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

Michael Stahl mst at kemper.freedesktop.org
Mon Jun 4 04:26:07 PDT 2012


 writerfilter/source/rtftok/rtfdocumentimpl.cxx |   20 ++++++++++----------
 writerfilter/source/rtftok/rtfsprm.cxx         |   14 ++++++++++++++
 writerfilter/source/rtftok/rtfsprm.hxx         |    2 ++
 3 files changed, 26 insertions(+), 10 deletions(-)

New commits:
commit d366ae2031d13f3198cb58c715f18e8c8ce8686f
Author: Miklos Vajna <vmiklos at suse.cz>
Date:   Wed May 30 11:43:41 2012 +0200

    fdo#49968 speed up RTF import of repeated character/paragraph properties
    
    Most RTF documents (produced by Word/Writer) reset character and
    paragraph properties at the start of each paragraph. Because of this,
    appending properties of the same type didn't cause any noticable
    performance problems. However, it's valid to not reset these properties,
    and in this case a longer document takes forever to import.
    
    Filter these duplicates at the tokenizer level for trivial properties to
    get acceptable import speed.
    
    Also fixes rhbz#825548 in an easier-to-backport way.
    
    Change-Id: Id0b7289323d45ff0d747c74bb78d8eb7def0cfc2
    (cherry picked from commit a5328cf5605cdc243522eddcaffe7336196a4900)
    
    Signed-off-by: Michael Stahl <mstahl at redhat.com>

diff --git a/writerfilter/source/rtftok/rtfdocumentimpl.cxx b/writerfilter/source/rtftok/rtfdocumentimpl.cxx
index 063d94b..ea14c57 100644
--- a/writerfilter/source/rtftok/rtfdocumentimpl.cxx
+++ b/writerfilter/source/rtftok/rtfdocumentimpl.cxx
@@ -1664,7 +1664,7 @@ int RTFDocumentImpl::dispatchFlag(RTFKeyword nKeyword)
     if (nParam >= 0)
     {
         RTFValue::Pointer_t pValue(new RTFValue(nParam));
-        m_aStates.top().aParagraphSprms->push_back(make_pair(NS_sprm::LN_PJc, pValue));
+        m_aStates.top().aParagraphSprms.set(NS_sprm::LN_PJc, pValue);
         return 0;
     }
 
