Re: Getting rid of backups in flex.

(I'm responding late. Sorry if someone already gave similar explanations.)


" <casse@netcourrier.com>"@free.fr writes:
> "Robert Minsk" <robertminsk@yahoo.com> a écrit :
>
> > Given the following rules I am still getting 1 backup I'm trying to get
> > rid of.
> >
> > DIGIT [[:digit:]]
> > EXP [Ee][+-]?{DIGIT}+
> > EXPERR [Ee][+-]?
> > %%
> > [-+]?{DIGIT}+ { return tInt; }
> >
> > [-+]?{DIGIT}*"."{DIGIT}+{EXP}? |
> > [-+]?{DIGIT}+"."{DIGIT}*{EXP}? |
> > [-+]?{DIGIT}+{EXP} { return tFloat }
> >
> > [-+]?{DIGIT}*"."{DIGIT}+{EXPERR}? |
> > [-+]?{DIGIT}+"."{DIGIT}*{EXPERR}? |
> > [-+]?{DIGIT}+{EXPERR}
> > . { return tError }
> >
> > I get:
> > State #9 is non-accepting -
> > associated rule line numbers:
> > 7 11
> > out-transitions: [ 0-9 ]
> > jam-transitions: EOF [ \000-/ :-\377 ]
> >
> > 1 backing up (non-accepting) states.
> >
> > I have not been able to fix this backup. Can you get rid of the one
> > backup? Is there a better way to code this?


State #9 is non-accepting because when the automaton has read
                +.
or
                -.
it is in a situation where it has to read at least a digit.


I suggest adding another error-catching regular expression, namely:
                [-+]"."


Adding the following error-catching regular expression:
                ..
(as someone suggested previously) is not correct because this
expression would take precedence on the normal regular expressions
when the input would start with a single-digit integer. For example,
if the input starts with:
                9),...
the automaton would do a match of length 2 and return `tError' instead
of matching the length 1 string "9" and returning `tInt'.


--
                Danny Dubé
                dube@IRO.UMontreal.CA

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