non-pch build fix

[wxWidgets.git] / src / common / translation.cpp

/////////////////////////////////////////////////////////////////////////////
// Name:        src/common/translation.cpp
// Purpose:     Internationalization and localisation for wxWidgets
// Author:      Vadim Zeitlin, Vaclav Slavik,
//              Michael N. Filippov <michael@idisys.iae.nsk.su>
//              (2003/09/30 - PluralForms support)
// Created:     2010-04-23
// RCS-ID:      $Id$
// Copyright:   (c) 1998 Vadim Zeitlin <zeitlin@dptmaths.ens-cachan.fr>
// Licence:     wxWindows licence
/////////////////////////////////////////////////////////////////////////////

// ============================================================================
// declaration
// ============================================================================

// ----------------------------------------------------------------------------
// headers
// ----------------------------------------------------------------------------

// For compilers that support precompilation, includes "wx.h".
#include "wx/wxprec.h"

#ifdef __BORLANDC__
    #pragma hdrstop
#endif

#if wxUSE_INTL

#ifndef WX_PRECOMP
    #include "wx/dynarray.h"
    #include "wx/string.h"
    #include "wx/intl.h"
    #include "wx/log.h"
    #include "wx/utils.h"
    #include "wx/hashmap.h"
    #include "wx/module.h"
#endif // WX_PRECOMP

// standard headers
#include <ctype.h>
#include <stdlib.h>

#include "wx/file.h"
#include "wx/filename.h"
#include "wx/tokenzr.h"
#include "wx/fontmap.h"
#include "wx/scopedptr.h"
#include "wx/stdpaths.h"
#include "wx/hashset.h"

// ----------------------------------------------------------------------------
// simple types
// ----------------------------------------------------------------------------

typedef wxUint32 size_t32;

// ----------------------------------------------------------------------------
// constants
// ----------------------------------------------------------------------------

// magic number identifying the .mo format file
const size_t32 MSGCATALOG_MAGIC    = 0x950412de;
const size_t32 MSGCATALOG_MAGIC_SW = 0xde120495;

#define TRACE_I18N wxS("i18n")

// the constants describing the format of ll_CC locale string
static const size_t LEN_LANG = 2;

// ----------------------------------------------------------------------------
// global functions
// ----------------------------------------------------------------------------

namespace
{

// get just the language part
inline wxString ExtractLang(const wxString& langFull)
{
    return langFull.Left(LEN_LANG);
}

} // anonymous namespace


// ============================================================================
// implementation
// ============================================================================

// ----------------------------------------------------------------------------
// Plural forms parser
// ----------------------------------------------------------------------------

/*
                                Simplified Grammar

Expression:
    LogicalOrExpression '?' Expression ':' Expression
    LogicalOrExpression

LogicalOrExpression:
    LogicalAndExpression "||" LogicalOrExpression   // to (a || b) || c
    LogicalAndExpression

LogicalAndExpression:
    EqualityExpression "&&" LogicalAndExpression    // to (a && b) && c
    EqualityExpression

EqualityExpression:
    RelationalExpression "==" RelationalExperession
    RelationalExpression "!=" RelationalExperession
    RelationalExpression

RelationalExpression:
    MultiplicativeExpression '>' MultiplicativeExpression
    MultiplicativeExpression '<' MultiplicativeExpression
    MultiplicativeExpression ">=" MultiplicativeExpression
    MultiplicativeExpression "<=" MultiplicativeExpression
    MultiplicativeExpression

MultiplicativeExpression:
    PmExpression '%' PmExpression
    PmExpression

PmExpression:
    N
    Number
    '(' Expression ')'
*/

class wxPluralFormsToken
{
public:
    enum Type
    {
        T_ERROR, T_EOF, T_NUMBER, T_N, T_PLURAL, T_NPLURALS, T_EQUAL, T_ASSIGN,
        T_GREATER, T_GREATER_OR_EQUAL, T_LESS, T_LESS_OR_EQUAL,
        T_REMINDER, T_NOT_EQUAL,
        T_LOGICAL_AND, T_LOGICAL_OR, T_QUESTION, T_COLON, T_SEMICOLON,
        T_LEFT_BRACKET, T_RIGHT_BRACKET
    };
    Type type() const { return m_type; }
    void setType(Type type) { m_type = type; }
    // for T_NUMBER only
    typedef int Number;
    Number number() const { return m_number; }
    void setNumber(Number num) { m_number = num; }
private:
    Type m_type;
    Number m_number;
};


class wxPluralFormsScanner
{
public:
    wxPluralFormsScanner(const char* s);
    const wxPluralFormsToken& token() const { return m_token; }
    bool nextToken();  // returns false if error
private:
    const char* m_s;
    wxPluralFormsToken m_token;
};

wxPluralFormsScanner::wxPluralFormsScanner(const char* s) : m_s(s)
{
    nextToken();
}

