Table Driven Madness

I've contemplated further work on NedOS, but the current barrier is pretty high. In order to enable interrupts (new style - ACPIA, not legacy PIC) I need to parse the AML tables. Retrieving the code isn't hard, but parsing it is a nightmare.

I have the source code of the Intel reference parser. It reminds me of some code from an earlier project - the table driven state machine.

Instead of a human readable handler (say for a FPS):
void onEvent(type)
{
switch(type)
{
case KEY_A: move_left();
case KEY_D: move_right();
case KEY_SPACE: fire();
}
}

you have this:
while( curState != END)
{
get_input(&input);
output = table[curState].out;
curState = table[curState][input].nextState;
}

On our project this made at least some sense. The hardware under development had the same tables, which allows you to fix bugs by patching the table ROM.

For pure software, I don't see the point. Now, instead of being able to read through some code and figure out what is going on, I get to state at tables like this:
/* Index Name Parser Args Interpreter Args ObjectType Class Type Flags */

/* 00 */ ACPI_OP ("Zero", ARGP_ZERO_OP, ARGI_ZERO_OP, ACPI_TYPE_INTEGER, AML_CLASS_ARGUMENT, AML_TYPE_CONSTANT, AML_CONSTANT),
/* 01 */ ACPI_OP ("One", ARGP_ONE_OP, ARGI_ONE_OP, ACPI_TYPE_INTEGER, AML_CLASS_ARGUMENT, AML_TYPE_CONSTANT, AML_CONSTANT),
/* 02 */ ACPI_OP ("Alias", ARGP_ALIAS_OP, ARGI_ALIAS_OP, ACPI_TYPE_LOCAL_ALIAS, AML_CLASS_NAMED_OBJECT, AML_TYPE_NAMED_SIMPLE, AML_HAS_ARGS | AML_NSOBJECT | AML_NSOPCODE | AML_NSNODE | AML_NAMED),
/* 03 */ ACPI_OP ("Name", ARGP_NAME_OP, ARGI_NAME_OP, ACPI_TYPE_ANY, AML_CLASS_NAMED_OBJECT, AML_TYPE_NAMED_COMPLEX, AML_HAS_ARGS | AML_NSOBJECT | AML_NSOPCODE | AML_NSNODE | AML_NAMED),
/* 04 */ ACPI_OP ("ByteConst", ARGP_BYTE_OP, ARGI_BYTE_OP, ACPI_TYPE_INTEGER, AML_CLASS_ARGUMENT, AML_TYPE_LITERAL, AML_CONSTANT),
/* 05 */ ACPI_OP ("WordConst", ARGP_WORD_OP, ARGI_WORD_OP, ACPI_TYPE_INTEGER, AML_CLASS_ARGUMENT, AML_TYPE_LITERAL, AML_CONSTANT),
/* 06 */ ACPI_OP ("DwordConst", ARGP_DWORD_OP, ARGI_DWORD_OP, ACPI_TYPE_INTEGER, AML_CLASS_ARGUMENT, AML_TYPE_LITERAL, AML_CONSTANT),
/* 07 */ ACPI_OP ("String", ARGP_STRING_OP, ARGI_STRING_OP, ACPI_TYPE_STRING, AML_CLASS_ARGUMENT, AML_TYPE_LITERAL, AML_CONSTANT),
/* 08 */ ACPI_OP ("Scope", ARGP_SCOPE_OP, ARGI_SCOPE_OP, ACPI_TYPE_LOCAL_SCOPE, AML_CLASS_NAMED_OBJECT, AML_TYPE_NAMED_NO_OBJ, AML_HAS_ARGS | AML_NSOBJECT | AML_NSOPCODE | AML_NSNODE | AML_NAMED),

Oh yea, that makes total sense. I see exactly what's happening here...