[Libreoffice-commits] .: Branch 'libreoffice-3-5' - 2 commits - sot/source writerfilter/source

Caolán McNamara caolan at kemper.freedesktop.org
Fri May 11 08:34:53 PDT 2012


 sot/source/sdstor/stgstrms.cxx              |  133 ++++++++++++++--
 sot/source/sdstor/stgstrms.hxx              |    4 
 writerfilter/source/dmapper/PropertyMap.cxx |  224 ++++++++++++++--------------
 3 files changed, 233 insertions(+), 128 deletions(-)

New commits:
commit ae218f73ff4eaa4191e620abee8235f73ea0e1fc
Author: Noel Power <noel.power at novell.com>
Date:   Fri May 11 09:50:31 2012 +0100

    Revert "another partial fix for bnc#693238" ( and fixes fdo#44292 )
    
    This reverts commit cd1c434c4bd4756fa1355e906cb2e8de4aae0618.
    
    Conflicts:
    
    	writerfilter/source/dmapper/PropertyMap.cxx
    
    Change-Id: I I83ae3418438aa02db744e6b6eb5a9c199938c91f
    (cherry picked from commit 2dee7d6205832a1212790ac1ab168a6b6fce686d)
    
    Signed-off-by: Caolán McNamara <caolanm at redhat.com>

diff --git a/writerfilter/source/dmapper/PropertyMap.cxx b/writerfilter/source/dmapper/PropertyMap.cxx
index 4f03839..503052f 100644
--- a/writerfilter/source/dmapper/PropertyMap.cxx
+++ b/writerfilter/source/dmapper/PropertyMap.cxx
@@ -843,130 +843,130 @@ void SectionPropertyMap::CloseSectionGroup( DomainMapper_Impl& rDM_Impl )
         if( m_nColumnCount > 0 && xSection.is())
             ApplyColumnProperties( xSection );
     }
-    //get the properties and create appropriate page styles
-    uno::Reference< beans::XPropertySet > xFollowPageStyle = GetPageStyle( rDM_Impl.GetPageStyles(), rDM_Impl.GetTextFactory(), false );
-
-    if( m_nDzaGutter > 0 )
+    else
     {
-        //todo: iGutterPos from DocProperties are missing
-        if( m_bGutterRTL )
-            m_nRightMargin += m_nDzaGutter;
-        else
-            m_nLeftMargin += m_nDzaGutter;
-    }
-    operator[]( PropertyDefinition( PROP_LEFT_MARGIN, false )) =  uno::makeAny( m_nLeftMargin  );
-    operator[]( PropertyDefinition( PROP_RIGHT_MARGIN, false )) = uno::makeAny( m_nRightMargin );
-
-    /*** if headers/footers are available then the top/bottom margins of the
-        header/footer are copied to the top/bottom margin of the page
-    */
-    CopyLastHeaderFooter( false, rDM_Impl );
-    PrepareHeaderFooterProperties( false );
-
-    const ::rtl::OUString sTrayIndex = rPropNameSupplier.GetName( PROP_PRINTER_PAPER_TRAY_INDEX );
-    if( m_nPaperBin >= 0 )
-        xFollowPageStyle->setPropertyValue( sTrayIndex, uno::makeAny( m_nPaperBin ) );
-    uno::Reference< text::XTextColumns > xColumns;
-    if( m_nColumnCount > 0 )
-        xColumns = ApplyColumnProperties( xFollowPageStyle );
-
-    //prepare text grid properties
-    sal_Int32 nHeight = 1;
-    PropertyMap::iterator aElement = find(PropertyDefinition( PROP_HEIGHT, false ));
-    if( aElement != end())
-        aElement->second >>= nHeight;
-
-    sal_Int32 nWidth = 1;
-    aElement = find(PropertyDefinition( PROP_WIDTH, false ));
-    if( aElement != end())
-        aElement->second >>= nWidth;
-
-    text::WritingMode eWritingMode = text::WritingMode_LR_TB;
-    aElement = find(PropertyDefinition( PROP_WRITING_MODE, false ));
-    if( aElement != end())
-        aElement->second >>= eWritingMode;
-
-
-
-    sal_Int32 nTextAreaHeight = eWritingMode == text::WritingMode_LR_TB ?
-    nHeight - m_nTopMargin - m_nBottomMargin :
-    nWidth - m_nLeftMargin - m_nRightMargin;
-
-    operator[]( PropertyDefinition( PROP_GRID_LINES, false )) =
-    uno::makeAny( static_cast<sal_Int16>(nTextAreaHeight/m_nGridLinePitch));
+        //get the properties and create appropriate page styles
+        uno::Reference< beans::XPropertySet > xFollowPageStyle = GetPageStyle( rDM_Impl.GetPageStyles(), rDM_Impl.GetTextFactory(), false );
 