bool wxPluralFormsScanner::nextToken()
{
    wxPluralFormsToken::Type type = wxPluralFormsToken::T_ERROR;
    while (isspace((unsigned char) *m_s))
    {
        ++m_s;
    }
    if (*m_s == 0)
    {
        type = wxPluralFormsToken::T_EOF;
    }
    else if (isdigit((unsigned char) *m_s))
    {
        wxPluralFormsToken::Number number = *m_s++ - '0';
        while (isdigit((unsigned char) *m_s))
        {
            number = number * 10 + (*m_s++ - '0');
        }
        m_token.setNumber(number);
        type = wxPluralFormsToken::T_NUMBER;
    }
    else if (isalpha((unsigned char) *m_s))
    {
        const char* begin = m_s++;
        while (isalnum((unsigned char) *m_s))
        {
            ++m_s;
        }
        size_t size = m_s - begin;
        if (size == 1 && memcmp(begin, "n", size) == 0)
        {
            type = wxPluralFormsToken::T_N;
        }
        else if (size == 6 && memcmp(begin, "plural", size) == 0)
        {
            type = wxPluralFormsToken::T_PLURAL;
        }
        else if (size == 8 && memcmp(begin, "nplurals", size) == 0)
        {
            type = wxPluralFormsToken::T_NPLURALS;
        }
    }
    else if (*m_s == '=')
    {
        ++m_s;
        if (*m_s == '=')
        {
            ++m_s;
            type = wxPluralFormsToken::T_EQUAL;
        }
        else
        {
            type = wxPluralFormsToken::T_ASSIGN;
        }
    }
    else if (*m_s == '>')
    {
        ++m_s;
        if (*m_s == '=')
        {
            ++m_s;
            type = wxPluralFormsToken::T_GREATER_OR_EQUAL;
        }
        else
        {
            type = wxPluralFormsToken::T_GREATER;
        }
    }
    else if (*m_s == '<')
    {
        ++m_s;
        if (*m_s == '=')
        {
            ++m_s;
            type = wxPluralFormsToken::T_LESS_OR_EQUAL;
        }
        else
        {
            type = wxPluralFormsToken::T_LESS;
        }
    }
    else if (*m_s == '%')
    {
        ++m_s;
        type = wxPluralFormsToken::T_REMINDER;
    }
    else if (*m_s == '!' && m_s[1] == '=')
    {
        m_s += 2;
        type = wxPluralFormsToken::T_NOT_EQUAL;
    }
    else if (*m_s == '&' && m_s[1] == '&')
    {
        m_s += 2;
        type = wxPluralFormsToken::T_LOGICAL_AND;
    }
    else if (*m_s == '|' && m_s[1] == '|')
    {
        m_s += 2;
        type = wxPluralFormsToken::T_LOGICAL_OR;
    }
    else if (*m_s == '?')
    {
        ++m_s;
        type = wxPluralFormsToken::T_QUESTION;
    }
    else if (*m_s == ':')
    {
        ++m_s;
        type = wxPluralFormsToken::T_COLON;
    } else if (*m_s == ';') {
        ++m_s;
        type = wxPluralFormsToken::T_SEMICOLON;
    }
    else if (*m_s == '(')
    {
        ++m_s;
        type = wxPluralFormsToken::T_LEFT_BRACKET;
    }
    else if (*m_s == ')')
    {
        ++m_s;
        type = wxPluralFormsToken::T_RIGHT_BRACKET;
    }
    m_token.setType(type);
    return type != wxPluralFormsToken::T_ERROR;
}

class wxPluralFormsNode;

// NB: Can't use wxDEFINE_SCOPED_PTR_TYPE because wxPluralFormsNode is not
//     fully defined yet:
class wxPluralFormsNodePtr
{
public:
    wxPluralFormsNodePtr(wxPluralFormsNode *p = NULL) : m_p(p) {}
    ~wxPluralFormsNodePtr();
    wxPluralFormsNode& operator*() const { return *m_p; }
    wxPluralFormsNode* operator->() const { return m_p; }
    wxPluralFormsNode* get() const { return m_p; }
    wxPluralFormsNode* release();
    void reset(wxPluralFormsNode *p);

private:
    wxPluralFormsNode *m_p;
};

class wxPluralFormsNode
{
public:
    wxPluralFormsNode(const wxPluralFormsToken& token) : m_token(token) {}
    const wxPluralFormsToken& token() const { return m_token; }
    const wxPluralFormsNode* node(size_t i) const
        { return m_nodes[i].get(); }
    void setNode(size_t i, wxPluralFormsNode* n);
    wxPluralFormsNode* releaseNode(size_t i);
    wxPluralFormsToken::Number evaluate(wxPluralFormsToken::Number n) const;

private:
    wxPluralFormsToken m_token;
    wxPluralFormsNodePtr m_nodes[3];
};

wxPluralFormsNodePtr::~wxPluralFormsNodePtr()
{
    delete m_p;
}
wxPluralFormsNode* wxPluralFormsNodePtr::release()
{
    wxPluralFormsNode *p = m_p;
    m_p = NULL;
    return p;
}
void wxPluralFormsNodePtr::reset(wxPluralFormsNode *p)
{
    if (p != m_p)
    {
        delete m_p;
        m_p = p;
    }
}


void wxPluralFormsNode::setNode(size_t i, wxPluralFormsNode* n)
{
    m_nodes[i].reset(n);
}

wxPluralFormsNode*  wxPluralFormsNode::releaseNode(size_t i)
{
    return m_nodes[i].release();
}

