Re: LL(1) parser table

Hi,
I beleive, it is not a LL(1) grammar. LR is a better parsing technique
b'cos it is stateful, meaning the decision about which branch to take
for such a situation (wether to use first or follow) is decided based on
the state of the parser, how much of the parsing is done. It is not at
all difficult to create an LR parser for the same grammar.


Need any help in creating a LR parser, do get back.


Kaarthik


lee wrote:


> Hi all,
>
> I am trying construct a LL1 parser table for the following grammar.
>
> S -> E
> E -> ++E | E++ | E-E | E/E | id
>
> where symbols have usual meanings.
>
> However in the final table this grammar seems to be ambiguous. I
> actually first transformed the above into this disambiguous form.
>
> S -> E
> E -> E-T | T
> T -> T/F | F
> F -> F++ | G
> G -> ++E | id
>
> And after that I removed all left-recursions.
>
> S -> E
> E -> TE'
> E'-> -TE'
> E'-> epsilon
> T -> FT'
> T'-> /FT'
> T'-> epsilon
> F -> GF'
> F'-> ++F'
> F'-> epsilon
> G -> ++E
> G -> id
>
> Still after all this grammar is ambiguous. Because F' FOLLOW set is
> {/,-,++,$} and there are two entries for [F',++] comination.
>
> Does this means the above grammar is not suitable for LL1 parsing?

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