b9d1006ebb
#i118662# remove berkeleyDB from module l10tools Patch contributed by Oliver-Rainer Wittmann http://svn.apache.org/viewvc?view=revision&revision=1213189 reportdesign: remove extra items from Help menu Patch contributed by Ariel Constenla-Haile http://svn.apache.org/viewvc?view=revision&revision=1201215
698 lines
25 KiB
C++
698 lines
25 KiB
C++
/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*- */
|
|
/*
|
|
* This file is part of the LibreOffice project.
|
|
*
|
|
* This Source Code Form is subject to the terms of the Mozilla Public
|
|
* License, v. 2.0. If a copy of the MPL was not distributed with this
|
|
* file, You can obtain one at http://mozilla.org/MPL/2.0/.
|
|
*
|
|
* This file incorporates work covered by the following license notice:
|
|
*
|
|
* Licensed to the Apache Software Foundation (ASF) under one or more
|
|
* contributor license agreements. See the NOTICE file distributed
|
|
* with this work for additional information regarding copyright
|
|
* ownership. The ASF licenses this file to you under the Apache
|
|
* License, Version 2.0 (the "License"); you may not use this file
|
|
* except in compliance with the License. You may obtain a copy of
|
|
* the License at http://www.apache.org/licenses/LICENSE-2.0 .
|
|
*/
|
|
|
|
|
|
#if defined __SUNPRO_CC
|
|
#pragma disable_warn
|
|
#elif defined _MSC_VER
|
|
#pragma warning(push, 1)
|
|
#endif
|
|
|
|
#include "pdfparse.hxx"
|
|
|
|
// workaround windows compiler: do not include multi_pass.hpp
|
|
#include <boost/spirit/include/classic_core.hpp>
|
|
#include <boost/spirit/include/classic_utility.hpp>
|
|
#include <boost/spirit/include/classic_error_handling.hpp>
|
|
#include <boost/spirit/include/classic_file_iterator.hpp>
|
|
#include <boost/bind.hpp>
|
|
#include <string.h>
|
|
|
|
#include <rtl/strbuf.hxx>
|
|
#include <rtl/alloc.h>
|
|
|
|
// disable warnings again because someone along the line has enabled them
|
|
#if defined __SUNPRO_CC
|
|
#pragma disable_warn
|
|
#elif defined _MSC_VER
|
|
#pragma warning(push, 1)
|
|
#endif
|
|
|
|
using namespace boost::spirit;
|
|
using namespace pdfparse;
|
|
|
|
using ::rtl::OString;
|
|
using ::rtl::OStringBuffer;
|
|
|
|
class StringEmitContext : public EmitContext
|
|
{
|
|
OStringBuffer m_aBuf;
|
|
public:
|
|
StringEmitContext() : EmitContext(), m_aBuf(256) {}
|
|
virtual ~StringEmitContext() {}
|
|
virtual bool write( const void* pBuf, unsigned int nLen ) throw()
|
|
{
|
|
m_aBuf.append( (const sal_Char*)pBuf, nLen );
|
|
return true;
|
|
}
|
|
virtual unsigned int getCurPos() throw() { return m_aBuf.getLength(); }
|
|
virtual bool copyOrigBytes( unsigned int nOrigOffset, unsigned int nLen ) throw()
|
|
{ return (nOrigOffset+nLen < static_cast<unsigned int>(m_aBuf.getLength()) ) ?
