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

Review comments

Browse Source
This commit is contained in:
jpirnay
2026-04-20 13:28:56 +02:00
parent 3fb9ac9e29
commit 29284c85c9
10 changed files with 129 additions and 127 deletions
Download Patch File Download Diff File Expand all files Collapse all files
lib/Epub/Epub/parsers/ChapterHtmlSlimParser.cpp
+12 -7
View File
@@ -593,9 +593,7 @@ void XMLCALL ChapterHtmlSlimParser::startElement(void* userData, const XML_Char*
(self->currentPageNextY + totalImageHeightWithSpacing > self->viewportHeight)) {
LOG_DBG("EHP", "Image page break: currentY=%d needed=%d viewportH=%d", self->currentPageNextY,
totalImageHeightWithSpacing, self->viewportHeight);
const uint32_t byteOff =
self->activeParser ? static_cast<uint32_t>(XML_GetCurrentByteIndex(self->activeParser)) : 0;
self->paragraphLutPerPage.push_back({byteOff, self->xpathParagraphIndex});
self->paragraphLutPerPage.push_back({self->lastBodyChildByteOffset, self->xpathParagraphIndex});
self->completePageFn(std::move(self->currentPage));
self->completedPageCount++;
self->currentPage.reset(new Page());
@@ -689,8 +687,16 @@ void XMLCALL ChapterHtmlSlimParser::startElement(void* userData, const XML_Char*
// check so that hidden <p> elements are still counted, matching ChapterXPathIndexer's
// counting (pure XML, no CSS). This ensures paragraph indices in the section cache LUT
// align with KOReader's crengine XPath indices.
if (self->xpathBodyDepth >= 0 && self->depth == self->xpathBodyDepth + 1 && strcmp(name, "p") == 0) {
self->xpathParagraphIndex++;
// At the same time, record the byte offset of every direct-body-child element start:
// the forward mapper's partial-parse heuristic requires the seek hint to land on a
// body-child boundary, otherwise partialBaseDepth can misidentify wrapped paragraphs.
if (self->xpathBodyDepth >= 0 && self->depth == self->xpathBodyDepth + 1) {
if (self->activeParser) {
self->lastBodyChildByteOffset = static_cast<uint32_t>(XML_GetCurrentByteIndex(self->activeParser));
}
if (strcmp(name, "p") == 0) {
self->xpathParagraphIndex++;
}
}
if (matches(name, SKIP_TAGS, NUM_SKIP_TAGS)) {
@@ -1438,8 +1444,7 @@ ParsedText::LineProcessResult ChapterHtmlSlimParser::addLineToPage(std::shared_p
}
if (currentPageNextY + lineHeight > viewportHeight) {
const uint32_t byteOff = activeParser ? static_cast<uint32_t>(XML_GetCurrentByteIndex(activeParser)) : 0;
paragraphLutPerPage.push_back({byteOff, xpathParagraphIndex});
paragraphLutPerPage.push_back({lastBodyChildByteOffset, xpathParagraphIndex});
completePageFn(std::move(currentPage));
completedPageCount++;
currentPage.reset(new Page());
lib/Epub/Epub/parsers/ChapterHtmlSlimParser.h
+7 -1
View File
@@ -92,9 +92,15 @@ class ChapterHtmlSlimParser {
// without reparsing, and current page can generate an XPath without reparsing.
uint16_t xpathParagraphIndex = 0; // current <p> sibling index (1-based)
int xpathBodyDepth = -1; // depth of the <body> element (-1 = not yet seen)
// Byte offset of the most recent direct-body-child element start (any tag at xpathBodyDepth+1).
// Recorded at the same depth condition that increments xpathParagraphIndex, so the stored
// offset is guaranteed to land on a body-child element boundary. This keeps the XPath forward
// mapper's partial-parse heuristic reliable for wrapped chapters: without this, the offset
// could point mid-way into a nested <div>/<section>, which confuses partialBaseDepth.
uint32_t lastBodyChildByteOffset = 0;
struct ParagraphLutEntry {
uint32_t xhtmlByteOffset; // Expat byte offset at page break — used to seek near target paragraph
uint32_t xhtmlByteOffset; // byte offset of most recent body-child element start at page break
uint16_t paragraphIndex; // 1-based <p> index at page completion
};
std::vector<ParagraphLutEntry> paragraphLutPerPage; // deep LUT: one entry per page