+//-----------------------------------------------------------------------------
+// wxHtmlPageBreakCell
+//-----------------------------------------------------------------------------
+
+// Since html isn't a page-layout language, it doesn't support page
+// page breaks directly--that requires CSS2 support. But a page-break
+// facility is handy, and has been requested more than once on the
+// mailing lists. This wxHtml tag handler implements just enough of
+// CSS2 to support a page break by recognizing only
+// <div style="page-break-before:always">
+//
+// wxHtml maintains page breaks in wxHtmlPrintout::m_PageBreaks. The
+// tag handler below adds appropriate offsets to that array member.
+// wxHtmlDCRenderer::Render() accesses that array and makes a new page
+// begin after each page-break tag.
+
+// The page-break handler does all its work in AdjustPagebreak(). For
+// all tag handlers, that function adjusts the page-break position.
+// For other tags, it determines whether the html element can fit on
+// the remainder of the page; if it cannot fit, but must not be split,
+// then the function moves the page break provided in the argument up,
+// and returns 'true' to inform the caller that the argument was
+// modified.
+//
+// Due to its special purpose, the page-break facility differs from
+// other tags. It takes up no space, but it behaves as though there is
+// never enough room to fit it on the remainder of the page--it always
+// forces a page break. Therefore, unlike other elements that trigger
+// a page break, it would never 'fit' on the following page either.
+// Therefore it's necessary to compare each pagebreak candidate to the
+// array wxHtmlPrintout::m_PageBreaks of pagebreaks already set, and
+// set a new one only if it's not in that array.
+
+class wxHtmlPageBreakCell : public wxHtmlCell
+{
+public:
+ wxHtmlPageBreakCell() {}
+
+ bool AdjustPagebreak(int* pagebreak,
+ int* known_pagebreaks = NULL,
+ int number_of_pages = 0) const;
+ void Draw(wxDC& WXUNUSED(dc),
+ int WXUNUSED(x), int WXUNUSED(y),
+ int WXUNUSED(view_y1), int WXUNUSED(view_y2),
+ wxHtmlRenderingInfo& WXUNUSED(info)) {}
+
+private:
+ DECLARE_NO_COPY_CLASS(wxHtmlPageBreakCell)
+};
+
+// Comparison routine for bsearch into an int* array of pagebreaks.
+extern "C" int wxCMPFUNC_CONV wxInteger_compare(void const* i0, void const* i1)
+{
+ return *(int*)i0 - *(int*)i1;
+}
+
+bool wxHtmlPageBreakCell::AdjustPagebreak(int* pagebreak, int* known_pagebreaks, int number_of_pages) const
+{
+ // When we are counting pages, 'known_pagebreaks' is non-NULL.
+ // That's the only time we change 'pagebreak'. Otherwise, pages
+ // were already counted, 'known_pagebreaks' is NULL, and we don't
+ // do anything except return false.
+ //
+ // We also simply return false if the 'pagebreak' argument is
+ // less than (vertically above) or the same as the current
+ // vertical position. Otherwise we'd be setting a pagebreak above
+ // the current cell, which is incorrect, or duplicating a
+ // pagebreak that has already been set.
+ if(NULL == known_pagebreaks || *pagebreak <= m_PosY)
+ {
+ return false;
+ }
+
+ // m_PosY is only the vertical offset from the parent. The pagebreak
+ // required here is the total page offset, so m_PosY must be added
+ // to the parent's offset and height.
+ int total_height = m_PosY + GetParent()->GetPosY() + GetParent()->GetHeight();
+
+ // Search the array of pagebreaks to see whether we've already set
+ // a pagebreak here. The standard bsearch() function is appropriate
+ // because the array of pagebreaks through known_pagebreaks[number_of_pages]
+ // is known to be sorted in strictly increasing order. '1 + number_of_pages'
+ // is used as a bsearch() argument because the array contains a leading
+ // zero plus one element for each page.
+ int* where = (int*) bsearch(&total_height, known_pagebreaks,
+ 1 + number_of_pages, sizeof(int),
+ wxInteger_compare);
+ // Add a pagebreak only if there isn't one already set here.
+ if(NULL != where)
+ {
+ return false;
+ }
+ else
+ {
+ *pagebreak = m_PosY;
+ return true;
+ }
+}