From: james.harris.1@nospicedham.gmail.com
On 29/06/2017 23:40, Rod Pemberton wrote:
> On Thu, 29 Jun 2017 09:57:41 +0100
> James Harris wrote:
>
>> On 29/06/2017 01:40, Rod Pemberton wrote:
>>> On Wed, 28 Jun 2017 11:21:21 +0100
>>> James Harris wrote:
>>>> On 28/06/2017 04:31, Rod Pemberton wrote:
...
> To avoid the function calls for multiple character translations, you
> can create a lookup table.
>
> #define ARY 256
> char lower[ARY];
>
> /* In main() or as a void func(void) */
> int i;
> for(i=0;i {
> lower[i]=tolower(i);
> }
That's a good way to go. It's fast, simple and would work with any
character set that will fit in 8 bits.
...
> When "if (x == MIGHT)", the XOR result is 0x20, meaning that the two
> characters are either an upper- and lower-case alphabetic which are the
> same when lower-cased, or they're two different graphics characters
> 0x20 apart.
That's far from obvious but I have to say that for ASCII it seems to be
correct. And it may work for EBCDIC too (with an XOR of 0x40).
...
> Using the same identity/constraint, your 'x==MIGHT' code should look
> something like this:
>
>> int jh_stricmp(char *s, char *t)
>> {
>> int i;
>> char x; /* The xor of each pair of characters */
>>
>> for (i = 0; s[i]; ++i)
>> {
>> x = s[i] ^ t[i];
>> if (x == 0) /* The chars match (in this case, they are
>> identical) */
>> continue;
>> if (x == MIGHT) /* The chars might match (upper & lower case) */
> if (alpha[(int)s[i]]) /* They do match */
> continue;
>> break;
>> }
>> return s[i] - t[i];
>> }
I could understand an (unsigned) cast because chars might be signed but
why the (int) cast? Wouldn't it just sign-extend a negative number but
still leave it negative?
...
> Sigh, the true value for Linux is___() ctype.h functions doesn't fit
> into a char ... When cast to a char or masked with 0xFF, the true value
> is zero (false). Unbelievable! So, it needs two logical-nots !! to
> fix. A simple assignment works for tolower(). That tells me someone
> once knew what they were doing. It also makes you wonder if anyone
> coding for Linux had any C experience prior to Linux. Who would
> intentionally cause a failure for a function intended to be used to
> fill a character array for use as a lookup table? This is just F-word
> insane. They must've broken every C program in existence that use
> ctype.h functions, which was not written specifically for Linux. I
> guess it's a good thing I don't usually use the ctype.h functions.
>
Understood.
--
James Harris
--- SoupGate-Win32 v1.05
* Origin: you cannot sedate... all the things you hate (1:229/2)
|