|
|
write( m_aBuf.getStr() + nOrigOffset, nLen ) : false; }
|
|
virtual unsigned int readOrigBytes( unsigned int nOrigOffset, unsigned int nLen, void* pBuf ) throw()
|
|
{
|
|
if( nOrigOffset+nLen < static_cast<unsigned int>(m_aBuf.getLength()) )
|
|
{
|
|
memcpy( pBuf, m_aBuf.getStr()+nOrigOffset, nLen );
|
|
return nLen;
|
|
}
|
|
return 0;
|
|
}
|
|
|
|
OString getString() { return m_aBuf.makeStringAndClear(); }
|
|
};
|
|
|
|
template< class iteratorT >
|
|
class PDFGrammar : public grammar< PDFGrammar<iteratorT> >
|
|
{
|
|
public:
|
|
|
|
PDFGrammar( const iteratorT& first )
|
|
: m_fDouble( 0.0 ), m_aGlobalBegin( first ) {}
|
|
~PDFGrammar()
|
|
{
|
|
if( !m_aObjectStack.empty() )
|
|
delete m_aObjectStack.front();
|
|
}
|
|
|
|
double m_fDouble;
|
|
std::vector< unsigned int > m_aUIntStack;
|
|
std::vector< PDFEntry* > m_aObjectStack;
|
|
rtl::OString m_aErrorString;
|
|
iteratorT m_aGlobalBegin;
|
|
|
|
public:
|
|
struct pdf_string_parser
|
|
{
|
|
typedef nil_t result_t;
|
|
template <typename ScannerT>
|
|
std::ptrdiff_t
|
|
operator()(ScannerT const& scan, result_t&) const
|
|
{
|
|
std::ptrdiff_t len = 0;
|
|
|
|
int nBraceLevel = 0;
|
|
while( ! scan.at_end() )
|
|
{
|
|
char c = *scan;
|
|
if( c == ')' )
|
|
{
|
|
nBraceLevel--;
|
|
if( nBraceLevel < 0 )
|
|
break;
|
|
}
|
|
else if( c == '(' )
|
|
nBraceLevel++;
|
|
else if( c == '\\' ) // ignore escaped braces
|
|
{
|
|
++len;
|
|
++scan;
|
|
if( scan.at_end() )
|
|
break;
|
|
}
|
|
++len;
|
|
++scan;
|
|
}
|
|
return scan.at_end() ? -1 : len;
|
|
}
|
|
};
|
|
|
|
template< typename ScannerT >
|
|
struct definition
|
|
{
|
|
definition( const PDFGrammar<iteratorT>& rSelf )
|
|
{
|
|
PDFGrammar<iteratorT>* pSelf = const_cast< PDFGrammar<iteratorT>* >( &rSelf );
|
|
|
|
// workaround workshop compiler: comment_p doesn't work
|
|
// comment = comment_p("%")[boost::bind(&PDFGrammar::pushComment, pSelf, _1, _2 )];
|
|
comment = lexeme_d[ (ch_p('%') >> *(~ch_p('\r') & ~ch_p('\n')) >> eol_p)[boost::bind(&PDFGrammar::pushComment, pSelf, _1, _2 )] ];
|
|
|
|
boolean = (str_p("true") | str_p("false"))[boost::bind(&PDFGrammar::pushBool, pSelf, _1, _2)];
|
|
|
|
// workaround workshop compiler: confix_p doesn't work
|
|
//stream = confix_p( "stream", *anychar_p, "endstream" )[boost::bind(&PDFGrammar::emitStream, pSelf, _1, _2 )];
|
|
stream = (str_p("stream") >> *(anychar_p - str_p("endstream")) >> str_p("endstream"))[boost::bind(&PDFGrammar::emitStream, pSelf, _1, _2 )];
|
|
|
|
name = lexeme_d[
|
|
ch_p('/')
|
|
>> (*(anychar_p-chset_p("\t\n\f\r ()<>[]{}/%")-ch_p('\0')))
|
|
[boost::bind(&PDFGrammar::pushName, pSelf, _1, _2)] ];
|
|
|
|
// workaround workshop compiler: confix_p doesn't work
|
|
//stringtype = ( confix_p("(",*anychar_p, ")") |
|
|
// confix_p("<",*xdigit_p, ">") )
|
|
// [boost::bind(&PDFGrammar::pushString,pSelf, _1, _2)];
|
|
|
|
