Programming Forums > Application Development > C++

programmingnoob · Jul 11th, 2006, 5:13 PM

Quote:

Originally Posted by The Dark

Try something like:

tem<string>*     procstmt()
	 {
		  if (NT->theStruct.type != procedure)
		 {
		     if(compound()==NULL) return NULL; 
 
			 if (NT->theStruct.type != scln){  return NULL;}
			 NT = Si.cget();
		 }
		 else{ NT = Si.cget();
			y = proc();
			if (y != NULL) {
                          if (procstmt() == NULL)
                            return NULL;
                        }
			else return NULL; 
		 }
                 // Return something not NULL
	 }

the problem is I dont think we are allowed to have a function call itself!
thats why i followed the longer route

The Dark · Jul 11th, 2006, 6:03 PM

That sounds a bit harsh, if the spec calls itself, your program should be able to!

programmingnoob · Jul 11th, 2006, 6:44 PM

Quote:

Originally Posted by The Dark

That sounds a bit harsh, if the spec calls itself, your program should be able to!

i know!! ... isnt it unfair?!

let me give you an example though ...
//<dec-list> ::= <dec> | <dec-list> ; <dec>

(Toggle Plain Text)

item<string>*	 declist()
	 {
		 if (dec() == NULL) { return NULL;}
		 while (NT->theStruct.type == cma) 
		 {NT = Si.cget();
		 if (dec() == NULL) {return NULL;}}
		 NT = Si.cget();
	     return NT;
	 }

mikaoj · Jul 11th, 2006, 6:53 PM

[EDIT] Shit, I was on the wrong site. Sorry! [/EDIT]

Delete this post, please.

The Dark · Jul 11th, 2006, 7:49 PM

How about something like:

(Toggle Plain Text)

item<string>*     procstmt()
{
  while (true)
  {
    if (NT->theStruct.type != procedure)
    {
      if (compound()==NULL)
        return NULL; 

      if (NT->theStruct.type != scln)
      {
        return NULL;
      }
      NT = Si.cget();
      break;
    }
    else
    {
      NT = Si.cget();
      if (proc()==NULL)
        return NULL;
      // Loop around again
    }
  }
  return OK;
}

(I had to reformat your code to get my head around it)

Also in your last example:

(Toggle Plain Text)

while (NT->theStruct.type == cma)

Should be

(Toggle Plain Text)

while (NT->theStruct.type == scln)

programmingnoob · Jul 19th, 2006, 9:11 PM

Quote:

Originally Posted by The Dark

How about something like:

(Toggle Plain Text)

item<string>*     procstmt()
{
  while (true)
  {
    if (NT->theStruct.type != procedure)
    {
      if (compound()==NULL)
        return NULL; 

      if (NT->theStruct.type != scln)
      {
        return NULL;
      }
      NT = Si.cget();
      break;
    }
    else
    {
      NT = Si.cget();
      if (proc()==NULL)
        return NULL;
      // Loop around again
    }
  }
  return OK;
}

(I had to reformat your code to get my head around it)

Also in your last example:

(Toggle Plain Text)

while (NT->theStruct.type == cma)

Should be

(Toggle Plain Text)

while (NT->theStruct.type == scln)

thanks! but i didnt really try your code, i fixed a couple of things in my function definiton and other things, and now my yes-no parser is working now.

programmingnoob · Jul 19th, 2006, 9:23 PM

now, i have to create abstract syntax tree. how do i do that?
i mean i dont actually understand the concept i think

for every single rule, the abstract syntax tree is given ... but I just do not know what they are talking about ...

Arevos · Jul 20th, 2006, 5:35 AM

Quote:

Originally Posted by programmingnoob

for every single rule, the abstract syntax tree is given ... but I just do not know what they are talking about ...

The concept is really pretty simple. It just sounds complex.

Take the following expression:

(Toggle Plain Text)

x = y + 1

The AST for this might look like:

(Toggle Plain Text)

The assignment operator has the greatest presidence, so it sits at the top of the tree. Hopefully you can see how the AST represents the parsed code in memory. With an AST, the code is easier to access, and easier to evaluate.

From a data structure point of view, each node of the tree has a number of pointers to its children (and possible one to its parent). You can represent the node as a struct, or a class.

programmingnoob · Jul 22nd, 2006, 12:19 AM

thanks arevos, i was just lacking confidence to actually implement the tree

programmingnoob · Jul 22nd, 2006, 12:45 AM

As you guys know, I am done with my parser that only validates a given expression. However I did a little bit of cheating with the rule of stmt:-

<stmt> ::= <assign> | <call> | <pread> | <pwriteln> |
<compound> | <conditional>

Now what I did was just take a look at the first token, and determine which funcion should be called. Here are the corresponding rules that clash though:

<call> ::= <proc-name> ( <params> )

<assign> ::= $id := <exp>

They would clash because both <proc-name> and $id require the first token to be $id. So I wouldnt know which function to call unless I take a look at the second token.
However, if I do get the second token, then I can not call the respective functions (assign or call in this case) because it would mess thing up, if you know what i am talking about.

