 Pcbnew: major swig fix.
* Switched hashtables.h over to std::undordered_map from boost version.
* Added new macros DECL_VEC_FOR_SWIG() and DECL_MAP_FOR_SWIG() in macros.h.
These along with future DECL_HASH_FOR_SWIG() unify the declaration to swig
and C++ so that the resultant type name is common in both languages, and
the types AGREE.
* Fixed swigging of NETINFO_ITEM and NETINFO_LIST via magic.
* Newly exposed (python wrapped) are: D_PADS, TRACKS (was TRACK_PTRS),
NETNAME_MAP, NETCODE_MAP, wxString (without constructor purposely, read
comment in wx.i), MARKERS, ZONE_CONTAINERS, NETCLASSPTR, KICAD_T types.
* std::vector<SOMETHING*> tends to end up named SOMETHINGS in C++ and python.
Having the name consistent between like types is helpful, and between
languages. std::map<> ends up as SOMETHING_MAP.
* NETINFO_LIST::m_netNames and NETINFO_LIST::m_netCodes are now std::map
instead of hashtables, because swig does not yet support std::unordered_map.
* You can now get to any netclass or net info. NETNAMES_MAP and NETCODES_MAP
are traversable basically the same as a python dictionary using a python
string (not wsString) as the key! The wxString typemap converts python
string to wxString before the lookup happens. Iteration also works.
9 years ago  Pcbnew: major swig fix.
* Switched hashtables.h over to std::undordered_map from boost version.
* Added new macros DECL_VEC_FOR_SWIG() and DECL_MAP_FOR_SWIG() in macros.h.
These along with future DECL_HASH_FOR_SWIG() unify the declaration to swig
and C++ so that the resultant type name is common in both languages, and
the types AGREE.
* Fixed swigging of NETINFO_ITEM and NETINFO_LIST via magic.
* Newly exposed (python wrapped) are: D_PADS, TRACKS (was TRACK_PTRS),
NETNAME_MAP, NETCODE_MAP, wxString (without constructor purposely, read
comment in wx.i), MARKERS, ZONE_CONTAINERS, NETCLASSPTR, KICAD_T types.
* std::vector<SOMETHING*> tends to end up named SOMETHINGS in C++ and python.
Having the name consistent between like types is helpful, and between
languages. std::map<> ends up as SOMETHING_MAP.
* NETINFO_LIST::m_netNames and NETINFO_LIST::m_netCodes are now std::map
instead of hashtables, because swig does not yet support std::unordered_map.
* You can now get to any netclass or net info. NETNAMES_MAP and NETCODES_MAP
are traversable basically the same as a python dictionary using a python
string (not wsString) as the key! The wxString typemap converts python
string to wxString before the lookup happens. Iteration also works.
9 years ago |
|
/*
* This program source code file is part of KiCad, a free EDA CAD application. * * Copyright (C) 2013 SoftPLC Corporation, Dick Hollenbeck <dick@softplc.com> * Copyright (C) 2013-2021 KiCad Developers, see AUTHORS.txt for contributors. * * @author Dick Hollenbeck * * This program is free software; you can redistribute it and/or * modify it under the terms of the GNU General Public License * as published by the Free Software Foundation; either version 2 * of the License, or (at your option) any later version. * * This program is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU General Public License for more details. * * You should have received a copy of the GNU General Public License * along with this program; if not, you may find one here: * http://www.gnu.org/licenses/old-licenses/gpl-2.0.html
* or you may search the http://www.gnu.org website for the version 2 license,
* or you may write to the Free Software Foundation, Inc., * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA */
#ifndef UTF8_H_
#define UTF8_H_
#include <string>
#include <wx/string.h>
#if defined(DEBUG)
#define UTF8_VERIFY // Might someday be a hidden cmake config option
#endif
/**
* Test a C string to see if it is UTF8 encoded. * * An ASCII string is a valid UTF8 string. */bool IsUTF8( const char* aString );
#if defined(UTF8_VERIFY)
#define MAYBE_VERIFY_UTF8(x) wxASSERT( IsUTF8(x) )
#else
#define MAYBE_VERIFY_UTF8(x) // nothing
#endif
/**
* An 8 bit string that is assuredly encoded in UTF8, and supplies special conversion * support to and from wxString, to and from std::string, and has non-mutating iteration * over Unicode characters. * * I've been careful to supply only conversion facilities and not try and duplicate * wxString() with many member functions. There are multiple ways to create text into * a std::string without the need of too many member functions: * * - richio.h's StrPrintf(). * - std::ostringstream. * * Because this class uses no virtuals, it should be possible to cast any std::string * into a UTF8 using this kind of cast: (UTF8 &) without construction or copying being * the effect of the cast. Be sure the source std::string holds UTF8 encoded text before * you do that. */class UTF8{public: UTF8( const wxString& o );
/// This is a constructor for which you could end up with
/// non-UTF8 encoding, but that would be your fault.
UTF8( const char* txt ) : m_s( txt ) { MAYBE_VERIFY_UTF8( c_str() ); }
/// For use with _() function on wx 2.8.
/// BTW _() on wx >= 2.9 returns wxString, not wchar_t* like on 2.8.