-    sal_Int32 nCharWidth = 423; //240 twip/ 12 pt
-    //todo: is '0' the right index here?
-    const StyleSheetEntryPtr pEntry = rDM_Impl.GetStyleSheetTable()->FindStyleSheetByISTD(::rtl::OUString::valueOf(static_cast<sal_Int32>(0), 16));
-    if( pEntry.get( ) )
-    {
-        PropertyMap::iterator aElement_ = pEntry->pProperties->find(PropertyDefinition( PROP_CHAR_HEIGHT_ASIAN, false ));
-        if( aElement_ != pEntry->pProperties->end())
+        if( m_nDzaGutter > 0 )
         {
-            double fHeight = 0;
-            if( aElement_->second >>= fHeight )
-                nCharWidth = ConversionHelper::convertTwipToMM100( (long)( fHeight * 20.0 + 0.5 ));
+            //todo: iGutterPos from DocProperties are missing
+            if( m_bGutterRTL )
+                m_nRightMargin += m_nDzaGutter;
+            else
+                m_nLeftMargin += m_nDzaGutter;
+        }
+        operator[]( PropertyDefinition( PROP_LEFT_MARGIN, false )) =  uno::makeAny( m_nLeftMargin  );
+        operator[]( PropertyDefinition( PROP_RIGHT_MARGIN, false )) = uno::makeAny( m_nRightMargin );
+
+        /*** if headers/footers are available then the top/bottom margins of the
+            header/footer are copied to the top/bottom margin of the page
+          */
+        CopyLastHeaderFooter( false, rDM_Impl );
+        PrepareHeaderFooterProperties( false );
+
+        const ::rtl::OUString sTrayIndex = rPropNameSupplier.GetName( PROP_PRINTER_PAPER_TRAY_INDEX );
+        if( m_nPaperBin >= 0 )
+            xFollowPageStyle->setPropertyValue( sTrayIndex, uno::makeAny( m_nPaperBin ) );
+        uno::Reference< text::XTextColumns > xColumns;
+        if( m_nColumnCount > 0 )
+            xColumns = ApplyColumnProperties( xFollowPageStyle );
+
+        //prepare text grid properties
+        sal_Int32 nHeight = 1;
+        PropertyMap::iterator aElement = find(PropertyDefinition( PROP_HEIGHT, false ));
+        if( aElement != end())
+            aElement->second >>= nHeight;
+
+        sal_Int32 nWidth = 1;
+        aElement = find(PropertyDefinition( PROP_WIDTH, false ));
+        if( aElement != end())
+            aElement->second >>= nWidth;
+
+        text::WritingMode eWritingMode = text::WritingMode_LR_TB;
+        aElement = find(PropertyDefinition( PROP_WRITING_MODE, false ));
+        if( aElement != end())
+            aElement->second >>= eWritingMode;
+
+
+
+        sal_Int32 nTextAreaHeight = eWritingMode == text::WritingMode_LR_TB ?
+            nHeight - m_nTopMargin - m_nBottomMargin :
+            nWidth - m_nLeftMargin - m_nRightMargin;
+
+        operator[]( PropertyDefinition( PROP_GRID_LINES, false )) =
+                uno::makeAny( static_cast<sal_Int16>(nTextAreaHeight/m_nGridLinePitch));
+
+        sal_Int32 nCharWidth = 423; //240 twip/ 12 pt
+        //todo: is '0' the right index here?
+        const StyleSheetEntryPtr pEntry = rDM_Impl.GetStyleSheetTable()->FindStyleSheetByISTD(::rtl::OUString::valueOf(static_cast<sal_Int32>(0), 16));
+        if( pEntry.get( ) )
+        {
+            PropertyMap::iterator aElement_ = pEntry->pProperties->find(PropertyDefinition( PROP_CHAR_HEIGHT_ASIAN, false ));
+            if( aElement_ != pEntry->pProperties->end())
+            {
+                double fHeight = 0;
+                if( aElement_->second >>= fHeight )
+                    nCharWidth = ConversionHelper::convertTwipToMM100( (long)( fHeight * 20.0 + 0.5 ));
+            }
         }
-    }
 
