diff options
author | naruse <naruse@b2dd03c8-39d4-4d8f-98ff-823fe69b080e> | 2008-03-05 13:54:36 +0000 |
---|---|---|
committer | naruse <naruse@b2dd03c8-39d4-4d8f-98ff-823fe69b080e> | 2008-03-05 13:54:36 +0000 |
commit | 59091322fd1344e46d0e17d187461937ac9393d6 (patch) | |
tree | 792f7d97e9e9d5dc7843fee263a2516eb4ffb4c0 /string.c | |
parent | 25d3c61306bc6b57de3ffa58c4f109adaa7c87fc (diff) | |
download | ruby-59091322fd1344e46d0e17d187461937ac9393d6.tar.gz ruby-59091322fd1344e46d0e17d187461937ac9393d6.tar.xz ruby-59091322fd1344e46d0e17d187461937ac9393d6.zip |
* string.c (is_utf8_lead_byte, count_utf8_lead_bytes_with_ulong):
defined for UTF-8 optimization.
* string.c (str_strlen): use is_utf8_lead_byte and
count_utf8_lead_bytes_with_ulong.
* string.c (str_utf8_nth) ditto.
git-svn-id: http://svn.ruby-lang.org/repos/ruby/trunk@15695 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Diffstat (limited to 'string.c')
-rw-r--r-- | string.c | 70 |
1 files changed, 35 insertions, 35 deletions
@@ -755,6 +755,24 @@ rb_enc_strlen_cr(const char *p, const char *e, rb_encoding *enc, int *cr) return c; } +#ifdef NONASCII_MASK +#define is_utf8_lead_byte(c) (((c)&0xC0) != 0x80) +static inline const long +count_utf8_lead_bytes_with_ulong(const unsigned long *s) +{ + unsigned long d = *s; + d |= ~(d>>1); + d >>= 6; + d &= NONASCII_MASK >> 3; + d += (d>>8); + d += (d>>16); +#if NONASCII_MASK == 0x8080808080808080UL + d += (d>>32); +#endif + return (long)(d&0xF); +} +#endif + static long str_strlen(VALUE str, rb_encoding *enc) { @@ -774,26 +792,19 @@ str_strlen(VALUE str, rb_encoding *enc) const VALUE lowbits = sizeof(unsigned long) - 1; s = (const unsigned long*)(~lowbits & ((VALUE)p + lowbits)); t = (const unsigned long*)(~lowbits & (VALUE)e); - for (len=0; p<(const char *)s; p++) { - if (((*p)&0xC0) != 0x80) len++; + while (p < (const char *)s) { + if (is_utf8_lead_byte(*p)) len++; + p++; } while (s < t) { - unsigned long d = *s; - d = ~d | (d<<1); - d &= NONASCII_MASK; - d >>= 7; - d += (d>>8); - d += (d>>16); -#if NONASCII_MASK == 0x8080808080808080UL - d = d + (d>>32); -#endif - len += (long)(d&0xF); + len += count_utf8_lead_bytes_with_ulong(s); s++; } - p = (const char *)t; + p = (const char *)s; } - for (; p<e; p++) { - if (((*p)&0xC0) != 0x80) len++; + while (p < e) { + if (is_utf8_lead_byte(*p)) len++; + p++; } return len; } @@ -1162,33 +1173,22 @@ str_utf8_nth(const char *p, const char *e, int nth) const VALUE lowbits = sizeof(unsigned long) - 1; s = (const unsigned long*)(~lowbits & ((VALUE)p + lowbits)); t = (const unsigned long*)(~lowbits & (VALUE)e); - for (; p<(const char *)s && 0<nth; p++) { - if (((*p)&0xC0) != 0x80) nth--; + while (p < (const char *)s) { + if (is_utf8_lead_byte(*p)) nth--; + p++; } while (s < t) { - unsigned long d = *s++; - d = ~d | (d<<1); - d &= NONASCII_MASK; - d >>= 7; - d += (d>>8); - d += (d>>16); -#if NONASCII_MASK == 0x8080808080808080UL - d += (d>>32); -#endif - nth -= (long)(d&0xF); - if (nth < 8) { - t = s; - break; - } + nth -= count_utf8_lead_bytes_with_ulong(s); + if (nth < sizeof(long)) break; + s++; } - p = (char *)t; + p = (char *)s; } if (0 < nth) { while (p < e) { - if (((*p)&0xC0) != 0x80) { + if (is_utf8_lead_byte(*p)) { nth--; - if (nth < 0) - break; + if (nth < 0) break; } p++; } |