stringtype = ( ( ch_p('(') >> functor_parser<pdf_string_parser>() >> ch_p(')') ) |
|
|
( ch_p('<') >> *xdigit_p >> ch_p('>') ) )
|
|
[boost::bind(&PDFGrammar::pushString,pSelf, _1, _2)];
|
|
|
|
null_object = str_p( "null" )[boost::bind(&PDFGrammar::pushNull, pSelf, _1, _2)];
|
|
|
|
#ifdef USE_ASSIGN_ACTOR
|
|
objectref = ( uint_p[push_back_a(pSelf->m_aUIntStack)]
|
|
>> uint_p[push_back_a(pSelf->m_aUIntStack)]
|
|
>> ch_p('R')
|
|
>> eps_p
|
|
)[boost::bind(&PDFGrammar::pushObjectRef, pSelf, _1, _2)];
|
|
#else
|
|
objectref = ( uint_p[boost::bind(&PDFGrammar::push_back_action_uint, pSelf, _1)]
|
|
>> uint_p[boost::bind(&PDFGrammar::push_back_action_uint, pSelf, _1)]
|
|
>> ch_p('R')
|
|
>> eps_p
|
|
)[boost::bind(&PDFGrammar::pushObjectRef, pSelf, _1, _2)];
|
|
#endif
|
|
|
|
#ifdef USE_ASSIGN_ACTOR
|
|
simple_type = objectref | name |
|
|
( real_p[assign_a(pSelf->m_fDouble)] >> eps_p )
|
|
[boost::bind(&PDFGrammar::pushDouble, pSelf, _1, _2)]
|
|
| stringtype | boolean | null_object;
|
|
#else
|
|
simple_type = objectref | name |
|
|
( real_p[boost::bind(&PDFGrammar::assign_action_double, pSelf, _1)] >> eps_p )
|
|
[boost::bind(&PDFGrammar::pushDouble, pSelf, _1, _2)]
|
|
| stringtype | boolean | null_object;
|
|
#endif
|
|
|
|
dict_begin = str_p( "<<" )[boost::bind(&PDFGrammar::beginDict, pSelf, _1, _2)];
|
|
dict_end = str_p( ">>" )[boost::bind(&PDFGrammar::endDict, pSelf, _1, _2)];
|
|
|
|
array_begin = str_p("[")[boost::bind(&PDFGrammar::beginArray,pSelf, _1, _2)];
|
|
array_end = str_p("]")[boost::bind(&PDFGrammar::endArray,pSelf, _1, _2)];
|
|
|
|
#ifdef USE_ASSIGN_ACTOR
|
|
object_begin= uint_p[push_back_a(pSelf->m_aUIntStack)]
|
|
>> uint_p[push_back_a(pSelf->m_aUIntStack)]
|
|
>> str_p("obj" )[boost::bind(&PDFGrammar::beginObject, pSelf, _1, _2)];
|
|
#else
|
|
object_begin= uint_p[boost::bind(&PDFGrammar::push_back_action_uint, pSelf, _1)]
|
|
>> uint_p[boost::bind(&PDFGrammar::push_back_action_uint, pSelf, _1)]
|
|
>> str_p("obj" )[boost::bind(&PDFGrammar::beginObject, pSelf, _1, _2)];
|
|
#endif
|
|
object_end = str_p( "endobj" )[boost::bind(&PDFGrammar::endObject, pSelf, _1, _2)];
|
|
|
|
xref = str_p( "xref" ) >> uint_p >> uint_p
|
|
>> lexeme_d[
|
|
+( repeat_p(10)[digit_p]
|
|
>> blank_p
|
|
>> repeat_p(5)[digit_p]
|
|
>> blank_p
|
|
>> ( ch_p('n') | ch_p('f') )
|
|
>> repeat_p(2)[space_p]
|
|
) ];
|
|
|
|
dict_element= dict_begin | comment | simple_type
|
|
| array_begin | array_end | dict_end;
|
|
|
|
object = object_begin
|
|
>> *dict_element
|
|
>> !stream
|
|
>> object_end;
|
|
|
|
trailer = str_p( "trailer" )[boost::bind(&PDFGrammar::beginTrailer,pSelf,_1,_2)]
|
|
>> *dict_element
|
|
>> str_p("startxref")
|
|
>> uint_p
|
|
>> str_p("%%EOF")[boost::bind(&PDFGrammar::endTrailer,pSelf,_1,_2)];
|
|
|
|
#ifdef USE_ASSIGN_ACTOR
|
|
pdfrule = ! (lexeme_d[
|
|
str_p( "%PDF-" )
|
|
>> uint_p[push_back_a(pSelf->m_aUIntStack)]
|
|
>> ch_p('.')