-    //dxtCharSpace
-    if(m_nDxtCharSpace)
-    {
-        sal_Int32 nCharSpace = m_nDxtCharSpace;
-        //main lives in top 20 bits, and is signed.
-        sal_Int32 nMain = (nCharSpace & 0xFFFFF000);
-        nMain /= 0x1000;
-        nCharWidth += ConversionHelper::convertTwipToMM100( nMain * 20 );
-
-        sal_Int32 nFraction = (nCharSpace & 0x00000FFF);
-        nFraction = (nFraction * 20)/0xFFF;
-        nCharWidth += ConversionHelper::convertTwipToMM100( nFraction );
-    }
-    operator[]( PropertyDefinition( PROP_GRID_BASE_HEIGHT, false )) = uno::makeAny( nCharWidth );
-    sal_Int32 nRubyHeight = m_nGridLinePitch - nCharWidth;
-    if(nRubyHeight < 0 )
-        nRubyHeight = 0;
-    operator[]( PropertyDefinition( PROP_GRID_RUBY_HEIGHT, false )) = uno::makeAny( nRubyHeight );
+        //dxtCharSpace
+        if(m_nDxtCharSpace)
+        {
+            sal_Int32 nCharSpace = m_nDxtCharSpace;
+            //main lives in top 20 bits, and is signed.
+            sal_Int32 nMain = (nCharSpace & 0xFFFFF000);
+            nMain /= 0x1000;
+            nCharWidth += ConversionHelper::convertTwipToMM100( nMain * 20 );
+
+            sal_Int32 nFraction = (nCharSpace & 0x00000FFF);
+            nFraction = (nFraction * 20)/0xFFF;
+            nCharWidth += ConversionHelper::convertTwipToMM100( nFraction );
+        }
+        operator[]( PropertyDefinition( PROP_GRID_BASE_HEIGHT, false )) = uno::makeAny( nCharWidth );
+        sal_Int32 nRubyHeight = m_nGridLinePitch - nCharWidth;
+        if(nRubyHeight < 0 )
+            nRubyHeight = 0;
+        operator[]( PropertyDefinition( PROP_GRID_RUBY_HEIGHT, false )) = uno::makeAny( nRubyHeight );
 
-    sal_Int16 nGridMode = text::TextGridMode::NONE;
+        sal_Int16 nGridMode = text::TextGridMode::NONE;
 
-    switch (m_nGridType)
-    {
-        case NS_ooxml::LN_Value_wordprocessingml_ST_DocGrid_lines:
-            nGridMode = text::TextGridMode::LINES;
-            break;
-        case NS_ooxml::LN_Value_wordprocessingml_ST_DocGrid_linesAndChars:
-            nGridMode = text::TextGridMode::LINES_AND_CHARS;
-            break;
-        default:
-            break;
-    }
+        switch (m_nGridType)
+        {
+            case NS_ooxml::LN_Value_wordprocessingml_ST_DocGrid_lines:
+                nGridMode = text::TextGridMode::LINES;
+                break;
+            case NS_ooxml::LN_Value_wordprocessingml_ST_DocGrid_linesAndChars:
+                nGridMode = text::TextGridMode::LINES_AND_CHARS;
+                break;
+            default:
+                break;
+        }
 
-    operator[](PropertyDefinition(PROP_GRID_MODE, false)) = uno::makeAny(nGridMode);
+        operator[](PropertyDefinition(PROP_GRID_MODE, false)) = uno::makeAny(nGridMode);
 
-    _ApplyProperties( xFollowPageStyle );
+        _ApplyProperties( xFollowPageStyle );
 