wxPluralFormsToken::Number
wxPluralFormsNode::evaluate(wxPluralFormsToken::Number n) const
{
    switch (token().type())
    {
        // leaf
        case wxPluralFormsToken::T_NUMBER:
            return token().number();
        case wxPluralFormsToken::T_N:
            return n;
        // 2 args
        case wxPluralFormsToken::T_EQUAL:
            return node(0)->evaluate(n) == node(1)->evaluate(n);
        case wxPluralFormsToken::T_NOT_EQUAL:
            return node(0)->evaluate(n) != node(1)->evaluate(n);
        case wxPluralFormsToken::T_GREATER:
            return node(0)->evaluate(n) > node(1)->evaluate(n);
        case wxPluralFormsToken::T_GREATER_OR_EQUAL:
            return node(0)->evaluate(n) >= node(1)->evaluate(n);
        case wxPluralFormsToken::T_LESS:
            return node(0)->evaluate(n) < node(1)->evaluate(n);
        case wxPluralFormsToken::T_LESS_OR_EQUAL:
            return node(0)->evaluate(n) <= node(1)->evaluate(n);
        case wxPluralFormsToken::T_REMINDER:
            {
                wxPluralFormsToken::Number number = node(1)->evaluate(n);
                if (number != 0)
                {
                    return node(0)->evaluate(n) % number;
                }
                else
                {
                    return 0;
                }
            }
        case wxPluralFormsToken::T_LOGICAL_AND:
            return node(0)->evaluate(n) && node(1)->evaluate(n);
        case wxPluralFormsToken::T_LOGICAL_OR:
            return node(0)->evaluate(n) || node(1)->evaluate(n);
        // 3 args
        case wxPluralFormsToken::T_QUESTION:
            return node(0)->evaluate(n)
                ? node(1)->evaluate(n)
                : node(2)->evaluate(n);
        default:
            return 0;
    }
}


class wxPluralFormsCalculator
{
public:
    wxPluralFormsCalculator() : m_nplurals(0), m_plural(0) {}

    // input: number, returns msgstr index
    int evaluate(int n) const;

    // input: text after "Plural-Forms:" (e.g. "nplurals=2; plural=(n != 1);"),
    // if s == 0, creates default handler
    // returns 0 if error
    static wxPluralFormsCalculator* make(const char* s = 0);

    ~wxPluralFormsCalculator() {}

    void  init(wxPluralFormsToken::Number nplurals, wxPluralFormsNode* plural);

private:
    wxPluralFormsToken::Number m_nplurals;
    wxPluralFormsNodePtr m_plural;
};

wxDEFINE_SCOPED_PTR_TYPE(wxPluralFormsCalculator)

void wxPluralFormsCalculator::init(wxPluralFormsToken::Number nplurals,
                                wxPluralFormsNode* plural)
{
    m_nplurals = nplurals;
    m_plural.reset(plural);
}

int wxPluralFormsCalculator::evaluate(int n) const
{
    if (m_plural.get() == 0)
    {
        return 0;
    }
    wxPluralFormsToken::Number number = m_plural->evaluate(n);
    if (number < 0 || number > m_nplurals)
    {
        return 0;
    }
    return number;
}


class wxPluralFormsParser
{
public:
    wxPluralFormsParser(wxPluralFormsScanner& scanner) : m_scanner(scanner) {}
    bool parse(wxPluralFormsCalculator& rCalculator);

private:
    wxPluralFormsNode* parsePlural();
    // stops at T_SEMICOLON, returns 0 if error
    wxPluralFormsScanner& m_scanner;
    const wxPluralFormsToken& token() const;
    bool nextToken();

    wxPluralFormsNode* expression();
    wxPluralFormsNode* logicalOrExpression();
    wxPluralFormsNode* logicalAndExpression();
    wxPluralFormsNode* equalityExpression();
    wxPluralFormsNode* multiplicativeExpression();
    wxPluralFormsNode* relationalExpression();
    wxPluralFormsNode* pmExpression();
};

bool wxPluralFormsParser::parse(wxPluralFormsCalculator& rCalculator)
{
    if (token().type() != wxPluralFormsToken::T_NPLURALS)
        return false;
    if (!nextToken())
        return false;
    if (token().type() != wxPluralFormsToken::T_ASSIGN)
        return false;
    if (!nextToken())
        return false;
    if (token().type() != wxPluralFormsToken::T_NUMBER)
        return false;
    wxPluralFormsToken::Number nplurals = token().number();
    if (!nextToken())
        return false;
    if (token().type() != wxPluralFormsToken::T_SEMICOLON)
        return false;
    if (!nextToken())
        return false;
    if (token().type() != wxPluralFormsToken::T_PLURAL)
        return false;
    if (!nextToken())
        return false;
    if (token().type() != wxPluralFormsToken::T_ASSIGN)
        return false;
    if (!nextToken())
        return false;
    wxPluralFormsNode* plural = parsePlural();
    if (plural == 0)
        return false;
    if (token().type() != wxPluralFormsToken::T_SEMICOLON)
        return false;
    if (!nextToken())
        return false;
    if (token().type() != wxPluralFormsToken::T_EOF)
        return false;
    rCalculator.init(nplurals, plural);
    return true;
}

wxPluralFormsNode* wxPluralFormsParser::parsePlural()
{
    wxPluralFormsNode* p = expression();
    if (p == NULL)
    {
        return NULL;
    }
    wxPluralFormsNodePtr n(p);
    if (token().type() != wxPluralFormsToken::T_SEMICOLON)
    {
        return NULL;
    }
    return n.release();
}

const wxPluralFormsToken& wxPluralFormsParser::token() const
{
    return m_scanner.token();
}

bool wxPluralFormsParser::nextToken()
{
    if (!m_scanner.nextToken())
        return false;
    return true;
}