|
|
>> uint_p[push_back_a(pSelf->m_aUIntStack)]
|
|
>> *((~ch_p('\r') & ~ch_p('\n')))
|
|
>> eol_p
|
|
])[boost::bind(&PDFGrammar::haveFile,pSelf, _1, _2)]
|
|
>> *( comment | object | ( xref >> trailer ) );
|
|
#else
|
|
pdfrule = ! (lexeme_d[
|
|
str_p( "%PDF-" )
|
|
>> uint_p[boost::bind(&PDFGrammar::push_back_action_uint, pSelf, _1)]
|
|
>> ch_p('.')
|
|
>> uint_p[boost::bind(&PDFGrammar::push_back_action_uint, pSelf, _1)]
|
|
>> *((~ch_p('\r') & ~ch_p('\n')))
|
|
>> eol_p
|
|
])[boost::bind(&PDFGrammar::haveFile,pSelf, _1, _2)]
|
|
>> *( comment | object | ( xref >> trailer ) );
|
|
#endif
|
|
}
|
|
rule< ScannerT > comment, stream, boolean, name, stringtype, null_object, simple_type,
|
|
objectref, array, value, dict_element, dict_begin, dict_end,
|
|
array_begin, array_end, object, object_begin, object_end,
|
|
xref, trailer, pdfrule;
|
|
|
|
const rule< ScannerT >& start() const { return pdfrule; }
|
|
};
|
|
|
|
#ifndef USE_ASSIGN_ACTOR
|
|
void push_back_action_uint( unsigned int i )
|
|
{
|
|
m_aUIntStack.push_back( i );
|
|
}
|
|
void assign_action_double( double d )
|
|
{
|
|
m_fDouble = d;
|
|
}
|
|
#endif
|
|
|
|
void parseError( const char* pMessage, iteratorT pLocation )
|
|
{
|
|
throw_( pLocation, pMessage );
|
|
}
|
|
|
|
rtl::OString iteratorToString( iteratorT first, iteratorT last ) const
|
|
{
|
|
rtl::OStringBuffer aStr( 32 );
|
|
while( first != last )
|
|
{
|
|
aStr.append( *first );
|
|
++first;
|
|
}
|
|
return aStr.makeStringAndClear();
|
|
}
|
|
|
|
void haveFile( iteratorT pBegin, SAL_UNUSED_PARAMETER iteratorT /*pEnd*/ )
|
|
{
|
|
if( m_aObjectStack.empty() )
|
|
{
|
|
PDFFile* pFile = new PDFFile();
|
|
pFile->m_nMinor = m_aUIntStack.back();
|
|
m_aUIntStack.pop_back();
|
|
pFile->m_nMajor = m_aUIntStack.back();
|
|
m_aUIntStack.pop_back();
|
|
m_aObjectStack.push_back( pFile );
|
|
}
|
|
else
|
|
parseError( "found file header in unusual place", pBegin );
|
|
}
|
|
|
|
void pushComment( iteratorT first, iteratorT last )
|
|
{
|
|
// add a comment to the current stack element
|
|
PDFComment* pComment =
|
|
new PDFComment(iteratorToString(first,last));
|
|
if( m_aObjectStack.empty() )
|
|
m_aObjectStack.push_back( new PDFPart() );
|
|
PDFContainer* pContainer = dynamic_cast<PDFContainer*>(m_aObjectStack.back());
|
|
if( pContainer == NULL )
|
|
parseError( "comment without container", first );
|
|
pContainer->m_aSubElements.push_back( pComment );
|
|
}
|
|
|
|
void insertNewValue( PDFEntry* pNewValue, iteratorT pPos )
|
|
{
|
|
PDFContainer* pContainer = NULL;
|
|
const char* pMsg = NULL;
|
|
if( ! m_aObjectStack.empty() &&
|
|
(pContainer = dynamic_cast<PDFContainer*>(m_aObjectStack.back())) != NULL )
|
|
{
|
|
if( dynamic_cast<PDFDict*>(pContainer) == NULL &&