-    //todo: creating a "First Page" style depends on HasTitlePage und _fFacingPage_
-    if( m_bTitlePage )
-    {
-        CopyLastHeaderFooter( true, rDM_Impl );
-        PrepareHeaderFooterProperties( true );
-        uno::Reference< beans::XPropertySet > xFirstPageStyle = GetPageStyle(
-        rDM_Impl.GetPageStyles(), rDM_Impl.GetTextFactory(), true );
-        _ApplyProperties( xFirstPageStyle );
-
-        sal_Int32 nPaperBin = m_nFirstPaperBin >= 0 ? m_nFirstPaperBin : m_nPaperBin >= 0 ? m_nPaperBin : 0;
-        if( nPaperBin )
-            xFollowPageStyle->setPropertyValue( sTrayIndex, uno::makeAny( nPaperBin ) );
-        if( xColumns.is() )
-            xFollowPageStyle->setPropertyValue(
-        rPropNameSupplier.GetName( PROP_TEXT_COLUMNS ), uno::makeAny( xColumns ));
-    }
+        //todo: creating a "First Page" style depends on HasTitlePage und _fFacingPage_
+        if( m_bTitlePage )
+        {
+            CopyLastHeaderFooter( true, rDM_Impl );
+            PrepareHeaderFooterProperties( true );
+            uno::Reference< beans::XPropertySet > xFirstPageStyle = GetPageStyle(
+                                rDM_Impl.GetPageStyles(), rDM_Impl.GetTextFactory(), true );
+            _ApplyProperties( xFirstPageStyle );
+
+            sal_Int32 nPaperBin = m_nFirstPaperBin >= 0 ? m_nFirstPaperBin : m_nPaperBin >= 0 ? m_nPaperBin : 0;
+            if( nPaperBin )
+                xFollowPageStyle->setPropertyValue( sTrayIndex, uno::makeAny( nPaperBin ) );
+            if( xColumns.is() )
+                xFollowPageStyle->setPropertyValue(
+                    rPropNameSupplier.GetName( PROP_TEXT_COLUMNS ), uno::makeAny( xColumns ));
+        }
 
-    ApplyBorderToPageStyles( rDM_Impl.GetPageStyles( ), rDM_Impl.GetTextFactory( ), m_nBorderParams );
+        ApplyBorderToPageStyles( rDM_Impl.GetPageStyles( ), rDM_Impl.GetTextFactory( ), m_nBorderParams );
 
-    if ( m_nBreakType != 0 )
-    {
         try
         {
             {
commit e7c861e3ec9398436eda16bb748af22a2d9afc27
Author: Caolán McNamara <caolanm at redhat.com>
Date:   Thu May 10 15:43:47 2012 +0100

    Related: fdo#47644 for huge files build a page chain cache for seeks
    
    For huge ole2 structured storage files build a page chain cache on
    demand to speed up long distance seeks
    
    i.e. reduces .doc parse time for fdo#47644 from 1 minute 7 seconds to 18
    seconds for me

diff --git a/sot/source/sdstor/stgstrms.cxx b/sot/source/sdstor/stgstrms.cxx
index 5e6c1ea..e982837 100644
--- a/sot/source/sdstor/stgstrms.cxx
+++ b/sot/source/sdstor/stgstrms.cxx
@@ -26,9 +26,10 @@
  *
  ************************************************************************/
 
+#include <algorithm>
 
 #include <string.h>     // memcpy()
-
+#include <sal/log.hxx>
 #include <osl/file.hxx>
 #include <tools/tempfile.hxx>
 #include <tools/debug.hxx>
@@ -334,6 +335,56 @@ void StgStrm::SetEntry( StgDirEntry& r )
     r.SetDirty();
 }
 
