Re: Semantics make a grammar ambiguous

Hi Robert,


Robert Sherry wrote in message 99-04-099...
>I gather that your scanner does the lookup into the symbol table.


Yes, it does.


>However, if you were to switch to having it return ID for all
>variables and type names you might introduce shift/reduce conflicts in
>your grammar.


:) You're right again! I did this and got exactly what you describe - many
conflicts.


> I had a problem like this when I was trying to write a parser
>for a spice like language. Spice is a language for describing analog
>circuits. I solved the problem by having the parser feed information
>back to the scanner about what to expect. Knows the scanner would know
>whether it should expect a type or a identifier.


This is quite clever - thanks! Never quite thought of using the
parser to tell the scanner what it can and can't output. I will do
this - I can see already that this is what I can do and I can leave
the grammer alone.


>The text books say that this is a bad idea but it might solve your problem.


A bad idea that solves a nasty problem...Isn't that a good idea? :)


>This idea has
>another problem. Consider the statement: <snip>


Yes, it does. But I've heard the not long ago that Standard C
practice was to disallow a default int type. I guess I have the
situation now that register <id OR typedef_name> <id>; always sees the
typedef because my scanner does a lookup and changes an ID token to
<typedef> token when a <typedef> is present (the whole reason for my
problem).


> I hope this helps. Good Luck.


Very much so. Thanks Robert for your feedback and idea!


Bill Auerbach

Return to the comp.compilers page.
Search the comp.compilers archives again.