|
|
dynamic_cast<PDFArray*>(pContainer) == NULL )
|
|
{
|
|
PDFObject* pObj = dynamic_cast<PDFObject*>(pContainer);
|
|
if( pObj )
|
|
{
|
|
if( pObj->m_pObject == NULL )
|
|
pObj->m_pObject = pNewValue;
|
|
else
|
|
{
|
|
pMsg = "second value for object";
|
|
pContainer = NULL;
|
|
}
|
|
}
|
|
else if( dynamic_cast<PDFDict*>(pNewValue) )
|
|
{
|
|
PDFTrailer* pTrailer = dynamic_cast<PDFTrailer*>(pContainer);
|
|
if( pTrailer )
|
|
{
|
|
if( pTrailer->m_pDict == NULL )
|
|
pTrailer->m_pDict = dynamic_cast<PDFDict*>(pNewValue);
|
|
else
|
|
pContainer = NULL;
|
|
}
|
|
else
|
|
pContainer = NULL;
|
|
}
|
|
else
|
|
pContainer = NULL;
|
|
}
|
|
}
|
|
if( pContainer )
|
|
pContainer->m_aSubElements.push_back( pNewValue );
|
|
else
|
|
{
|
|
if( ! pMsg )
|
|
{
|
|
if( dynamic_cast<PDFContainer*>(pNewValue) )
|
|
pMsg = "array without container";
|
|
else
|
|
pMsg = "value without container";
|
|
}
|
|
delete pNewValue;
|
|
parseError( pMsg, pPos );
|
|
}
|
|
}
|
|
|
|
void pushName( iteratorT first, iteratorT last )
|
|
{
|
|
insertNewValue( new PDFName(iteratorToString(first,last)), first );
|
|
}
|
|
|
|
void pushDouble( iteratorT first, SAL_UNUSED_PARAMETER iteratorT /*last*/ )
|
|
{
|
|
insertNewValue( new PDFNumber(m_fDouble), first );
|
|
}
|
|
|
|
void pushString( iteratorT first, iteratorT last )
|
|
{
|
|
insertNewValue( new PDFString(iteratorToString(first,last)), first );
|
|
}
|
|
|
|
void pushBool( iteratorT first, iteratorT last )
|
|
{
|
|
insertNewValue( new PDFBool( (last-first == 4) ), first );
|
|
}
|
|
|
|
void pushNull( iteratorT first, SAL_UNUSED_PARAMETER iteratorT )
|
|
{
|
|
insertNewValue( new PDFNull(), first );
|
|
}
|
|
|
|
|
|
void beginObject( iteratorT first, SAL_UNUSED_PARAMETER iteratorT /*last*/ )
|
|
{
|
|
if( m_aObjectStack.empty() )
|
|
m_aObjectStack.push_back( new PDFPart() );
|
|
|
|
unsigned int nGeneration = m_aUIntStack.back();
|
|
m_aUIntStack.pop_back();
|
|
unsigned int nObject = m_aUIntStack.back();
|
|
m_aUIntStack.pop_back();
|
|
|
|
PDFObject* pObj = new PDFObject( nObject, nGeneration );
|
|
pObj->m_nOffset = first - m_aGlobalBegin;
|
|
|
|
PDFContainer* pContainer = dynamic_cast<PDFContainer*>(m_aObjectStack.back());
|
|
if( pContainer &&
|
|
( dynamic_cast<PDFFile*>(pContainer) ||
|
|
dynamic_cast<PDFPart*>(pContainer) ) )
|
|
{
|
|
pContainer->m_aSubElements.push_back( pObj );
|
|
m_aObjectStack.push_back( pObj );
|
|
}
|
|
else
|
|
parseError( "object in wrong place", first );
|
|
}
|
|
|
|
void endObject( iteratorT first, SAL_UNUSED_PARAMETER iteratorT )
|
|
{
|
|
if( m_aObjectStack.empty() )
|
|
parseError( "endobj without obj", first );
|
|
else if( dynamic_cast<PDFObject*>(m_aObjectStack.back()) == NULL )
|
|
parseError( "spurious endobj", first );