wxPluralFormsNode* wxPluralFormsParser::expression()
{
    wxPluralFormsNode* p = logicalOrExpression();
    if (p == NULL)
        return NULL;
    wxPluralFormsNodePtr n(p);
    if (token().type() == wxPluralFormsToken::T_QUESTION)
    {
        wxPluralFormsNodePtr qn(new wxPluralFormsNode(token()));
        if (!nextToken())
        {
            return 0;
        }
        p = expression();
        if (p == 0)
        {
            return 0;
        }
        qn->setNode(1, p);
        if (token().type() != wxPluralFormsToken::T_COLON)
        {
            return 0;
        }
        if (!nextToken())
        {
            return 0;
        }
        p = expression();
        if (p == 0)
        {
            return 0;
        }
        qn->setNode(2, p);
        qn->setNode(0, n.release());
        return qn.release();
    }
    return n.release();
}

wxPluralFormsNode*wxPluralFormsParser::logicalOrExpression()
{
    wxPluralFormsNode* p = logicalAndExpression();
    if (p == NULL)
        return NULL;
    wxPluralFormsNodePtr ln(p);
    if (token().type() == wxPluralFormsToken::T_LOGICAL_OR)
    {
        wxPluralFormsNodePtr un(new wxPluralFormsNode(token()));
        if (!nextToken())
        {
            return 0;
        }
        p = logicalOrExpression();
        if (p == 0)
        {
            return 0;
        }
        wxPluralFormsNodePtr rn(p);    // right
        if (rn->token().type() == wxPluralFormsToken::T_LOGICAL_OR)
        {
            // see logicalAndExpression comment
            un->setNode(0, ln.release());
            un->setNode(1, rn->releaseNode(0));
            rn->setNode(0, un.release());
            return rn.release();
        }


        un->setNode(0, ln.release());
        un->setNode(1, rn.release());
        return un.release();
    }
    return ln.release();
}

wxPluralFormsNode* wxPluralFormsParser::logicalAndExpression()
{
    wxPluralFormsNode* p = equalityExpression();
    if (p == NULL)
        return NULL;
    wxPluralFormsNodePtr ln(p);   // left
    if (token().type() == wxPluralFormsToken::T_LOGICAL_AND)
    {
        wxPluralFormsNodePtr un(new wxPluralFormsNode(token()));  // up
        if (!nextToken())
        {
            return NULL;
        }
        p = logicalAndExpression();
        if (p == 0)
        {
            return NULL;
        }
        wxPluralFormsNodePtr rn(p);    // right
        if (rn->token().type() == wxPluralFormsToken::T_LOGICAL_AND)
        {
// transform 1 && (2 && 3) -> (1 && 2) && 3
//     u                  r
// l       r     ->   u      3
//       2   3      l   2
            un->setNode(0, ln.release());
            un->setNode(1, rn->releaseNode(0));
            rn->setNode(0, un.release());
            return rn.release();
        }

        un->setNode(0, ln.release());
        un->setNode(1, rn.release());
        return un.release();
    }
    return ln.release();
}

wxPluralFormsNode* wxPluralFormsParser::equalityExpression()
{
    wxPluralFormsNode* p = relationalExpression();
    if (p == NULL)
        return NULL;
    wxPluralFormsNodePtr n(p);
    if (token().type() == wxPluralFormsToken::T_EQUAL
        || token().type() == wxPluralFormsToken::T_NOT_EQUAL)
    {
        wxPluralFormsNodePtr qn(new wxPluralFormsNode(token()));
        if (!nextToken())
        {
            return NULL;
        }
        p = relationalExpression();
        if (p == NULL)
        {
            return NULL;
        }
        qn->setNode(1, p);
        qn->setNode(0, n.release());
        return qn.release();
    }
    return n.release();
}

wxPluralFormsNode* wxPluralFormsParser::relationalExpression()
{
    wxPluralFormsNode* p = multiplicativeExpression();
    if (p == NULL)
        return NULL;
    wxPluralFormsNodePtr n(p);
    if (token().type() == wxPluralFormsToken::T_GREATER
            || token().type() == wxPluralFormsToken::T_LESS
            || token().type() == wxPluralFormsToken::T_GREATER_OR_EQUAL
            || token().type() == wxPluralFormsToken::T_LESS_OR_EQUAL)
    {
        wxPluralFormsNodePtr qn(new wxPluralFormsNode(token()));
        if (!nextToken())
        {
            return NULL;
        }
        p = multiplicativeExpression();
        if (p == NULL)
        {
            return NULL;
        }
        qn->setNode(1, p);
        qn->setNode(0, n.release());
        return qn.release();
    }
    return n.release();
}

wxPluralFormsNode* wxPluralFormsParser::multiplicativeExpression()
{
    wxPluralFormsNode* p = pmExpression();
    if (p == NULL)
        return NULL;
    wxPluralFormsNodePtr n(p);
    if (token().type() == wxPluralFormsToken::T_REMINDER)
    {
        wxPluralFormsNodePtr qn(new wxPluralFormsNode(token()));
        if (!nextToken())
        {
            return NULL;
        }
        p = pmExpression();
        if (p == NULL)
        {
            return NULL;
        }
        qn->setNode(1, p);
        qn->setNode(0, n.release());
        return qn.release();
    }
    return n.release();
}