UTF8( const wchar_t* txt );
UTF8( const std::string& o ) : m_s( o ) { MAYBE_VERIFY_UTF8( c_str() ); }
UTF8() { }
~UTF8() // Needed mainly to build python wrapper
{ }
// expose some std::string functions publicly, since base class must be private.
const char* c_str() const { return m_s.c_str(); } bool empty() const { return m_s.empty(); }
std::string::size_type find( char c ) const { return m_s.find( c ); } std::string::size_type find( char c, size_t s ) const { return m_s.find( c, s ); }
void clear() { m_s.clear(); } std::string::size_type length() const { return m_s.length(); } std::string::size_type size() const { return m_s.size(); } int compare( const std::string& s ) const { return m_s.compare( s ); }
bool operator==( const UTF8& rhs ) const { return m_s == rhs.m_s; } bool operator==( const std::string& rhs ) const { return m_s == rhs; } bool operator==( const char* s ) const { return m_s == s; }
std::string::size_type find_first_of( const std::string& str, std::string::size_type pos = 0 ) const { return m_s.find_first_of( str, pos ); }
UTF8& operator+=( const UTF8& str ) { m_s += str.m_s; MAYBE_VERIFY_UTF8( c_str() ); return *this; }
UTF8& operator+=( char ch ) { m_s.operator+=( ch ); MAYBE_VERIFY_UTF8( c_str() ); return *this; }
UTF8& operator+=( const char* s ) { m_s.operator+=( s ); MAYBE_VERIFY_UTF8( c_str() ); return *this; }
/// Append a wide (unicode) char to the UTF8 string.
/// if this wide char is not a ASCII7 char, it will be added as a UTF8 multibyte sequence
/// @param w_ch is a UTF-16 value (can be a UTF-32 on Linux)
UTF8& operator+=( unsigned w_ch );
// std::string::npos is not constexpr, so we can't use it in an
// initializer.
static constexpr std::string::size_type npos = -1;
UTF8& operator=( const wxString& o );
UTF8& operator=( const std::string& o ) { m_s = o; MAYBE_VERIFY_UTF8( c_str() ); return *this; }
UTF8& operator=( const char* s ) { m_s = s; MAYBE_VERIFY_UTF8( c_str() ); return *this; }
UTF8& operator=( char c ) { m_s = c; MAYBE_VERIFY_UTF8( c_str() ); return *this; }
// a substring of a UTF8 is not necessarily a UTF8 if a multibyte character
// was split, so return std::string not UTF8
std::string substr( size_t pos = 0, size_t len = npos ) const { return m_s.substr( pos, len ); }
operator const std::string& () const { return m_s; } //operator std::string& () { return m_s; }
//operator std::string () const { return m_s; }
wxString wx_str() const; operator wxString () const;
// "Read only" iterating over bytes is done with these, use the uni_iter to iterate
// over UTF8 (multi-byte) characters
std::string::const_iterator begin() const { return m_s.begin(); } std::string::const_iterator end() const { return m_s.end(); }
#ifndef SWIG
/**
* uni_iter * is a non-mutating iterator that walks through unicode code points in the UTF8 encoded * string. The normal ++(), ++(int), ->(), and *() operators are all supported * for read only access and some return an unsigned holding the unicode character * appropriate for the respective operator. */ class uni_iter { public: uni_iter() // Needed only to build python wrapper, not used outside the wrapper
{ it = nullptr; }
uni_iter( const uni_iter& o ) { it = o.it; }
/// pre-increment and return uni_iter at new position
const uni_iter& operator++() { it += uni_forward( it ); return *this; }
/// post-increment and return uni_iter at initial position
uni_iter operator++( int ) { uni_iter ret = *this;
it += uni_forward( it ); return ret; }
/// return unicode at current position
unsigned operator->() const { unsigned result;
// grab the result, do not advance
uni_forward( it, &result ); return result; }
/// return unicode at current position
unsigned operator*() const { unsigned result;
// grab the result, do not advance
uni_forward( it, &result ); return result; }
uni_iter operator-( int aVal ) const { return uni_iter( (char*) it - aVal ); }
bool operator==( const uni_iter& other ) const { return it == other.it; } bool operator!=( const uni_iter& other ) const { return it != other.it; }
/// Since the ++ operators advance more than one byte, this is your best
/// loop termination test, < end(), not == end().
bool operator< ( const uni_iter& other ) const { return it < other.it; } bool operator<=( const uni_iter& other ) const { return it <= other.it; } bool operator> ( const uni_iter& other ) const { return it > other.it; } bool operator>=( const uni_iter& other ) const { return it >= other.it; }
private: friend class UTF8;
const unsigned char* it;
// private constructor
uni_iter( const char* start ) : it( (const unsigned char*) start ) { } };
/**
* Returns a @a uni_iter initialized to the start of "this" UTF8 byte sequence. */ uni_iter ubegin() const { return uni_iter( m_s.data() ); }
/**
* Return a @a uni_iter initialized to the end of "this" UTF8 byte sequence. */ uni_iter uend() const { return uni_iter( m_s.data() + m_s.size() ); }
/**
* Advance over a single UTF8 encoded multibyte character, capturing the Unicode character * as it goes, and returning the number of bytes consumed. * * @param aSequence is the UTF8 byte sequence, must be aligned on start of character. * @param aResult is where to put the unicode character, and may be NULL if no interest. * @return the count of bytes consumed. */ static int uni_forward( const unsigned char* aSequence, unsigned* aResult = nullptr );#endif // SWIG
protected: std::string m_s;};
#endif // UTF8_H_
|