RE: bit pattern matching
I'm afraid my similarity.e routine has nothing to do with the problem, if I
understood well what Allen explained.
I then started juggling these ideas:
1) Using (x xor_bits (x -1)) and_bits x, to determine the maximum power of 2
that exactly divides x.
2) Invert the order of the bits, i.e, 00110101 -> 10101100.
3) x and_bits (x + x): if not zero, there are 2 or more contiguous 1 bits.
, but I had no luck.
Maybe Allen can provide us with a more detailed example of the thing he
wants to do, or better a more extensive view of it.
Regards.
----- Original Message -----
From: Kat <gertie at visionsix.com>
To: <EUforum at topica.com>
Sent: Thursday, May 27, 2004 4:18 AM
Subject: Re: bit pattern matching
>
>
> On 26 May 2004, at 23:42, Allen Robnett wrote:
>
> >
> > From: "Ricardo M. Forno" <rforno at uyuyuy.com>
> > Subject: RE: bit manipulation
> >
> >
> > <<If you are searching for an exact match (lets say 100110010), and you
have it
> > in a character string (a sequence), you migth try this:>>
> >
> > The matching that I want to do is probably a little unusual.
> > First I want to align the most significant bits of the two samples.
> > Then I want to determine the number of bits that match, up to the first
> > mismatch, but not beyond. The ratio of the number of matching bits to
the length
> > of the primary sample should be a useful measure of the degree of match.
>
> Oh, then you want to search email records for DCuny's difference code,
> which i asked for some 4 or 5 yrs ago. It's a great starting point for
tweaking
> to your specific application of this sort. It works as he submitted it to
the
> email list, but never made it to the user contribs page, afaik. I looked
for it
> here for an hour, since i did significant tweaking, speed tests, and
verification
> tests for data loss, but i'm afraid since no one was interested back then,
> that's some of the code i deleted.
>
> Look up similarity.e by R. M. Forno too.
>
> Kat
>
>
>
>
|
Not Categorized, Please Help
|
|