I think the main problem is there is no way i can put the token back, so what i basically did is i didnt call the functions "call" and "assign", instead i just wrote them out like this:

(Toggle Plain Text)

 if (NT->theStruct.type==id) 
		 { 
			 NT = Si.cget(); if (NT->theStruct.type==asgn){ NT = Si.cget(); if (exp() == NULL) {return NULL;}}
			 else if (NT->theStruct.type==lp) {NT = Si.cget(); if (params() == NULL) { return NULL;} if (NT->theStruct.type != rp){return NULL;}NT = Si.cget();}
			return NT; 
		 }

but that's cheating! isnt it?

but what's the way out?

Jul 11th, 2006, 6:53 PM	#34
mikaoj Programmer Join Date: Aug 2005 Location: Norway Posts: 56 Rep Power: 0	[EDIT] Shit, I was on the wrong site. Sorry! [/EDIT] Delete this post, please. __________________ Heh.

Jul 11th, 2006, 7:49 PM	#35
The Dark Expert Programmer Join Date: Jun 2005 Posts: 893 Rep Power: 4	How about something like: (Toggle Plain Text) item<string>* procstmt() { while (true) { if (NT->theStruct.type != procedure) { if (compound()==NULL) return NULL; if (NT->theStruct.type != scln) { return NULL; } NT = Si.cget(); break; } else { NT = Si.cget(); if (proc()==NULL) return NULL; // Loop around again } } return OK; } item<string>* procstmt() { while (true) { if (NT->theStruct.type != procedure) { if (compound()==NULL) return NULL; if (NT->theStruct.type != scln) { return NULL; } NT = Si.cget(); break; } else { NT = Si.cget(); if (proc()==NULL) return NULL; // Loop around again } } return OK; } (I had to reformat your code to get my head around it) Also in your last example: (Toggle Plain Text) while (NT->theStruct.type == cma) while (NT->theStruct.type == cma) Should be (Toggle Plain Text) while (NT->theStruct.type == scln) while (NT->theStruct.type == scln)

Jul 22nd, 2006, 12:45 AM	#40
programmingnoob Hobbyist Programmer Join Date: Feb 2006 Posts: 155 Rep Power: 3	As you guys know, I am done with my parser that only validates a given expression. However I did a little bit of cheating with the rule of stmt:- <stmt> ::= <assign> \| <call> \| <pread> \| <pwriteln> \| <compound> \| <conditional> Now what I did was just take a look at the first token, and determine which funcion should be called. Here are the corresponding rules that clash though: <call> ::= <proc-name> ( <params> ) <assign> ::= $id := <exp> They would clash because both <proc-name> and $id require the first token to be $id. So I wouldnt know which function to call unless I take a look at the second token. However, if I do get the second token, then I can not call the respective functions (assign or call in this case) because it would mess thing up, if you know what i am talking about. I think the main problem is there is no way i can put the token back, so what i basically did is i didnt call the functions "call" and "assign", instead i just wrote them out like this: (Toggle Plain Text) if (NT->theStruct.type==id) { NT = Si.cget(); if (NT->theStruct.type==asgn){ NT = Si.cget(); if (exp() == NULL) {return NULL;}} else if (NT->theStruct.type==lp) {NT = Si.cget(); if (params() == NULL) { return NULL;} if (NT->theStruct.type != rp){return NULL;}NT = Si.cget();} return NT; } if (NT->theStruct.type==id) { NT = Si.cget(); if (NT->theStruct.type==asgn){ NT = Si.cget(); if (exp() == NULL) {return NULL;}} else if (NT->theStruct.type==lp) {NT = Si.cget(); if (params() == NULL) { return NULL;} if (NT->theStruct.type != rp){return NULL;}NT = Si.cget();} return NT; } but that's cheating! isnt it? but what's the way out?

Thread Tools
Show Printable Version Email this Page
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode

Jul 11th, 2006, 6:03 PM	#32
The Dark Expert Programmer Join Date: Jun 2005 Posts: 893 Rep Power: 4	That sounds a bit harsh, if the spec calls itself, your program should be able to!

Jul 19th, 2006, 9:23 PM	#37
programmingnoob Hobbyist Programmer Join Date: Feb 2006 Posts: 155 Rep Power: 3	now, i have to create abstract syntax tree. how do i do that? i mean i dont actually understand the concept i think for every single rule, the abstract syntax tree is given ... but I just do not know what they are talking about ...

Jul 22nd, 2006, 12:19 AM	#39
programmingnoob Hobbyist Programmer Join Date: Feb 2006 Posts: 155 Rep Power: 3	thanks arevos, i was just lacking confidence to actually implement the tree

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)

Programming Forums > Application Development > C++

writing a parser in c++