+namespace lcl
+{
+#if defined(__GXX_EXPERIMENTAL_CXX0X__) || __cplusplus >= 201103L
+    using std::is_sorted;
+#else
+    template <typename iter> bool is_sorted(iter aStart, iter aEnd)
+    {
+        if (aStart == aEnd)
+            return true;
+
+        for (iter aNext = aStart + 1; aNext != aEnd; aStart = aNext, ++aNext)
+        {
+            if (*aNext < *aStart)
+                return false;
+        }
+
+        return true;
+    }
+#endif
+}
+
+bool StgStrm::buildPageChainCache()
+{
+    if (nSize > 0)
+        m_aPagesCache.reserve(nSize/nPageSize);
+
+    sal_Int32 nBgn = nStart;
+    while (nBgn >= 0)
+    {
+        m_aPagesCache.push_back(nBgn);
+        sal_Int32 nOldBgn = nBgn;
+        nBgn = pFat->GetNextPage(nBgn);
+        if (nBgn == nOldBgn)
+            return false;
+    }
+
+    m_bSortedPageChain = lcl::is_sorted(m_aPagesCache.begin(), m_aPagesCache.end());
+
+    SAL_WARN_IF(!m_bSortedPageChain, "sot", "unsorted page chain, that's suspicious");
+
+    return true;
+}
+
+//See fdo#47644 for a .doc with a vast amount of pages where seeking around the
+//document takes a colossal amount of time
+//
+//There's a cost to building a page cache, so only build one if the number of
+//pages to seek through hits some sufficiently high value where it's worth it.
+#define ARBITRARY_LARGE_AMOUNT_OF_PAGES 256
+
 // Compute page number and offset for the given byte position.
 // If the position is behind the size, set the stream right
 // behind the EOF.
@@ -355,7 +406,7 @@ sal_Bool StgStrm::Pos2Page( sal_Int32 nBytePos )
         return sal_True;
     if( nNew > nOld )
     {
-        // the new position is behind the current, so an incremental
+        // the new position is after the current, so an incremental
         // positioning is OK. Set the page relative position
         nRel = nNew - nOld;
         nBgn = nPage;
@@ -369,13 +420,59 @@ sal_Bool StgStrm::Pos2Page( sal_Int32 nBytePos )
     }
     // now, traverse the FAT chain.
     nRel /= nPageSize;
+
     sal_Int32 nLast = STG_EOF;
-    while( nRel && nBgn >= 0 )
+    if (m_aPagesCache.empty() && nRel < ARBITRARY_LARGE_AMOUNT_OF_PAGES)
+    {
+        while (nRel && nBgn >= 0)
+        {
+            nLast = nBgn;
+            nBgn = pFat->GetNextPage( nBgn );
+            nRel--;
+        }
+    }
+    else if (nBgn >= 0)
     {
-        nLast = nBgn;
-        nBgn = pFat->GetNextPage( nBgn );
-        nRel--;
+        //Seeking large distances is slow, so if we're starting seeking (some
+        //fairly arbitrary) large distances, build a cache and re-use it for
+        //subsequent seeks
+        if (m_aPagesCache.empty())
+        {
+            SAL_INFO("sot", "kicking off large seek helper\n");
+            buildPageChainCache();
+        }
+
+        std::vector<sal_Int32>::iterator aI;
+
+        if (m_bSortedPageChain)
+            aI = std::lower_bound(m_aPagesCache.begin(), m_aPagesCache.end(), nBgn);
+        else
+            aI = std::find(m_aPagesCache.begin(), m_aPagesCache.end(), nBgn);
+
+        if (aI == m_aPagesCache.end())
+        {
+            SAL_WARN("sot", "Unknown page position");
+            nBgn = STG_EOF;
+        }
+        else
+        {
+            size_t nBgnDistance = std::distance(m_aPagesCache.begin(), aI);
+
+            size_t nIndex = nBgnDistance + nRel;
+
+            if (nIndex > m_aPagesCache.size())
+            {
+                nRel = m_aPagesCache.size() - nBgnDistance;
+                nIndex = m_aPagesCache.size() - 1;
+            }
+            else
+                nRel = 0;
+
+            nLast = nIndex ? m_aPagesCache[nIndex - 1] : STG_EOF;
+            nBgn = m_aPagesCache[nIndex];
+        }
     }
+
     // special case: seek to 1st byte of new, unallocated page
     // (in case the file size is a multiple of the page size)
     if( nBytePos == nSize && nBgn == STG_EOF && !nRel && !nOffset )
@@ -404,6 +501,8 @@ StgPage* StgStrm::GetPhysPage( sal_Int32 nBytePos, sal_Bool bForce )
 
