplest level, you don't need
to know perl at all to make PDB descriptors.

Annotated description:
For example, we will look at pika_display_new, specified in gdisplay.pdb.

sub display_new { 

We start with the name of our PDB function (not including the "pika_" prefix).

    $blurb = 'Create a new display for the specified image.';

This directly corresponds to the "blurb" field in the ProcRecord.

    $help = <<'HELP';
Creates a new display for the specified image. If the image already has a
display, another is added. Multiple displays are handled transparently by the
PIKA. The newly created display is returned and can be subsequently destroyed
with a call to 'pika-display-delete'. This procedure only makes sense for use
with the PIKA UI.
HELP

This is the help field. Notice because it is a long string, we used HERE
document syntax to split it over multiple lines. Any extra whitespace
in $blurb or $help, including newlines, is automatically stripped, so you
don't have to worry about that.

    &std_pdb_misc;

This is the "author", "copyright", and "date" fields. Since S&P are quite
common, they get a special shortcut which fills these in for you. Stuff
like this is defined in stddefs.pdb.

    @inargs = ( &std_image_arg );

You specify arguments in a list. Again, your basic image is very common,
so it gets a shortcut.

    @outargs = (
        { name => 'display', type => 'display',
          desc => 'The new display', alias => 'gdisp', init => 1 }
    );

This is a real argument. It has a name, type, description at a minimum.
"alias" lets you use the alias name in your invoker code, but the real
name is still shown in the ProcRecord. This is useful not only as a
shorthand, but for grabbing variables defined somewhere else (or constants),
in conjunction with the "no_declare" flag. "init" simply says initialize
this variable to a dummy value (in this case to placate gcc warnings)

    %invoke = (
        headers => [ qw("gdisplay.h") ],

These are the headers needed for the functions you call.

        vars => [ 'guint scale = 0x101' ],

Extra variables can be put here for your invoker.

        code => <<'CODE'
{
  if (gimage->layers == NULL)
    success = FALSE;
  else
    success = ((gdisp = gdisplay_new (gimage, scale)) != NULL);
}
CODE

The actual invoker code. Since it's a multiline block, we put curly braces
in the beginning.