|
|
else
|
|
m_aObjectStack.pop_back();
|
|
}
|
|
|
|
void pushObjectRef( iteratorT first, SAL_UNUSED_PARAMETER iteratorT )
|
|
{
|
|
unsigned int nGeneration = m_aUIntStack.back();
|
|
m_aUIntStack.pop_back();
|
|
unsigned int nObject = m_aUIntStack.back();
|
|
m_aUIntStack.pop_back();
|
|
insertNewValue( new PDFObjectRef(nObject,nGeneration), first );
|
|
}
|
|
|
|
void beginDict( iteratorT first, SAL_UNUSED_PARAMETER iteratorT )
|
|
{
|
|
PDFDict* pDict = new PDFDict();
|
|
pDict->m_nOffset = first - m_aGlobalBegin;
|
|
|
|
insertNewValue( pDict, first );
|
|
// will not come here if insertion fails (exception)
|
|
m_aObjectStack.push_back( pDict );
|
|
}
|
|
void endDict( iteratorT first, SAL_UNUSED_PARAMETER iteratorT )
|
|
{
|
|
PDFDict* pDict = NULL;
|
|
if( m_aObjectStack.empty() )
|
|
parseError( "dictionary end without begin", first );
|
|
else if( (pDict = dynamic_cast<PDFDict*>(m_aObjectStack.back())) == NULL )
|
|
parseError( "spurious dictionary end", first );
|
|
else
|
|
m_aObjectStack.pop_back();
|
|
|
|
PDFEntry* pOffender = pDict->buildMap();
|
|
if( pOffender )
|
|
{
|
|
StringEmitContext aCtx;
|
|
aCtx.write( "offending dictionary element: ", 30 );
|
|
pOffender->emit( aCtx );
|
|
m_aErrorString = aCtx.getString();
|
|
parseError( m_aErrorString.getStr(), first );
|
|
}
|
|
}
|
|
|
|
void beginArray( iteratorT first, SAL_UNUSED_PARAMETER iteratorT )
|
|
{
|
|
PDFArray* pArray = new PDFArray();
|
|
pArray->m_nOffset = first - m_aGlobalBegin;
|
|
|
|
insertNewValue( pArray, first );
|
|
// will not come here if insertion fails (exception)
|
|
m_aObjectStack.push_back( pArray );
|
|
}
|
|
|
|
void endArray( iteratorT first, SAL_UNUSED_PARAMETER iteratorT )
|
|
{
|
|
if( m_aObjectStack.empty() )
|
|
parseError( "array end without begin", first );
|
|
else if( dynamic_cast<PDFArray*>(m_aObjectStack.back()) == NULL )
|
|
parseError( "spurious array end", first );
|
|
else
|
|
m_aObjectStack.pop_back();
|
|
}
|
|
|
|
void emitStream( iteratorT first, iteratorT last )
|
|
{
|
|
if( m_aObjectStack.empty() )
|
|
parseError( "stream without object", first );
|
|
PDFObject* pObj = dynamic_cast<PDFObject*>(m_aObjectStack.back());
|
|
if( pObj && pObj->m_pObject )
|
|
{
|
|
if( pObj->m_pStream )
|
|
parseError( "multiple streams in object", first );
|
|
|
|
PDFDict* pDict = dynamic_cast<PDFDict*>(pObj->m_pObject);
|
|
if( pDict )
|
|
{
|
|
PDFStream* pStream = new PDFStream( first - m_aGlobalBegin, last - m_aGlobalBegin, pDict );
|
|
|
|
pObj->m_pStream = pStream;
|
|
pObj->m_aSubElements.push_back( pStream );
|
|
}
|
|
}
|
|
else
|
|
parseError( "stream without object", first );
|
|
}
|
|
|
|
void beginTrailer( iteratorT first, SAL_UNUSED_PARAMETER iteratorT )
|
|
{
