CODEWITHSUNDEEP
whitespacevariable-assignmentantlr4
user1511417
user1511417

Reputation: 1940

ANTLR v4 multi-assignment

In Terence Parr'S Definitiv ANTLR Reference on page 39 an assignment ist defined like this:

stat: expr (NL|SEMI)            # printExpr
| var EQL expr (NL|SEMI)        # assign        // x = 5; y = x     
| (NL|SEMI)                     # blank
;
...
WS  : [ \r\t]+ -> skip;
NL  : '\r' ? '\n';
SEMI: ';';
...

This works fine.

But, how can I achieve something like this x = 6 y=x ?

Here: two assignments separated by a single whitespace (WS is skipped, btw.)

Upvotes: 1

Views: 101

Answers (1)

GRosenberg
GRosenberg

Reputation: 5991

A language designed to be dependent on occasional whitespace significance is nearly always a bad idea, and moreso when trying to parse it using a context-free parser. That said, there are two alternatives, at least for this example:

First, put the WS on the hidden channel. Then, use a predicate to test for a WS token in the token stream where it is significant:

| var EQL expr ( {atWS()}? | NL | SEMI )   # assign

Second, don't hide or skip the WS. Then, use like any other token where it should be significant:

| var WS? EQL WS? expr ( WS | NL | SEMI )  # assign

And, add a rule to capture otherwise unused WS:

ws: WS ;  // will pollute the parse tree, but can be ignored there

Both alternatives have consequences, moderate to severe, depending on what you are trying to accomplish.

Upvotes: 1

Related Questions