wxPluralFormsNode* wxPluralFormsParser::pmExpression()
{
    wxPluralFormsNodePtr n;
    if (token().type() == wxPluralFormsToken::T_N
        || token().type() == wxPluralFormsToken::T_NUMBER)
    {
        n.reset(new wxPluralFormsNode(token()));
        if (!nextToken())
        {
            return NULL;
        }
    }
    else if (token().type() == wxPluralFormsToken::T_LEFT_BRACKET) {
        if (!nextToken())
        {
            return NULL;
        }
        wxPluralFormsNode* p = expression();
        if (p == NULL)
        {
            return NULL;
        }
        n.reset(p);
        if (token().type() != wxPluralFormsToken::T_RIGHT_BRACKET)
        {
            return NULL;
        }
        if (!nextToken())
        {
            return NULL;
        }
    }
    else
    {
        return NULL;
    }
    return n.release();
}

wxPluralFormsCalculator* wxPluralFormsCalculator::make(const char* s)
{
    wxPluralFormsCalculatorPtr calculator(new wxPluralFormsCalculator);
    if (s != NULL)
    {
        wxPluralFormsScanner scanner(s);
        wxPluralFormsParser p(scanner);
        if (!p.parse(*calculator))
        {
            return NULL;
        }
    }
    return calculator.release();
}




// ----------------------------------------------------------------------------
// wxMsgCatalogFile corresponds to one disk-file message catalog.
//
// This is a "low-level" class and is used only by wxMsgCatalog
// NOTE: for the documentation of the binary catalog (.MO) files refer to
//       the GNU gettext manual:
//       http://www.gnu.org/software/autoconf/manual/gettext/MO-Files.html
// ----------------------------------------------------------------------------

WX_DECLARE_EXPORTED_STRING_HASH_MAP(wxString, wxMessagesHash);

class wxMsgCatalogFile
{
public:
    // ctor & dtor
    wxMsgCatalogFile();
    ~wxMsgCatalogFile();

    // load the catalog from disk
    bool Load(const wxString& filename,
              wxPluralFormsCalculatorPtr& rPluralFormsCalculator);

    // fills the hash with string-translation pairs
    bool FillHash(wxMessagesHash& hash, const wxString& msgIdCharset) const;

    // return the charset of the strings in this catalog or empty string if
    // none/unknown
    wxString GetCharset() const { return m_charset; }

private:
    // this implementation is binary compatible with GNU gettext() version 0.10

    // an entry in the string table
    struct wxMsgTableEntry
    {
        size_t32   nLen;           // length of the string
        size_t32   ofsString;      // pointer to the string
    };

    // header of a .mo file
    struct wxMsgCatalogHeader
    {
        size_t32  magic,          // offset +00:  magic id
                  revision,       //        +04:  revision
                  numStrings;     //        +08:  number of strings in the file
        size_t32  ofsOrigTable,   //        +0C:  start of original string table
                  ofsTransTable;  //        +10:  start of translated string table
        size_t32  nHashSize,      //        +14:  hash table size
                  ofsHashTable;   //        +18:  offset of hash table start
    };

    // all data is stored here
    wxMemoryBuffer m_data;

    // data description
    size_t32          m_numStrings;   // number of strings in this domain
    wxMsgTableEntry  *m_pOrigTable,   // pointer to original   strings
                     *m_pTransTable;  //            translated

    wxString m_charset;               // from the message catalog header


    // swap the 2 halves of 32 bit integer if needed
    size_t32 Swap(size_t32 ui) const
    {
        return m_bSwapped ? (ui << 24) | ((ui & 0xff00) << 8) |
                            ((ui >> 8) & 0xff00) | (ui >> 24)
                            : ui;
    }

    // just return the pointer to the start of the data as "char *" to
    // facilitate doing pointer arithmetic with it
    char *StringData() const
    {
        return static_cast<char *>(m_data.GetData());
    }

    const char *StringAtOfs(wxMsgTableEntry *pTable, size_t32 n) const
    {
        const wxMsgTableEntry * const ent = pTable + n;

        // this check could fail for a corrupt message catalog
        size_t32 ofsString = Swap(ent->ofsString);
        if ( ofsString + Swap(ent->nLen) > m_data.GetDataLen())
        {
            return NULL;
        }

        return StringData() + ofsString;
    }

    bool m_bSwapped;   // wrong endianness?

    wxDECLARE_NO_COPY_CLASS(wxMsgCatalogFile);
};


// ----------------------------------------------------------------------------
// wxMsgCatalog corresponds to one loaded message catalog.
//
// This is a "low-level" class and is used only by wxLocale (that's why
// it's designed to be stored in a linked list)
// ----------------------------------------------------------------------------

class wxMsgCatalog
{
public:
#if !wxUSE_UNICODE
    wxMsgCatalog() { m_conv = NULL; }
    ~wxMsgCatalog();
#endif

    // load the catalog from disk
    bool Load(const wxString& filename,
              const wxString& domain,
              const wxString& msgIdCharset);

    // get name of the catalog
    wxString GetDomain() const { return m_domain; }

    // get the translated string: returns NULL if not found
    const wxString *GetString(const wxString& sz, size_t n = size_t(-1)) const;

    // public variable pointing to the next element in a linked list (or NULL)
    wxMsgCatalog *m_pNext;

private:
    wxMessagesHash  m_messages; // all messages in the catalog
    wxString        m_domain;   // name of the domain

#if !wxUSE_UNICODE
    // the conversion corresponding to this catalog charset if we installed it
    // as the global one
    wxCSConv *m_conv;
#endif

    wxPluralFormsCalculatorPtr  m_pluralFormsCalculator;
};

