eric/Crosspoint
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

perf: always binary-search idref lookups in content.opf (#2433)

Browse Source 
Co-authored-by: Erica Jensen <erica@mailershaven.com>
This commit is contained in:
hooligan333
2026-07-11 22:56:28 -04:00
committed by GitHub
co-authored by Erica Jensen
parent bb06219c9c
commit 5fbc657aeb
2 changed files with 7 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files
lib/Epub/Epub/parsers/ContentOpfParser.cpp
+6 -5
View File
@@ -137,8 +137,10 @@ void XMLCALL ContentOpfParser::startElement(void* userData, const XML_Char* name
LOG_ERR("COF", "Couldn't open temp items file for reading. This is probably going to be a fatal error."); LOG_ERR("COF", "Couldn't open temp items file for reading. This is probably going to be a fatal error.");
} }
// Sort item index for binary search if we have enough items // Sort the (unconditionally-built) item index so every idref lookup uses binary
if (self->itemIndex.size() >= LARGE_SPINE_THRESHOLD) { // search. Without this, small/medium manifests fell back to an O(spine × manifest)
// linear rescan of .items.bin per itemref (up to ~200ms/item at large scale).
if (!self->itemIndex.empty()) {
std::sort(self->itemIndex.begin(), self->itemIndex.end(), [](const ItemIndexEntry& a, const ItemIndexEntry& b) { std::sort(self->itemIndex.begin(), self->itemIndex.end(), [](const ItemIndexEntry& a, const ItemIndexEntry& b) {
return a.idHash < b.idHash || (a.idHash == b.idHash && a.idLen < b.idLen); return a.idHash < b.idHash || (a.idHash == b.idHash && a.idLen < b.idLen);
}); });
@@ -284,9 +286,8 @@ void XMLCALL ContentOpfParser::startElement(void* userData, const XML_Char* name
++it; ++it;
} }
} else { } else {
// Slow path: linear scan (for small manifests, keeps original behavior) // Fallback linear scan, only reached when the index is empty (no manifest
// TODO: This lookup is slow as need to scan through all items each time. // items). The fast binary-search path above is used for all real manifests.
// It can take up to 200ms per item when getting to 1500 items.
self->tempItemStore.seek(0); self->tempItemStore.seek(0);
std::string itemId; std::string itemId;
while (self->tempItemStore.available()) { while (self->tempItemStore.available()) {
lib/Epub/Epub/parsers/ContentOpfParser.h
+1 -3
View File
@@ -32,7 +32,7 @@ class ContentOpfParser final : public Print {
HalFile tempItemStore; HalFile tempItemStore;
std::string coverItemId; std::string coverItemId;
// Index for fast idref→href lookup (used only for large EPUBs) // Index for fast idref→href lookup (binary search over .items.bin)
struct ItemIndexEntry { struct ItemIndexEntry {
uint32_t idHash; // FNV-1a hash of itemId uint32_t idHash; // FNV-1a hash of itemId
uint16_t idLen; // length for collision reduction uint16_t idLen; // length for collision reduction
@@ -41,8 +41,6 @@ class ContentOpfParser final : public Print {
std::deque<ItemIndexEntry> itemIndex; std::deque<ItemIndexEntry> itemIndex;
bool useItemIndex = false; bool useItemIndex = false;
static constexpr uint16_t LARGE_SPINE_THRESHOLD = 400;
// FNV-1a hash function // FNV-1a hash function
static uint32_t fnvHash(const std::string& s) { static uint32_t fnvHash(const std::string& s) {
uint32_t hash = 2166136261u; uint32_t hash = 2166136261u;