Mark: 9 revision(s)

2008-11-03T05:20:39Z

9 revision(s)

New page

comiler...
ok so we need a different function (perhaps) for each thing we want to do. Load idents, strings, perform operations, evaluate statements.

and now there's a whole heap of undefined references
cygdrive/c/Users/sdal039/AppData/Local/Temp/ccDd1mQd.o:parser.c:(.text+0x202): undefined reference to `_end_of_tokens'
/cygdrive/c/Users/sdal039/AppData/Local/Temp/ccDd1mQd.o:parser.c:(.text+0x212): undefined reference to `_error'
/cygdrive/c/Users/sdal039/AppData/Local/Temp/ccDd1mQd.o:parser.c:(.text+0x240): undefined reference to `_current'
/cygdrive/c/Users/sdal039/AppData/Local/Temp/ccDd1mQd.o:parser.c:(.text+0x2a6): undefined reference to `_advance'

and so on...

aha.. toy tokeniser is up now. errors gone.

ok this is stupidly hard :P
code so far:
static int strCount = 0;
void identifiers( Tree* pTree )
{
if (pTree != 0) {
if (isVariable( pTree->name )) { //variable assignment
printf(".globl %c\n\t%c .long 0\n", pTree->name, pTree->name );
}

if (isConstant( pTree->name )) { //string assignment
printf("S%d:\n\t.ascii \"%c\"\n", strCount, pTree->name);
strCount++;
}
}

int i;
for (i = 0; i < pTree->arity ; i++)
identifiers( pTree->arg[i] );
}

Well it does do something..
given the token stream 'a=1' it outputs
.globl a
a .long 0
S0:
.ascii "1"

adding two numbers...
Tree* compile(Tree* t)
{
if (t != 0) {
if (eqToken( t->name, TOK_ADD )) {//addition
Tree* left = compile( t->arg[0] );//get first opperand
Tree* right = compile( t->arg[1] );//get second opperand
printf("pop\%%c\npop\%%c\nadd \%%c, \%%c\npush\%%c", left->name, right->name, right->name, left->name, left->name);
}
}
return t;
}

segmentation fault...yaaaayyyy.
>_<

What makes this task hard is that without proper implementations of the tokeniser and parse tree it's really hard to write. toy-tokeniser doesn't really allow you to test things, similarly with the parse tree. for example, it can't easily differentiate between a variable 'i' and and if statement as it tokenises ifs to just 'i'. Can't represent quoted strings either as each token is simply 1 character.

Mark suggested doing these three tasks (or at least the first two) over two/three separate labs, in the order than they are used. this would make it easier at each stage and at the end we'd have a pretty solid parser with everyone working on one bit at a time.

SE250:lab-9:sdal039 - Revision history

Mark: 9 revision(s)