 sal_Bool StgStrm::Copy( sal_Int32 nFrom, sal_Int32 nBytes )
 {
+    m_aPagesCache.clear();
+
     sal_Int32 nTo = nStart;
     sal_Int32 nPgs = ( nBytes + nPageSize - 1 ) / nPageSize;
     while( nPgs-- )
@@ -430,6 +529,8 @@ sal_Bool StgStrm::Copy( sal_Int32 nFrom, sal_Int32 nBytes )
 
 sal_Bool StgStrm::SetSize( sal_Int32 nBytes )
 {
+    m_aPagesCache.clear();
+
     // round up to page size
     sal_Int32 nOld = ( ( nSize + nPageSize - 1 ) / nPageSize ) * nPageSize;
     sal_Int32 nNew = ( ( nBytes + nPageSize - 1 ) / nPageSize ) * nPageSize;
@@ -528,6 +629,8 @@ sal_Int32 StgFATStrm::GetPage( short nOff, sal_Bool bMake, sal_uInt16 *pnMasterA
         {
             if( bMake )
             {
+                m_aPagesCache.clear();
+
                 // create a new master page
                 nFAT = nMaxPage++;
                 pMaster = rIo.Copy( nFAT, STG_FREE );
@@ -582,6 +685,8 @@ sal_Int32 StgFATStrm::GetPage( short nOff, sal_Bool bMake, sal_uInt16 *pnMasterA
 
 sal_Bool StgFATStrm::SetPage( short nOff, sal_Int32 nNewPage )
 {
+    m_aPagesCache.clear();
+
     sal_Bool bRes = sal_True;
     if( nOff < rIo.aHdr.GetFAT1Size() )
         rIo.aHdr.SetFATPage( nOff, nNewPage );
@@ -631,6 +736,8 @@ sal_Bool StgFATStrm::SetPage( short nOff, sal_Int32 nNewPage )
 
 sal_Bool StgFATStrm::SetSize( sal_Int32 nBytes )
 {
+    m_aPagesCache.clear();
+
     // Set the number of entries to a multiple of the page size
     short nOld = (short) ( ( nSize + ( nPageSize - 1 ) ) / nPageSize );
     short nNew = (short) (
@@ -747,16 +854,10 @@ void StgDataStrm::Init( sal_Int32 nBgn, sal_Int32 nLen )
     {
         // determine the actual size of the stream by scanning
         // the FAT chain and counting the # of pages allocated
-        nSize = 0;
-        sal_Int32 nOldBgn = -1;
-        while( nBgn >= 0 && nBgn != nOldBgn )
-        {
-            nOldBgn = nBgn;
-            nBgn = pFat->GetNextPage( nBgn );
-            if( nBgn == nOldBgn )
-                rIo.SetError( ERRCODE_IO_WRONGFORMAT );
-            nSize += nPageSize;
-        }
+        bool bOk = buildPageChainCache();
+        if (!bOk)
+            rIo.SetError( ERRCODE_IO_WRONGFORMAT );
+        nSize = m_aPagesCache.size() * nPageSize;
     }
 }
 
diff --git a/sot/source/sdstor/stgstrms.hxx b/sot/source/sdstor/stgstrms.hxx
index a6a0ad1..cfae9e2 100644
--- a/sot/source/sdstor/stgstrms.hxx
+++ b/sot/source/sdstor/stgstrms.hxx
@@ -30,6 +30,7 @@
 #define _STGSTRMS_HXX
 
 #include <tools/stream.hxx>
+#include <vector>
 
 class StgIo;
 class StgStrm;
@@ -77,6 +78,9 @@ protected:
     sal_Int32 nPage;                        // current logical page
     short nOffset;                      // offset into current page
     short nPageSize;                    // logical page size
+    std::vector<sal_Int32> m_aPagesCache;
+    bool m_bSortedPageChain;
+    bool buildPageChainCache();
     sal_Bool  Copy( sal_Int32 nFrom, sal_Int32 nBytes );
     StgStrm( StgIo& );
 public:


More information about the Libreoffice-commits mailing list