@@ -2222,7 +2222,7 @@ int RTFDocumentImpl::dispatchValue(RTFKeyword nKeyword, int nParam)
     }
     if (nSprm > 0)
     {
-        m_aStates.top().aCharacterSprms->push_back(make_pair(nSprm, pIntValue));
+        m_aStates.top().aCharacterSprms.set(nSprm, pIntValue);
         // Language is a character property, but we should store it at a paragraph level as well for fields.
         if (nKeyword == RTF_LANG && m_bNeedPap)
             m_aStates.top().aParagraphSprms->push_back(make_pair(nSprm, pIntValue));
@@ -2240,7 +2240,7 @@ int RTFDocumentImpl::dispatchValue(RTFKeyword nKeyword, int nParam)
     }
     if (nSprm > 0)
     {
-        m_aStates.top().aParagraphSprms->push_back(make_pair(nSprm, pIntValue));
+        m_aStates.top().aParagraphSprms.set(nSprm, pIntValue);
         if (nKeyword == RTF_ITAP && nParam > 0)
             // Invalid tables may omit INTBL after ITAP
             dispatchFlag(RTF_INTBL);
@@ -2305,7 +2305,7 @@ int RTFDocumentImpl::dispatchValue(RTFKeyword nKeyword, int nParam)
             {
                 int nFontIndex = getFontIndex(nParam);
                 RTFValue::Pointer_t pValue(new RTFValue(nFontIndex));
-                m_aStates.top().aCharacterSprms->push_back(make_pair(NS_sprm::LN_CRgFtc0, pValue));
+                m_aStates.top().aCharacterSprms.set(NS_sprm::LN_CRgFtc0, pValue);
                 m_aStates.top().nCurrentEncoding = getEncoding(nFontIndex);
             }
             break;
@@ -2353,7 +2353,7 @@ int RTFDocumentImpl::dispatchValue(RTFKeyword nKeyword, int nParam)
                 m_aStates.top().aTableAttributes->push_back(make_pair(NS_rtf::LN_SGC, pValue)); // paragraph style
             }
             else
-                m_aStates.top().aParagraphAttributes->push_back(make_pair(NS_rtf::LN_ISTD, pIntValue));
+                m_aStates.top().aParagraphAttributes.set(NS_rtf::LN_ISTD, pIntValue);
             break;
         case RTF_CS:
             if (m_aStates.top().nDestinationState == DESTINATION_STYLESHEET)
@@ -2814,11 +2814,11 @@ int RTFDocumentImpl::dispatchValue(RTFKeyword nKeyword, int nParam)
             break;
         case RTF_SB:
             lcl_putNestedAttribute(m_aStates.top().aParagraphSprms,
-                    NS_ooxml::LN_CT_PPrBase_spacing, NS_ooxml::LN_CT_Spacing_before, pIntValue);
+                    NS_ooxml::LN_CT_PPrBase_spacing, NS_ooxml::LN_CT_Spacing_before, pIntValue, true);
             break;
         case RTF_SA:
             lcl_putNestedAttribute(m_aStates.top().aParagraphSprms,
-                    NS_ooxml::LN_CT_PPrBase_spacing, NS_ooxml::LN_CT_Spacing_after, pIntValue);
+                    NS_ooxml::LN_CT_PPrBase_spacing, NS_ooxml::LN_CT_Spacing_after, pIntValue, true);
             break;
         case RTF_DPX:
             m_aStates.top().aDrawingObject.nLeft = TWIP_TO_MM100(nParam);
@@ -2847,11 +2847,11 @@ int RTFDocumentImpl::dispatchValue(RTFKeyword nKeyword, int nParam)
             m_aStates.top().nBinaryToRead = nParam;
             break;
         case RTF_LI:
-            m_aStates.top().aParagraphSprms->push_back(make_pair(NS_sprm::LN_PDxaLeft, pIntValue));
+            m_aStates.top().aParagraphSprms.set(NS_sprm::LN_PDxaLeft, pIntValue);
             // It turns out \li should reset the \fi inherited from the stylesheet.
             // So set the direct formatting to zero, if we don't have such direct formatting yet.
             if (!m_aStates.top().aParagraphSprms.find(NS_sprm::LN_PDxaLeft1).get())
-                m_aStates.top().aParagraphSprms->push_back(make_pair(NS_sprm::LN_PDxaLeft1, RTFValue::Pointer_t(new RTFValue(0))));
+                m_aStates.top().aParagraphSprms.set(NS_sprm::LN_PDxaLeft1, RTFValue::Pointer_t(new RTFValue(0)));
             break;
         default:
 #if OSL_DEBUG_LEVEL > 1
@@ -2934,7 +2934,7 @@ int RTFDocumentImpl::dispatchToggle(RTFKeyword nKeyword, bool bParam, int nParam
     }
     if (nSprm >= 0)
     {
-        m_aStates.top().aCharacterSprms->push_back(make_pair(nSprm, pBoolValue));
+        m_aStates.top().aCharacterSprms.set(nSprm, pBoolValue);
         return 0;
     }
 
diff --git a/writerfilter/source/rtftok/rtfsprm.cxx b/writerfilter/source/rtftok/rtfsprm.cxx
index 976505f..72793fe 100644
--- a/writerfilter/source/rtftok/rtfsprm.cxx
+++ b/writerfilter/source/rtftok/rtfsprm.cxx
@@ -105,6 +105,20 @@ RTFValue::Pointer_t RTFSprms::find(Id nKeyword)
     return pValue;
 }
 
+void RTFSprms::set(Id nKeyword, RTFValue::Pointer_t pValue, bool bOverwrite)
+{
+    if (bOverwrite)
+    {
+        for (RTFSprms::Iterator_t i = m_aSprms.begin(); i != m_aSprms.end(); ++i)
+            if (i->first == nKeyword)
+            {
+                i->second = pValue;
+                return;
+            }
+    }
+    m_aSprms.push_back(std::make_pair(nKeyword, pValue));
+}
+
 bool RTFSprms::erase(Id nKeyword)
 {
     for (RTFSprms::Iterator_t i = m_aSprms.begin(); i != m_aSprms.end(); ++i)
diff --git a/writerfilter/source/rtftok/rtfsprm.hxx b/writerfilter/source/rtftok/rtfsprm.hxx
index dbf2e2f..df33e5f 100644
--- a/writerfilter/source/rtftok/rtfsprm.hxx
+++ b/writerfilter/source/rtftok/rtfsprm.hxx
@@ -43,6 +43,8 @@ namespace writerfilter {
                 RTFSprms(const RTFSprms& rSprms);
                 std::vector< std::pair<Id, RTFValue::Pointer_t> >* operator->();
                 RTFValue::Pointer_t find(Id nKeyword);
+                /// Does the same as ->push_back(), except that it can overwrite existing entries.
+                void set(Id nKeyword, RTFValue::Pointer_t pValue, bool bOverwrite = true);
                 bool erase(Id nKeyword);
             private:
                 std::vector< std::pair<Id, RTFValue::Pointer_t> > m_aSprms;


More information about the Libreoffice-commits mailing list