// ----------------------------------------------------------------------------
// wxMsgCatalogFile clas
// ----------------------------------------------------------------------------

wxMsgCatalogFile::wxMsgCatalogFile()
{
}

wxMsgCatalogFile::~wxMsgCatalogFile()
{
}

// open disk file and read in it's contents
bool wxMsgCatalogFile::Load(const wxString& filename,
                            wxPluralFormsCalculatorPtr& rPluralFormsCalculator)
{
    wxFile fileMsg(filename);
    if ( !fileMsg.IsOpened() )
        return false;

    // get the file size (assume it is less than 4Gb...)
    wxFileOffset lenFile = fileMsg.Length();
    if ( lenFile == wxInvalidOffset )
        return false;

    size_t nSize = wx_truncate_cast(size_t, lenFile);
    wxASSERT_MSG( nSize == lenFile + size_t(0), wxS("message catalog bigger than 4GB?") );

    // read the whole file in memory
    if ( fileMsg.Read(m_data.GetWriteBuf(nSize), nSize) != lenFile )
        return false;

    m_data.UngetWriteBuf(nSize);


    // examine header
    bool bValid = m_data.GetDataLen() > sizeof(wxMsgCatalogHeader);

    const wxMsgCatalogHeader *pHeader = (wxMsgCatalogHeader *)m_data.GetData();
    if ( bValid ) {
        // we'll have to swap all the integers if it's true
        m_bSwapped = pHeader->magic == MSGCATALOG_MAGIC_SW;

        // check the magic number
        bValid = m_bSwapped || pHeader->magic == MSGCATALOG_MAGIC;
    }

    if ( !bValid ) {
        // it's either too short or has incorrect magic number
        wxLogWarning(_("'%s' is not a valid message catalog."), filename.c_str());

        return false;
    }

    // initialize
    m_numStrings  = Swap(pHeader->numStrings);
    m_pOrigTable  = (wxMsgTableEntry *)(StringData() +
                    Swap(pHeader->ofsOrigTable));
    m_pTransTable = (wxMsgTableEntry *)(StringData() +
                    Swap(pHeader->ofsTransTable));

    // now parse catalog's header and try to extract catalog charset and
    // plural forms formula from it:

    const char* headerData = StringAtOfs(m_pOrigTable, 0);
    if ( headerData && headerData[0] == '\0' )
    {
        // Extract the charset:
        const char * const header = StringAtOfs(m_pTransTable, 0);
        const char *
            cset = strstr(header, "Content-Type: text/plain; charset=");
        if ( cset )
        {
            cset += 34; // strlen("Content-Type: text/plain; charset=")

            const char * const csetEnd = strchr(cset, '\n');
            if ( csetEnd )
            {
                m_charset = wxString(cset, csetEnd - cset);
                if ( m_charset == wxS("CHARSET") )
                {
                    // "CHARSET" is not valid charset, but lazy translator
                    m_charset.empty();
                }
            }
        }
        // else: incorrectly filled Content-Type header

        // Extract plural forms:
        const char * plurals = strstr(header, "Plural-Forms:");
        if ( plurals )
        {
            plurals += 13; // strlen("Plural-Forms:")
            const char * const pluralsEnd = strchr(plurals, '\n');
            if ( pluralsEnd )
            {
                const size_t pluralsLen = pluralsEnd - plurals;
                wxCharBuffer buf(pluralsLen);
                strncpy(buf.data(), plurals, pluralsLen);
                wxPluralFormsCalculator * const
                    pCalculator = wxPluralFormsCalculator::make(buf);
                if ( pCalculator )
                {
                    rPluralFormsCalculator.reset(pCalculator);
                }
                else
                {
                    wxLogVerbose(_("Failed to parse Plural-Forms: '%s'"),
                                 buf.data());
                }
            }
        }

        if ( !rPluralFormsCalculator.get() )
            rPluralFormsCalculator.reset(wxPluralFormsCalculator::make());
    }

    // everything is fine
    return true;
}

bool wxMsgCatalogFile::FillHash(wxMessagesHash& hash,
                                const wxString& msgIdCharset) const
{
    wxUnusedVar(msgIdCharset); // silence warning in Unicode build

    // conversion to use to convert catalog strings to the GUI encoding
    wxMBConv *inputConv = NULL;
    wxMBConv *inputConvPtr = NULL; // same as inputConv but safely deleteable

    if ( !m_charset.empty() )
    {
#if !wxUSE_UNICODE && wxUSE_FONTMAP
        // determine if we need any conversion at all
        wxFontEncoding encCat = wxFontMapperBase::GetEncodingFromName(m_charset);
        if ( encCat != wxLocale::GetSystemEncoding() )
#endif
        {
            inputConvPtr =
            inputConv = new wxCSConv(m_charset);
        }
    }
    else // no need or not possible to convert the encoding
    {
#if wxUSE_UNICODE
        // we must somehow convert the narrow strings in the message catalog to
        // wide strings, so use the default conversion if we have no charset
        inputConv = wxConvCurrent;
#endif
    }

