diff options
Diffstat (limited to '')
-rw-r--r-- | include/oox/helper/helper.hxx | 301 |
1 files changed, 301 insertions, 0 deletions
diff --git a/include/oox/helper/helper.hxx b/include/oox/helper/helper.hxx new file mode 100644 index 000000000..63718ca0e --- /dev/null +++ b/include/oox/helper/helper.hxx @@ -0,0 +1,301 @@ +/* -*- 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 . + */ + +#ifndef INCLUDED_OOX_HELPER_HELPER_HXX +#define INCLUDED_OOX_HELPER_HELPER_HXX + +#include <sal/config.h> + +#include <cstring> +#include <limits> + +#include <o3tl/safeint.hxx> +#include <osl/endian.h> +#include <rtl/math.hxx> +#include <sal/macros.h> +#include <sal/types.h> +#include <tools/color.hxx> + +namespace oox { + +// Helper macros ============================================================== + +namespace detail { + +//TODO: Temporary helper for STATIC_ARRAY_SELECT; ultimately, the latter should be replaced by a +// proper function (template): +template<typename T> constexpr std::make_unsigned_t<T> make_unsigned(T value) { + if constexpr (std::is_signed_v<T>) { + return o3tl::make_unsigned(value); + } else { + return value; + } +} + +} + +/** Expands to the 'index'-th element of a STATIC data array, or to 'def', if + 'index' is out of the array limits. */ +#define STATIC_ARRAY_SELECT( array, index, def ) \ + ((detail::make_unsigned(index) < SAL_N_ELEMENTS(array)) ? ((array)[static_cast<size_t>(index)]) : (def)) + +// Common constants =========================================================== + +const sal_uInt8 WINDOWS_CHARSET_ANSI = 0; +const sal_uInt8 WINDOWS_CHARSET_DEFAULT = 1; +const sal_uInt8 WINDOWS_CHARSET_SYMBOL = 2; +const sal_uInt8 WINDOWS_CHARSET_APPLE_ROMAN = 77; +const sal_uInt8 WINDOWS_CHARSET_SHIFTJIS = 128; +const sal_uInt8 WINDOWS_CHARSET_HANGEUL = 129; +const sal_uInt8 WINDOWS_CHARSET_JOHAB = 130; +const sal_uInt8 WINDOWS_CHARSET_GB2312 = 134; +const sal_uInt8 WINDOWS_CHARSET_BIG5 = 136; +const sal_uInt8 WINDOWS_CHARSET_GREEK = 161; +const sal_uInt8 WINDOWS_CHARSET_TURKISH = 162; +const sal_uInt8 WINDOWS_CHARSET_VIETNAMESE = 163; +const sal_uInt8 WINDOWS_CHARSET_HEBREW = 177; +const sal_uInt8 WINDOWS_CHARSET_ARABIC = 178; +const sal_uInt8 WINDOWS_CHARSET_BALTIC = 186; +const sal_uInt8 WINDOWS_CHARSET_RUSSIAN = 204; +const sal_uInt8 WINDOWS_CHARSET_THAI = 222; +const sal_uInt8 WINDOWS_CHARSET_EASTERN = 238; +const sal_uInt8 WINDOWS_CHARSET_OEM = 255; + + +const ::Color API_RGB_TRANSPARENT (ColorTransparency, 0xffffffff); ///< Transparent color for API calls. +const sal_uInt32 UNSIGNED_RGB_TRANSPARENT = static_cast<sal_uInt32>(-1); ///< Transparent color for unsigned int32 places. +const ::Color API_RGB_BLACK (0x000000); ///< Black color for API calls. +const ::Color API_RGB_GRAY (0x808080); ///< Gray color for API calls. +const ::Color API_RGB_WHITE (0xFFFFFF); ///< White color for API calls. + +const sal_Int16 API_LINE_SOLID = 0; +const sal_Int16 API_LINE_DOTTED = 1; +const sal_Int16 API_LINE_DASHED = 2; +const sal_Int16 API_FINE_LINE_DASHED = 14; + +const sal_Int16 API_LINE_NONE = 0; +const sal_Int16 API_LINE_HAIR = 2; +const sal_Int16 API_LINE_THIN = 35; +const sal_Int16 API_LINE_MEDIUM = 88; +const sal_Int16 API_LINE_THICK = 141; + +const sal_Int16 API_ESCAPE_NONE = 0; ///< No escapement. +const sal_Int16 API_ESCAPE_SUPERSCRIPT = 101; ///< Superscript: raise characters automatically (magic value 101). +const sal_Int16 API_ESCAPE_SUBSCRIPT = -101; ///< Subscript: lower characters automatically (magic value -101). + +const sal_Int8 API_ESCAPEHEIGHT_NONE = 100; ///< Relative character height if not escaped. +const sal_Int8 API_ESCAPEHEIGHT_DEFAULT = 58; ///< Relative character height if escaped. + + +// Limitate values ------------------------------------------------------------ + +template< typename ReturnType, typename Type > +inline ReturnType getLimitedValue( Type nValue, Type nMin, Type nMax ) +{ + return static_cast< ReturnType >( ::std::clamp( nValue, nMin, nMax ) ); +} + +template< typename ReturnType, typename Type > +inline ReturnType getIntervalValue( Type nValue, Type nBegin, Type nEnd ) +{ + static_assert(::std::numeric_limits< Type >::is_integer, "is integer"); + Type nInterval = nEnd - nBegin; + Type nCount = (nValue < nBegin) ? -((nBegin - nValue - 1) / nInterval + 1) : ((nValue - nBegin) / nInterval); + return static_cast< ReturnType >( nValue - nCount * nInterval ); +} + +template< typename ReturnType > +inline ReturnType getDoubleIntervalValue( double fValue, double fBegin, double fEnd ) +{ + double fInterval = fEnd - fBegin; + double fCount = (fValue < fBegin) ? -(::rtl::math::approxFloor( (fBegin - fValue - 1.0) / fInterval ) + 1.0) : ::rtl::math::approxFloor( (fValue - fBegin) / fInterval ); + return static_cast< ReturnType >( fValue - fCount * fInterval ); +} + +// Read from bitfields -------------------------------------------------------- + +/** Returns true, if at least one of the bits set in nMask is set in nBitField. */ +template< typename Type > +inline bool getFlag( Type nBitField, Type nMask ) +{ + return (nBitField & nMask) != 0; +} + +/** Returns nSet, if at least one bit of nMask is set in nBitField, otherwise nUnset. */ +template< typename ReturnType, typename Type > +inline ReturnType getFlagValue( Type nBitField, Type nMask, ReturnType nSet, ReturnType nUnset ) +{ + return getFlag( nBitField, nMask ) ? nSet : nUnset; +} + +/** Extracts a value from a bit field. + + Returns the data fragment from nBitField, that starts at bit nStartBit + (0-based, bit 0 is rightmost) with the width of nBitCount. The returned + value will be right-aligned (normalized). + For instance: extractValue<T>(0x4321,8,4) returns 3 (value in bits 8-11). + */ +template< typename ReturnType, typename Type > +inline ReturnType extractValue( Type nBitField, sal_uInt8 nStartBit, sal_uInt8 nBitCount ) +{ + sal_uInt64 nMask = 1; nMask <<= nBitCount; --nMask; + return static_cast< ReturnType >( nMask & (nBitField >> nStartBit) ); +} + +// Write to bitfields --------------------------------------------------------- + +/** Sets or clears (according to bSet) all set bits of nMask in ornBitField. */ +template< typename Type > +inline void setFlag( Type& ornBitField, Type nMask, bool bSet = true ) +{ + if( bSet ) ornBitField |= nMask; else ornBitField &= ~nMask; +} + + +/** Optional value, similar to ::std::optional<>, with convenience accessors. + */ +template< typename Type > +class OptValue +{ +public: + OptValue() : maValue(), mbHasValue( false ) {} + explicit OptValue( const Type& rValue ) : maValue( rValue ), mbHasValue( true ) {} + explicit OptValue( bool bHasValue, const Type& rValue ) : maValue( rValue ), mbHasValue( bHasValue ) {} + + bool has() const { return mbHasValue; } + bool operator!() const { return !mbHasValue; } + bool differsFrom( const Type& rValue ) const { return mbHasValue && (maValue != rValue); } + + const Type& get() const { return maValue; } + const Type& get( const Type& rDefValue ) const { return mbHasValue ? maValue : rDefValue; } + + void set( const Type& rValue ) { maValue = rValue; mbHasValue = true; } + Type& use() { mbHasValue = true; return maValue; } + + OptValue& operator=( const Type& rValue ) { set( rValue ); return *this; } + bool operator==( const OptValue& rValue ) const { + return ( ( !mbHasValue && rValue.mbHasValue == false ) || + ( mbHasValue == rValue.mbHasValue && maValue == rValue.maValue ) ); + } + void assignIfUsed( const OptValue& rValue ) { if( rValue.mbHasValue ) set( rValue.maValue ); } + +private: + Type maValue; + bool mbHasValue; +}; + + +/** Provides platform independent functions to convert from or to little-endian + byte order, e.g. for reading data from or writing data to memory or a + binary stream. + + On big-endian platforms, the byte order in the passed values is swapped, + this can be used for converting big-endian to and from little-endian data. + + On little-endian platforms, the conversion functions are implemented empty, + thus compilers should completely optimize away the function call. + */ +class ByteOrderConverter +{ +public: +#ifdef OSL_BIGENDIAN + static void convertLittleEndian( sal_Int8& ) {} // present for usage in templates + static void convertLittleEndian( sal_uInt8& ) {} // present for usage in templates + static void convertLittleEndian( char16_t& rnValue ) { swap2( reinterpret_cast< sal_uInt8* >( &rnValue ) ); } + static void convertLittleEndian( sal_Int16& rnValue ) { swap2( reinterpret_cast< sal_uInt8* >( &rnValue ) ); } + static void convertLittleEndian( sal_uInt16& rnValue ) { swap2( reinterpret_cast< sal_uInt8* >( &rnValue ) ); } + static void convertLittleEndian( sal_Int32& rnValue ) { swap4( reinterpret_cast< sal_uInt8* >( &rnValue ) ); } + static void convertLittleEndian( sal_uInt32& rnValue ) { swap4( reinterpret_cast< sal_uInt8* >( &rnValue ) ); } + static void convertLittleEndian( sal_Int64& rnValue ) { swap8( reinterpret_cast< sal_uInt8* >( &rnValue ) ); } + static void convertLittleEndian( sal_uInt64& rnValue ) { swap8( reinterpret_cast< sal_uInt8* >( &rnValue ) ); } + static void convertLittleEndian( float& rfValue ) { swap4( reinterpret_cast< sal_uInt8* >( &rfValue ) ); } + static void convertLittleEndian( double& rfValue ) { swap8( reinterpret_cast< sal_uInt8* >( &rfValue ) ); } + + template< typename Type > + inline static void convertLittleEndianArray( Type* pnArray, size_t nElemCount ); + + static void convertLittleEndianArray( sal_Int8*, size_t ) {} + static void convertLittleEndianArray( sal_uInt8*, size_t ) {} + +#else + template< typename Type > + static void convertLittleEndian( Type& ) {} + + template< typename Type > + static void convertLittleEndianArray( Type*, size_t ) {} + +#endif + + /** Writes a value to memory, while converting it to little-endian. + @param pDstBuffer The memory buffer to write the value to. + @param nValue The value to be written to memory in little-endian. + */ + template< typename Type > + inline static void writeLittleEndian( void* pDstBuffer, Type nValue ); + +#ifdef OSL_BIGENDIAN +private: + inline static void swap2( sal_uInt8* pnData ); + inline static void swap4( sal_uInt8* pnData ); + inline static void swap8( sal_uInt8* pnData ); +#endif +}; + + +template< typename Type > +inline void ByteOrderConverter::writeLittleEndian( void* pDstBuffer, Type nValue ) +{ + convertLittleEndian( nValue ); + memcpy( pDstBuffer, &nValue, sizeof( Type ) ); +} + +#ifdef OSL_BIGENDIAN +template< typename Type > +inline void ByteOrderConverter::convertLittleEndianArray( Type* pnArray, size_t nElemCount ) +{ + for( Type* pnArrayEnd = pnArray + nElemCount; pnArray != pnArrayEnd; ++pnArray ) + convertLittleEndian( *pnArray ); +} + +inline void ByteOrderConverter::swap2( sal_uInt8* pnData ) +{ + ::std::swap( pnData[ 0 ], pnData[ 1 ] ); +} + +inline void ByteOrderConverter::swap4( sal_uInt8* pnData ) +{ + ::std::swap( pnData[ 0 ], pnData[ 3 ] ); + ::std::swap( pnData[ 1 ], pnData[ 2 ] ); +} + +inline void ByteOrderConverter::swap8( sal_uInt8* pnData ) +{ + ::std::swap( pnData[ 0 ], pnData[ 7 ] ); + ::std::swap( pnData[ 1 ], pnData[ 6 ] ); + ::std::swap( pnData[ 2 ], pnData[ 5 ] ); + ::std::swap( pnData[ 3 ], pnData[ 4 ] ); +} +#endif + + +} // namespace oox + +#endif + +/* vim:set shiftwidth=4 softtabstop=4 expandtab: */ |