Skip to content

bpo-41972: Tweak fastsearch.h string search algorithms #27091

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
ambv merged 21 commits into from
Jul 19, 2021
Merged

bpo-41972: Tweak fastsearch.h string search algorithms #27091

Changes from 1 commit
Commits
Show all changes
21 commits
Select commit Hold shift + click to select a range
90a6a0c
Don't force inlining
sweeneyde Jul 6, 2021
19fa661
Use a Boyer-Moore skip table
sweeneyde Jul 7, 2021
d2219d5
Switch to two-way in more cases
sweeneyde Jul 8, 2021
3b8760f
Tighter comments
sweeneyde Jul 9, 2021
5d86b8d
Fix the direction of the cutoff inequality :)
sweeneyde Jul 10, 2021
2e9a9b1
Separate a gap loop
sweeneyde Jul 10, 2021
c84314c
update cutoffs
sweeneyde Jul 10, 2021
7517c60
tweak cutoffs, fix i - cut + 1 jump
sweeneyde Jul 10, 2021
e5ef33b
Refactor into smaller functions
sweeneyde Jul 11, 2021
a01fb70
Focus the table on the last character of the pattern, giving a tighte…
sweeneyde Jul 11, 2021
6be0c32
tweak cutoffs
sweeneyde Jul 11, 2021
f9fe807
📜🤖 Added by blurb_it.
blurb-it[bot] Jul 12, 2021
25f9b56
Fix sanitizer warnings
sweeneyde Jul 12, 2021
d3f8438
Merge branch 'fastersearch' of https://github.com/sweeneyde/cpython i…
sweeneyde Jul 12, 2021
4379ea6
Refactor for sanitizer
sweeneyde Jul 12, 2021
e336b79
Don't goto declarations
sweeneyde Jul 12, 2021
b8df3e1
Emphasize to the compiler that the loop is a loop
sweeneyde Jul 13, 2021
869f7d0
Use Py_LOCAL_INLINE
sweeneyde Jul 15, 2021
747a39a
Use Py_LOCAL_INLINE again
sweeneyde Jul 15, 2021
72143ef
Tweak thresholds based on most recent comparison.
sweeneyde Jul 16, 2021
4d7d102
Fix greater/less than sign.
sweeneyde Jul 16, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
Next Next commit
Refactor for sanitizer
  • Loading branch information
@sweeneyde
sweeneyde committed Jul 12, 2021
commit 4379ea689c268d40e4b6ea83a025469b1a8d8866
56 changes: 29 additions & 27 deletions Objects/stringlib/fastsearch.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -376,29 +376,18 @@ STRINGLIB(_two_way)(const STRINGLIB_CHAR *haystack, Py_ssize_t len_haystack,
Py_ssize_t memory = 0;
periodicwindowloop:
while (window_last < haystack_end) {
assert(memory == 0);
LOG_LINEUP();
Py_ssize_t shift = table[(*window_last) & TABLE_MASK];
if (shift > 0 && memory > 0) {
// A mismatch has been identified to the right of
// where i starts, so we can jump at least as far as
// if the mismatch occurred on the first comparison.
Py_ssize_t memory_shift = Py_MAX(cut, memory) - cut + 1;
LOG("Skip with Memory.\n");
window_last += Py_MAX(shift, memory_shift);
LOG_LINEUP();
memory = 0;
shift = table[(*window_last) & TABLE_MASK];
}
while (shift > 0 && window_last < haystack_end) {
window_last += shift;
if (shift) {
LOG("Horspool skip.\n");
window_last += shift;
shift = table[(*window_last) & TABLE_MASK];
LOG_LINEUP();
}
if (window_last >= haystack_end) {
break;
continue;
}
no_shift:
const STRINGLIB_CHAR *const window = window_last - len_needle + 1;
assert((window[len_needle - 1] & TABLE_MASK) ==
(needle[len_needle - 1] & TABLE_MASK));
Py_ssize_t i = Py_MAX(cut, memory);
for (; i < len_needle; i++) {
if (needle[i] != window[i]) {
Expand All @@ -413,7 +402,22 @@ STRINGLIB(_two_way)(const STRINGLIB_CHAR *haystack, Py_ssize_t len_haystack,
LOG("Left half does not match.\n");
window_last += period;
memory = len_needle - period;
goto periodicwindowloop;
if (window_last >= haystack_end) {
return -1;
}
shift = table[(*window_last) & TABLE_MASK];
if (shift) {
// A mismatch has been identified to the right
// of where i will next start, so we can jump
// at least as far as if the mismatch occurred
// on the first comparison.
Py_ssize_t mem_jump = Py_MAX(cut, memory) - cut + 1;
LOG("Skip with Memory.\n");
memory = 0;
window_last += Py_MAX(shift, mem_jump);
goto periodicwindowloop;
}
goto no_shift;
}
}
LOG("Found a match!\n");
Expand All @@ -429,16 +433,14 @@ STRINGLIB(_two_way)(const STRINGLIB_CHAR *haystack, Py_ssize_t len_haystack,
while (window_last < haystack_end) {
LOG_LINEUP();
Py_ssize_t shift = table[(*window_last) & TABLE_MASK];
while (shift > 0 && window_last < haystack_end) {
window_last += shift;
if (shift) {
LOG("Horspool skip.\n");
window_last += shift;
shift = table[(*window_last) & TABLE_MASK];
LOG_LINEUP();
}
if (window_last >= haystack_end) {
break;
continue;
}
const STRINGLIB_CHAR *window = window_last - len_needle + 1;
const STRINGLIB_CHAR *const window = window_last - len_needle + 1;
assert((window[len_needle - 1] & TABLE_MASK) ==
(needle[len_needle - 1] & TABLE_MASK));
for (Py_ssize_t i = cut; i < gap_jump_end; i++) {
if (needle[i] != window[i]) {
LOG("Early right half mismatch: jump by gap.\n");
Expand Down