Re: regexp question

[Philip Nienhuis]

Sergei Steshenko wrote:
> ----- Original Message -----
> > From: Philip Nienhuis<address@hidden>
> > To: William Krekeler<address@hidden>
> > Cc: "address@hidden"<address@hidden>; address@hidden
> > Sent: Tuesday, December 6, 2011 7:52 PM
> > Subject: Re: regexp question
> >
> > Sergei, Wiliam,
> >
> > 2 answers in one post:
> >
> > Sergei Steshenko wrote:
> > >   I guess you need 'aa' surrounded by not 'a'. Octave uses
> > PCRE; I am not familiar with nuances of Octave PCRE usage; in Perl I would write
> > the regular expression this way:
> > >   [^a]aa[^a]
> > >
> > >   and if/when it matches, it returns pointer to the character preceding the
> > 'aa' substring, i.e. in case of 'baab' it should return pointer
> > to the first 'b'.
> >
> > Thanks, Sergei. I already tried this and found it'll work, but unfortunately
> > not in a more complicated situation:
> >
> > octave:35>  tststr3 = 'aa aaaaa baa'     ## Patterns at start&
> > end
> > tststr3 = aa aaaaa baa
> > octave:36>  regexp (tststr3, "[^a]aa[^a]")
> > ans = [](1x0)                           ## Hey......
> >
> > but
> > octave:41>  tststr4 = ' aa aaaaa baa '   ## Note spaces at start and
> > end
> > tststr4 =  aa aaaaa baa
> > octave:42>  regexp (tststr4, "[^a]aa[^a]")
> > ans =
> >      1   11
> >
> > ... so it doesn't catch the pattern at start and end of line.
> [snip]
>
> I still suggest the Perl regular expressions tutorials/documentation I gave 
> links to.
>
> Straightforwardly the regular expression can be extended to (in Perl syntax) :
>
> (^|[^a])(aa)([^a]|$)
> #  $1    $2    $3
> .
>
> Not inside character class '^' means line beginning, and '$' means line end.
>
> In Perl terms the 'aa' part you are interest in is in $2.

Thank you, Sergei.

How do I get $2?

octave-3.5.0+:1> tststr3 = 'aa aaaaa baa'  # No spaces at ends
tststr3 = aa aaaaa baa
octave-3.5.0+:2> regexp (tststr3, "(^|[^a])(aa)([^a]|$)")
ans =
    1   10

octave-3.5.0+:3> tststr3(1) 

ans = a 

octave-3.5.0+:4> tststr3(10) 

ans = b

... so there's some extra interpretation involved to get the proper 
position. (Little wonder as line beginnings/-ends have no length.)


Anyway, I think a regexp() solution is doomed here as its execution time 
is -currently- excessive (see my previous post).
A while ago Rik wrote that regexprep() would be in the order of 20 X 
slower than strrep. The script in my previous post confirms this 
relative slowness of regexp vs. compiled script functions.

In conclusion, I think I'll try to cook up something with strfind().

Philip