[antlr-interest] Suppressing spurious warnings

[Steve Bennett]

On 2/28/08, Gavin Lambert <antlr at mirality.co.nz> wrote:
> You can suppress the warning by removing PRE_OPEN from the tokens
>  block and adding a fragment rule for it instead:
>
>  fragment PRE_OPEN: ' ';

Ok, but that's still kind of annoying: I have to add fake code in
order to suppress a warning. Since I want to use this grammar as
documentation for the language, any code that exists just to make
ANTLR shut up while not adding any information about the language is a
Bad Thing.

No chance of an options { suppress_warning_undefined_token; } ?

> At present, the only way to remove these warnings is to rewrite
>  your grammar to remove the ambiguity.  Note that while ANTLR often
>  does the right thing anyway, you can't just assume that.  You
>  should always either rewrite the grammar or test its behaviour
>  thoroughly.  Or both.

Why can't I assume that ANTLR will do the correct thing? Is there an
example of somewhere where it does the unexpected thing? I do plan on
testing thoroughly, btw.

Also, when you say "at present", is there a chance of that warning
suppression being added?

> True left-recursion is fatal to ANTLR, so must be corrected.


>  I have yet to see a case where ignoring this warning is a good
>  idea.

I'm sure you're right, I just can't find the source of it.

I notice also that for my grammar with backtrack=true, I don't get
that warning in ANTL 3.0.1, but in ANTLR 3.1b1 it becomes a fatal
error:

error(10):  internal error:
org.antlr.tool.Grammar.createLookaheadDFA(Grammar.java:1138): could
not even do k=1 for decision 20; reason: timed out (>1000ms)

warning(205): mediawiki12.g:885:44: ANTLR could not analyze this
decision in rule bold_and_italics; often this is because of recursive
rule references visible from the left edge of alternatives.  ANTLR
will re-analyze the decision with a fixed lookahead of k=1.  Consider
using "options {k=1;}" for that decision and possibly adding a
syntactic predicate.


Here's the code:
bold_and_italics options {backtrack=false;}:
     {textis("''") && text_italics}? => APOSTROPHES
{text_italics=false; System.out.print("</i>"); } ->
^(I_OFF)
    |{textis("''") && !text_italics}? => APOSTROPHES
{text_italics=true; System.out.print("<i>"); }   ->            ^(I_ON)
    |{textis("'''") && text_bold}? => APOSTROPHES    {text_bold=false;
System.out.print("</b>"); }    ->            ^(B_OFF)
    |{textis("'''") && !text_bold}? => APOSTROPHES   {text_bold=true;
System.out.print("<b>"); }      ->            ^(B_ON)
    |{textis("''''") && text_bold}? => APOSTROPHES   {text_bold=false;
System.out.print("'</b>"); }   -> APOSTROPHE ^(B_OFF)
    |{textis("''''") && !text_bold}? => APOSTROPHES  {text_bold=true;
System.out.print("'<b>"); }     -> APOSTROPHE ^(B_ON)
    |{textis("'''''") && text_bold && text_italics}? =>  APOSTROPHES
{text_bold=false; text_italics=false; System.out.print("'</b> </i>");}
-> ^(B_OFF) ^(I_OFF)
    |{textis("'''''") && text_bold && !text_italics}? => APOSTROPHES
{text_bold=false; text_italics=true; System.out.print("'</b> <i>");}
-> ^(B_OFF) ^(I_ON)
    |{textis("'''''") && !text_bold && text_italics}? => APOSTROPHES
{text_bold=true; text_italics=false; System.out.print("'<b> </i>"); }
-> ^(B_ON)  ^(I_OFF)
    |{textis("'''''") && !text_bold && !text_italics}? =>APOSTROPHES
{text_bold=true; text_italics=true; System.out.print("'<b> <i>");}
-> ^(B_ON)  ^(I_ON)
    // Hopefully we never get more than 6 or less than 2. The lexer
should take care of that.
    ;

Steve