On Wed, Jun 4, 2008 at 11:43 PM, Robert Cummings <robert@xxxxxxxxxxxxx>
wrote:
> On Wed, 2008-06-04 at 23:20 -0400, Nathan Nobbe wrote:
> >
> > i repeated your test using the time program and splitting the script into
> 2,
> > one for each strpos and stripos, to find similar results. imo, there is
> no
> > need for 2 comparisons for case-insensitive searches, because both
> arguments
> > can be converted to a single case prior to the search. obviously, there
> is
> > a small amount of overhead there the case-sensitive search is
> unencumbered
> > by. i guess i never sat down and thought about how that algorithm would
> > work (case-sensitive) =/.
> >
> > thanks for the tips rob. sorry to bother you richard.
>
> You would do two comparisons... why incur the overhead of a conversion
> if one is not necessary.
because it simplifies the algorithm, there is no need for conditional logic.
> First you do case sensitive match, if that
> fails then you try the alternative version comparison. It is inefficient
> to perform 2 conversions and a single comparison in contrast.
3 operations vs. 1 or potentially 2, sure.
> Similarly,
> it's very inefficient to convert two entire strings then perform a
> comparison.
then they could be converted one at a time as the strings were traversed to
increase efficiency.
> If the first characters differ then conversion of the rest
> of the strings was pointless.
good point.
> This is basic algorithms in computer
> science.
>
you really know how to rub it in there rob. but i was looking at the
implementation in the php code, looks like somebody likes my idea (this code
found in ext/standard/string.c). on the second line the haystack is
converted to lower case[1], then if it passes a couple of checks, the needle
is converted to lower case[2], and lastly the comparison is performed[3].
there is no logic to check both cases.
(i have placed a star beside the statements ive referred to).
...
haystack_dup = estrndup(haystack, haystack_len);
*[1] php_strtolower(haystack_dup, haystack_len);
if (Z_TYPE_P(needle) == IS_STRING) {
if (Z_STRLEN_P(needle) == 0 || Z_STRLEN_P(needle) > haystack_len) {
efree(haystack_dup);
RETURN_FALSE;
}
needle_dup = estrndup(Z_STRVAL_P(needle), Z_STRLEN_P(needle));
*[2] php_strtolower(needle_dup, Z_STRLEN_P(needle));
*[3] found = php_memnstr(haystack_dup + offset, needle_dup,
Z_STRLEN_P(needle), haystack_dup + haystack_len);
}
...
-nathan
