feat(string): Add generic string startsWith/endsWith implementations #1898
Conversation
slurmlord
changed the title
|
|
||
| const size_t strlen = strlen_t(str); | ||
| const size_t smallen = strlen_t(smaller); | ||
| if (strlen < smallen) |
There was a problem hiding this comment.
Updated the comment (originally from the AsciiString implementation). The check needs to be < so that startsWith("hello", "hello) == true which is the current behavior.
There was a problem hiding this comment.
Right. I read that wrong and original code is also like that. It is good to preserve the original implementation as much as possible because any discrepancy here could potentially cause mismatching in logic.
| } | ||
|
|
||
| inline int strncmp_t(const char *str1, const char *str2, size_t maxcount) { return ::strncmp(str1, str2, maxcount); }; | ||
| inline int strncmp_t(const wchar_t *str1, const wchar_t *str2, size_t maxcount) { return ::wcsncmp(str1, str2, maxcount); }; |
There was a problem hiding this comment.
Can you measure if this is really faster?
Generally I prefer to have same code paths on all platforms, so that there are less potential discrepancies in behavior.
There was a problem hiding this comment.
Tested with comparing 50 char strings.
When both strings were random, the strnicmp_t in this file was about 1/3 faster. Typically ~2100 vs ~3200 microseconds for 100000 iterations.
When testing the worst case, comparing one string with itself, the built-in strnicmp was about 2x faster. In the area of ~8500 vs ~13600 microseconds.
I guess it comes down to if we expect strings to match or not :) But I'll remove these specializations for consistency's sake .
There was a problem hiding this comment.
I wonder if the original impls use simd to load multiple bytes in one go.
xezon
added
Minor
3 participants
This PR adds the string functions
startsWith,startsWithNoCase,endsWithandendsWithNoCasetoWWLib/stringex.h. Supports bothchar*andwchar_t*.Replaced existing functions in AsciiString with calls to the corresponding new implementations:
AsciiString::startsWithAsciiString::startsWithNoCaseAsciiString::endsWithAsciiString::endsWithNoCaseAlso added the same functions for UnicodeString.
This was split out from #1594 where the need for a
endsWithNoCaseforwchar_tsurfaced.