Skip to main content

With Layout

rascal-0.42.2

Synopsis​

Defines a concrete syntax for Exp with layout.

Description​

In Rascal, the major difference between lexical syntax and non-lexical syntax is that:

  • Strings that are parsed according to the lexical syntax do not contain additional layout characters such as spaces, new lines, and source code comments.
  • Strings that are parsed according to the normal (non-lexical) syntax can contain layout characters between each element.
  • Which 'layout' (whitespace and/or source code comments) will be accepted has to be defined explicitly by the grammar author.

The following example extends the grammar for Exp in No Layout with a layout definition:

layout Whitespace = [\t-\n\r\ ]*;      ❶  
    
lexical IntegerLiteral = [0-9]+;           

start syntax Exp 
  = IntegerLiteral          
  | bracket "(" Exp ")"     
  > left Exp "*" Exp        
  > left Exp "+" Exp        
  ;
  • ❶ Using the layout definition, we define that the Whitespace non-terminal is used in between every symbol of the syntax productions in the current module.

And now we can use spaces in our definition of the eval function as well:

import String;
import ParseTree;                                                   

int eval(str txt) = eval(parse(#start[Exp], txt).top);              

int eval((Exp)`<IntegerLiteral l>`) = toInt("<l>");       
int eval((Exp)`<Exp e1> * <Exp e2>`) = eval(e1) * eval(e2);  
int eval((Exp)`<Exp e1> + <Exp e2>`) = eval(e1) + eval(e2); 
int eval((Exp)`( <Exp e> )`) = eval(e);                    

value main() {
  return eval(" 2+3");
}

test bool tstEval1() = eval(" 7") == 7;
test bool tstEval2() = eval("7 * 3") == 21;
test bool tstEval3() = eval("7 + 3") == 10;
test bool tstEval4() = eval(" 3 + 4*5 ") == 23;

Note that Pattern Matching will ignore all trees in layout positions, such that the parse tree of "1 + \n1" will match against <Exp e1> + <Exp e2>. The same goes for equality on parse trees.

For the above example Rascal will insert the Whitespace non-terminal between every element of the syntax rules for Exp. Moreover, for the start production (See No Layout) Whitespace will be added before and after the Exp.

Examples​

The effect of the layout definition is that before parser generation the following grammar is derived for Exp:

syntax Exp 
  = IntegerLiteral          
  | bracket "(" Whitespace Exp Whitespace ")"     
  > left Exp Whitespace "*" Whitespace Exp        
  > left Exp Whitespace "+" Whitespace Exp        
  ;

syntax start[Exp] = Whitespace Exp top Whitespace;

To put this all to the test:

rascal>eval("2 +  3");
int: 5
rascal>eval("2   +  3*4");
int: 14
rascal>eval("( 2+3 )* 4");
int: 20

Pitfalls​

  • If the grammar for Exp would contain an optional symbol, as in syntax Exp = Exp "+"? Exp, then it would be ambiguous. Does a space in "1 1", belong to the Whitespace before or after the missing +? To disambiguate the layout definition should be changed to layout Whitespace = [\ \t\n\r]* !>> [\ \t\n\r]. That will make sure the space goes with the first Whitespace, because even an empty Whitespace list must never be followed immediately by a space.