#if !wxUSE_UNICODE
    // conversion to apply to msgid strings before looking them up: we only
    // need it if the msgids are neither in 7 bit ASCII nor in the same
    // encoding as the catalog
    wxCSConv *sourceConv = msgIdCharset.empty() || (msgIdCharset == m_charset)
                            ? NULL
                            : new wxCSConv(msgIdCharset);
#endif // !wxUSE_UNICODE

    for (size_t32 i = 0; i < m_numStrings; i++)
    {
        const char *data = StringAtOfs(m_pOrigTable, i);
        if (!data)
            return false; // may happen for invalid MO files

        wxString msgid;
#if wxUSE_UNICODE
        msgid = wxString(data, *inputConv);
#else // ASCII
        if ( inputConv && sourceConv )
            msgid = wxString(inputConv->cMB2WC(data), *sourceConv);
        else
            msgid = data;
#endif // wxUSE_UNICODE

        data = StringAtOfs(m_pTransTable, i);
        if (!data)
            return false; // may happen for invalid MO files

        size_t length = Swap(m_pTransTable[i].nLen);
        size_t offset = 0;
        size_t index = 0;
        while (offset < length)
        {
            const char * const str = data + offset;

            wxString msgstr;
#if wxUSE_UNICODE
            msgstr = wxString(str, *inputConv);
#else
            if ( inputConv )
                msgstr = wxString(inputConv->cMB2WC(str), *wxConvUI);
            else
                msgstr = str;
#endif // wxUSE_UNICODE/!wxUSE_UNICODE

            if ( !msgstr.empty() )
            {
                hash[index == 0 ? msgid : msgid + wxChar(index)] = msgstr;
            }

            // skip this string
            // IMPORTANT: accesses to the 'data' pointer are valid only for
            //            the first 'length+1' bytes (GNU specs says that the
            //            final NUL is not counted in length); using wxStrnlen()
            //            we make sure we don't access memory beyond the valid range
            //            (which otherwise may happen for invalid MO files):
            offset += wxStrnlen(str, length - offset) + 1;
            ++index;
        }
    }

#if !wxUSE_UNICODE
    delete sourceConv;
#endif
    delete inputConvPtr;

    return true;
}


// ----------------------------------------------------------------------------
// wxMsgCatalog class
// ----------------------------------------------------------------------------

#if !wxUSE_UNICODE
wxMsgCatalog::~wxMsgCatalog()
{
    if ( m_conv )
    {
        if ( wxConvUI == m_conv )
        {
            // we only change wxConvUI if it points to wxConvLocal so we reset
            // it back to it too
            wxConvUI = &wxConvLocal;
        }

        delete m_conv;
    }
}
#endif // !wxUSE_UNICODE

bool wxMsgCatalog::Load(const wxString& filename,
                        const wxString& domain,
                        const wxString& msgIdCharset)
{
    wxMsgCatalogFile file;

    m_domain = domain;

    if ( !file.Load(filename, m_pluralFormsCalculator) )
        return false;

    if ( !file.FillHash(m_messages, msgIdCharset) )
        return false;

    return true;
}

const wxString *wxMsgCatalog::GetString(const wxString& str, size_t n) const
{
    int index = 0;
    if (n != size_t(-1))
    {
        index = m_pluralFormsCalculator->evaluate(n);
    }
    wxMessagesHash::const_iterator i;
    if (index != 0)
    {
        i = m_messages.find(wxString(str) + wxChar(index));   // plural
    }
    else
    {
        i = m_messages.find(str);
    }

    if ( i != m_messages.end() )
    {
        return &i->second;
    }
    else
        return NULL;
}


// ----------------------------------------------------------------------------
// wxTranslations
// ----------------------------------------------------------------------------

namespace
{

wxTranslations *gs_translations = NULL;
bool gs_translationsOwned = false;

} // anonymous namespace


/*static*/
wxTranslations *wxTranslations::Get()
{
    return gs_translations;
}

/*static*/
void wxTranslations::Set(wxTranslations *t)
{
    if ( gs_translationsOwned )
        delete gs_translations;
    gs_translations = t;
    gs_translationsOwned = true;
}

/*static*/
void wxTranslations::SetNonOwned(wxTranslations *t)
{
    if ( gs_translationsOwned )
        delete gs_translations;
    gs_translations = t;
    gs_translationsOwned = false;
}


wxTranslations::wxTranslations()
{
    m_pMsgCat = NULL;
    m_loader = new wxFileTranslationsLoader;
}


wxTranslations::~wxTranslations()
{
    delete m_loader;

    // free catalogs memory
    wxMsgCatalog *pTmpCat;
    while ( m_pMsgCat != NULL )
    {
        pTmpCat = m_pMsgCat;
        m_pMsgCat = m_pMsgCat->m_pNext;
        delete pTmpCat;
    }
}


void wxTranslations::SetLoader(wxTranslationsLoader *loader)
{
    wxCHECK_RET( loader, "loader can't be NULL" );

    delete m_loader;
    m_loader = loader;
}


void wxTranslations::SetLanguage(wxLanguage lang)
{
    if ( lang == wxLANGUAGE_DEFAULT )
        SetLanguage("");
    else
        SetLanguage(wxLocale::GetLanguageCanonicalName(lang));
}

void wxTranslations::SetLanguage(const wxString& lang)
{
    m_lang = lang;
}


bool wxTranslations::AddStdCatalog()
{
    if ( !AddCatalog(wxS("wxstd")) )
        return false;

