base32/64: code shrink

function                                             old     new   delta
decode_base64                                        178     173      -5
decode_base32                                        217     186     -31
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 0/2 up/down: 0/-36)             Total: -36 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
This commit is contained in:
Denys Vlasenko 2020-11-28 12:33:50 +01:00
parent 7467e905e5
commit f140b49f5f

View File

@ -92,8 +92,8 @@ const char* FAST_FUNC decode_base64(char **pp_dst, const char *src)
{
char *dst = *pp_dst;
unsigned ch = 0;
unsigned t;
int i = 0;
int t;
while ((t = (unsigned char)*src) != '\0') {
src++;
@ -117,11 +117,11 @@ const char* FAST_FUNC decode_base64(char **pp_dst, const char *src)
continue;
ch = (ch << 6) | t;
if (++i == 4) {
i = (i + 1) & 3;
if (i == 0) {
*dst++ = (char) (ch >> 16);
*dst++ = (char) (ch >> 8);
*dst++ = (char) ch;
i = 0;
if (ch & 0x1000000) { /* was last input char '='? */
dst--;
if (ch & (0x1000000 << 6)) /* was it "=="? */
@ -140,9 +140,9 @@ const char* FAST_FUNC decode_base64(char **pp_dst, const char *src)
const char* FAST_FUNC decode_base32(char **pp_dst, const char *src)
{
char *dst = *pp_dst;
uint64_t ch = 0;
int64_t ch = 0;
unsigned t;
int i = 0;
int t;
while ((t = (unsigned char)*src) != '\0') {
src++;
@ -150,44 +150,52 @@ const char* FAST_FUNC decode_base32(char **pp_dst, const char *src)
/* "if" forest is faster than strchr(bb_uuenc_tbl_base32, t) */
if (t >= '2' && t <= '7')
t = t - '2' + 26;
else if ((t|0x20) >= 'a' && (t|0x20) <= 'z')
t = (t|0x20) - 'a';
else if (t == '=' && i > 1)
t = 0;
else
else {
t = (t | 0x20) - 'a';
if (t > 25)
//TODO: add BASE64_FLAG_foo to die on bad char?
continue;
continue;
}
ch = (ch << 5) | (unsigned)t; /* cast prevents pointless sign-extension of t */
if (++i == 8) {
/* testcase:
* echo ' 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18' | base32 | base32 -d
* IOW, decoding of
* EAYSAMRAGMQDIIBVEA3CANZAHAQDSIBRGAQDCMJAGEZCAMJTEAYTIIBRGUQDCNRAGE3SAMJYBI==
* ====
* should correctly stitch together the tail.
*/
if (t == 0) {
const char *s = src;
while (*--s == '=')
t--;
}
ch = (ch << 5) | t;
i = (i + 1) & 7;
if (i == 0) {
*dst++ = (char) (ch >> 32);
if (src[-1] == '=') /* was last input char '='? */
goto tail;
*dst++ = (char) (ch >> 24);
*dst++ = (char) (ch >> 16);
*dst++ = (char) (ch >> 8);
*dst++ = (char) ch;
i = 0;
if (t < 0) /* was last input char '='? */
break;
}
}
if (t < 0) /* was last input char '='? */
/* -t is the count of =, must be 1, 3, 4 or 6 */
dst -= (-t + 1) * 2 / 3; /* discard last 1, 2, 3 or 4 bytes */
*pp_dst = dst;
/* i is zero here if full 8-char block was decoded */
return src - i;
tail:
{
const char *s = src;
while (*--s == '=')
i++;
/* Why duplicate the below code? Testcase:
* echo ' 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18' | base32 | base32 -d
* IOW, decoding of
* EAYSAMRAGMQDIIBVEA3CANZAHAQDSIBRGAQDCMJAGEZCAMJTEAYTIIBRGUQDCNRAGE3SAMJYBI==
* ====
* must correctly stitch together the tail, must not overwrite
* the tail before it is analyzed! (we can be decoding in-place)
* Else testcase fails, prints trailing extra NUL bytes.
*/
*dst++ = (char) (ch >> 24);
*dst++ = (char) (ch >> 16);
*dst++ = (char) (ch >> 8);
*dst++ = (char) ch;
dst -= (i+1) * 2 / 3; /* discard last 1, 2, 3 or 4 bytes */
}
*pp_dst = dst;
return src;
}
#endif