|
|
if( m_aObjectStack.empty() )
|
|
m_aObjectStack.push_back( new PDFPart() );
|
|
|
|
PDFTrailer* pTrailer = new PDFTrailer();
|
|
pTrailer->m_nOffset = first - m_aGlobalBegin;
|
|
|
|
PDFContainer* pContainer = dynamic_cast<PDFContainer*>(m_aObjectStack.back());
|
|
if( pContainer &&
|
|
( dynamic_cast<PDFFile*>(pContainer) ||
|
|
dynamic_cast<PDFPart*>(pContainer) ) )
|
|
{
|
|
pContainer->m_aSubElements.push_back( pTrailer );
|
|
m_aObjectStack.push_back( pTrailer );
|
|
}
|
|
else
|
|
parseError( "trailer in wrong place", first );
|
|
}
|
|
|
|
void endTrailer( iteratorT first, SAL_UNUSED_PARAMETER iteratorT )
|
|
{
|
|
if( m_aObjectStack.empty() )
|
|
parseError( "%%EOF without trailer", first );
|
|
else if( dynamic_cast<PDFTrailer*>(m_aObjectStack.back()) == NULL )
|
|
parseError( "spurious %%EOF", first );
|
|
else
|
|
m_aObjectStack.pop_back();
|
|
}
|
|
};
|
|
|
|
#ifdef WIN32
|
|
PDFEntry* PDFReader::read( const char* pBuffer, unsigned int nLen )
|
|
{
|
|
PDFGrammar<const char*> aGrammar( pBuffer );
|
|
|
|
try
|
|
{
|
|
#if OSL_DEBUG_LEVEL > 1
|
|
boost::spirit::parse_info<const char*> aInfo =
|
|
#endif
|
|
boost::spirit::parse( pBuffer,
|
|
pBuffer+nLen,
|
|
aGrammar,
|
|
boost::spirit::space_p );
|
|
#if OSL_DEBUG_LEVEL > 1
|
|
fprintf( stderr, "parseinfo: stop = %p (buff=%p, offset = %d), hit = %s, full = %s, length = %d\n",
|
|
aInfo.stop, pBuffer, aInfo.stop - pBuffer,
|
|
aInfo.hit ? "true" : "false",
|
|
aInfo.full ? "true" : "false",
|
|
(int)aInfo.length );
|
|
#endif
|
|
}
|
|
catch( const parser_error<const char*, const char*>& rError )
|
|
{
|
|
#if OSL_DEBUG_LEVEL > 1
|
|
fprintf( stderr, "parse error: %s at buffer pos %u\nobject stack:\n",
|
|
rError.descriptor, rError.where - pBuffer );
|
|
unsigned int nElem = aGrammar.m_aObjectStack.size();
|
|
for( unsigned int i = 0; i < nElem; i++ )
|
|
{
|
|
fprintf( stderr, " %s\n", typeid( *(aGrammar.m_aObjectStack[i]) ).name() );
|
|
}
|
|
#endif
|
|
}
|
|
|
|
PDFEntry* pRet = NULL;
|
|
unsigned int nEntries = aGrammar.m_aObjectStack.size();
|
|
if( nEntries == 1 )
|
|
{
|
|
pRet = aGrammar.m_aObjectStack.back();
|
|
aGrammar.m_aObjectStack.pop_back();
|
|
}
|
|
#if OSL_DEBUG_LEVEL > 1
|
|
else if( nEntries > 1 )
|
|
fprintf( stderr, "error got %u stack objects in parse\n", nEntries );
|
|
#endif
|
|
|
|
return pRet;
|
|
}
|
|
#endif
|
|
|
|
PDFEntry* PDFReader::read( const char* pFileName )
|
|
{
|
|
#ifdef WIN32
|
|
/* #i106583#
|
|
since converting to boost 1.39 file_iterator does not work anymore on all Windows systems
|
|
C++ stdlib istream_iterator does not allow "-" apparently
|
|
using spirit 2.0 doesn't work in our environment with the MSC
|
|
|
|
So for the time being bite the bullet and read the whole file.
