From e2cef885a25967605007248cb158671b765df002 Mon Sep 17 00:00:00 2001 From: Serhiy Storchaka Date: Sat, 13 Apr 2013 22:45:04 +0300 Subject: Issue #16061: Speed up str.replace() for replacing 1-character strings. --- Objects/stringlib/replace.h | 53 +++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 53 insertions(+) create mode 100644 Objects/stringlib/replace.h (limited to 'Objects/stringlib/replace.h') diff --git a/Objects/stringlib/replace.h b/Objects/stringlib/replace.h new file mode 100644 index 00000000000..ef318ed6dd5 --- /dev/null +++ b/Objects/stringlib/replace.h @@ -0,0 +1,53 @@ +/* stringlib: replace implementation */ + +#ifndef STRINGLIB_FASTSEARCH_H +#error must include "stringlib/fastsearch.h" before including this module +#endif + +Py_LOCAL_INLINE(void) +STRINGLIB(replace_1char_inplace)(STRINGLIB_CHAR* s, STRINGLIB_CHAR* end, + Py_UCS4 u1, Py_UCS4 u2, Py_ssize_t maxcount) +{ + *s = u2; + while (--maxcount && ++s != end) { + /* Find the next character to be replaced. + + If it occurs often, it is faster to scan for it using an inline + loop. If it occurs seldom, it is faster to scan for it using a + function call; the overhead of the function call is amortized + across the many characters that call covers. We start with an + inline loop and use a heuristic to determine whether to fall back + to a function call. */ + if (*s != u1) { + int attempts = 10; + /* search u1 in a dummy loop */ + while (1) { + if (++s == end) + return; + if (*s == u1) + break; + if (!--attempts) { + /* if u1 was not found for attempts iterations, + use FASTSEARCH() or memchr() */ +#if STRINGLIB_SIZEOF_CHAR == 1 + s++; + s = memchr(s, u1, end - s); + if (s == NULL) + return; +#else + Py_ssize_t i; + STRINGLIB_CHAR ch1 = (STRINGLIB_CHAR) u1; + s++; + i = FASTSEARCH(s, end - s, &ch1, 1, 0, FAST_SEARCH); + if (i < 0) + return; + s += i; +#endif + /* restart the dummy loop */ + break; + } + } + } + *s = u2; + } +} -- cgit v1.2.3