|
|
FIXME: give Spirit 2.x another try when we upgrade boost again.
|
|
*/
|
|
PDFEntry* pRet = NULL;
|
|
FILE* fp = fopen( pFileName, "rb" );
|
|
if( fp )
|
|
{
|
|
fseek( fp, 0, SEEK_END );
|
|
unsigned int nLen = (unsigned int)ftell( fp );
|
|
fseek( fp, 0, SEEK_SET );
|
|
char* pBuf = (char*)rtl_allocateMemory( nLen );
|
|
if( pBuf )
|
|
{
|
|
fread( pBuf, 1, nLen, fp );
|
|
pRet = read( pBuf, nLen );
|
|
rtl_freeMemory( pBuf );
|
|
}
|
|
fclose( fp );
|
|
}
|
|
return pRet;
|
|
#else
|
|
file_iterator<> file_start( pFileName );
|
|
if( ! file_start )
|
|
return NULL;
|
|
file_iterator<> file_end = file_start.make_end();
|
|
PDFGrammar< file_iterator<> > aGrammar( file_start );
|
|
|
|
try
|
|
{
|
|
#if OSL_DEBUG_LEVEL > 1
|
|
boost::spirit::parse_info< file_iterator<> > aInfo =
|
|
#endif
|
|
boost::spirit::parse( file_start,
|
|
file_end,
|
|
aGrammar,
|
|
boost::spirit::space_p );
|
|
#if OSL_DEBUG_LEVEL > 1
|
|
fprintf( stderr, "parseinfo: stop at offset = %ld, hit = %s, full = %s, length = %lu\n",
|
|
aInfo.stop - file_start,
|
|
aInfo.hit ? "true" : "false",
|
|
aInfo.full ? "true" : "false",
|
|
aInfo.length );
|
|
#endif
|
|
}
|
|
catch( const parser_error< const char*, file_iterator<> >& rError )
|
|
{
|
|
#if OSL_DEBUG_LEVEL > 1
|
|
fprintf( stderr, "parse error: %s at buffer pos %lu\nobject stack:\n",
|
|
rError.descriptor, rError.where - file_start );
|
|
size_t nElem = aGrammar.m_aObjectStack.size();
|
|
for( size_t i = 0; i < nElem; ++i )
|
|
{
|
|
fprintf( stderr, " %s\n", typeid( *(aGrammar.m_aObjectStack[i]) ).name() );
|
|
}
|
|
#endif
|
|
}
|
|
|
|
PDFEntry* pRet = NULL;
|
|
unsigned int nEntries = aGrammar.m_aObjectStack.size();
|
|
if( nEntries == 1 )
|
|
{
|
|
pRet = aGrammar.m_aObjectStack.back();
|
|
aGrammar.m_aObjectStack.pop_back();
|
|
}
|
|
#if OSL_DEBUG_LEVEL > 1
|
|
else if( nEntries > 1 )
|
|
{
|
|
fprintf( stderr, "error got %u stack objects in parse\n", nEntries );
|
|
for( unsigned int i = 0; i < nEntries; i++ )
|
|
{
|
|
fprintf( stderr, "%s\n", typeid(*aGrammar.m_aObjectStack[i]).name() );
|
|
PDFObject* pObj = dynamic_cast<PDFObject*>(aGrammar.m_aObjectStack[i]);
|
|
if( pObj )
|
|
fprintf( stderr, " -> object %d generation %d\n", pObj->m_nNumber, pObj->m_nGeneration );
|
|
else
|
|
fprintf( stderr, "(type %s)\n", typeid(*aGrammar.m_aObjectStack[i]).name() );
|
|
}
|
|
}
|
|
#endif
|
|
return pRet;
|
|
#endif // WIN32
|
|
}
|
|
|
|
#if defined __SUNPRO_CC
|
|
#pragma enable_warn
|
|
#elif defined _MSC_VER
|
|
#pragma warning(pop)
|
|
#endif
|
|
|
|
|
|
/* vim:set shiftwidth=4 softtabstop=4 expandtab: */
|