    // there may be a catalog with toolkit specific overrides, it is not
    // an error if this does not exist
    wxString port(wxPlatformInfo::Get().GetPortIdName());
    if ( !port.empty() )
    {
        AddCatalog(port.BeforeFirst(wxS('/')).MakeLower());
    }

    return true;
}


bool wxTranslations::AddCatalog(const wxString& domain)
{
    return AddCatalog(domain, wxLANGUAGE_ENGLISH_US);
}

#if !wxUSE_UNICODE
bool wxTranslations::AddCatalog(const wxString& domain,
                                wxLanguage msgIdLanguage,
                                const wxString& msgIdCharset)
{
    m_msgIdCharset[domain] = msgIdCharset;
    return AddCatalog(domain, msgIdLanguage);
}
#endif // !wxUSE_UNICODE

bool wxTranslations::AddCatalog(const wxString& domain,
                                wxLanguage msgIdLanguage)
{
    const wxString msgIdLang = wxLocale::GetLanguageCanonicalName(msgIdLanguage);
    const wxString domain_lang = ChooseLanguageForDomain(domain, msgIdLang);

    if ( domain_lang.empty() )
    {
        wxLogTrace(TRACE_I18N,
                    wxS("no suitable translation for domain '%s' found"),
                    domain);
        return false;
    }

    wxLogTrace(TRACE_I18N,
                wxS("adding '%s' translation for domain '%s' (msgid language '%s')"),
                domain_lang, domain, msgIdLang);

    // It is OK to not load catalog if the msgid language and m_language match,
    // in which case we can directly display the texts embedded in program's
    // source code:
    if ( msgIdLang == domain_lang )
        return true;

    wxCHECK_MSG( m_loader, false, "loader can't be NULL" );
    return m_loader->LoadCatalog(this, domain, domain_lang);
}


// check if the given catalog is loaded
bool wxTranslations::IsLoaded(const wxString& domain) const
{
    return FindCatalog(domain) != NULL;
}


bool wxTranslations::LoadCatalogFile(const wxString& filename,
                                     const wxString& domain)
{
    wxMsgCatalog *pMsgCat = new wxMsgCatalog;

#if wxUSE_UNICODE
    const bool ok = pMsgCat->Load(filename, domain, wxEmptyString/*unused*/);
#else
    const bool ok = pMsgCat->Load(filename, domain,
                                  m_msgIdCharset[domain]);
#endif

    if ( !ok )
    {
        // don't add it because it couldn't be loaded anyway
        delete pMsgCat;
        return false;
    }

    // add it to the head of the list so that in GetString it will
    // be searched before the catalogs added earlier
    pMsgCat->m_pNext = m_pMsgCat;
    m_pMsgCat = pMsgCat;

    return true;
}


wxString wxTranslations::ChooseLanguageForDomain(const wxString& WXUNUSED(domain),
                                                 const wxString& WXUNUSED(msgIdLang))
{
    // explicitly set language should always be respected
    if ( !m_lang.empty() )
        return m_lang;

    // TODO: if the default language is used, pick the best (by comparing
    //       available languages with user's preferences), instead of blindly
    //       trusting availability of system language translation
    return wxLocale::GetLanguageCanonicalName(wxLocale::GetSystemLanguage());
}


namespace
{
WX_DECLARE_HASH_SET(wxString, wxStringHash, wxStringEqual,
                    wxLocaleUntranslatedStrings);
}

/* static */
const wxString& wxTranslations::GetUntranslatedString(const wxString& str)
{
    static wxLocaleUntranslatedStrings s_strings;

    wxLocaleUntranslatedStrings::iterator i = s_strings.find(str);
    if ( i == s_strings.end() )
        return *s_strings.insert(str).first;

    return *i;
}


const wxString& wxTranslations::GetString(const wxString& origString,
                                          const wxString& domain) const
{
    return GetString(origString, origString, size_t(-1), domain);
}

const wxString& wxTranslations::GetString(const wxString& origString,
                                          const wxString& origString2,
                                          size_t n,
                                          const wxString& domain) const
{
    if ( origString.empty() )
        return GetUntranslatedString(origString);

    const wxString *trans = NULL;
    wxMsgCatalog *pMsgCat;

    if ( !domain.empty() )
    {
        pMsgCat = FindCatalog(domain);

        // does the catalog exist?
        if ( pMsgCat != NULL )
            trans = pMsgCat->GetString(origString, n);
    }
    else
    {
        // search in all domains
        for ( pMsgCat = m_pMsgCat; pMsgCat != NULL; pMsgCat = pMsgCat->m_pNext )
        {
            trans = pMsgCat->GetString(origString, n);
            if ( trans != NULL )   // take the first found
                break;
        }
    }

    if ( trans == NULL )
    {
        wxLogTrace
        (
            TRACE_I18N,
            "string \"%s\"%s not found in %slocale '%s'.",
            origString,
            ((long)n) != -1 ? wxString::Format("[%ld]", (long)n) : wxString(),
            !domain.empty() ? wxString::Format("domain '%s' ", domain) : wxString(),
            m_lang
        );

        if (n == size_t(-1))
            return GetUntranslatedString(origString);
        else
            return GetUntranslatedString(n == 1 ? origString : origString2);
    }

    return *trans;
}


wxString wxTranslations::GetHeaderValue(const wxString& header,
                                        const wxString& domain) const
{
    if ( header.empty() )
        return wxEmptyString;

    const wxString *trans = NULL;
    wxMsgCatalog *pMsgCat;

    if ( !domain.empty() )
    {
        pMsgCat = FindCatalog(domain);

        // does the catalog exist?
        if ( pMsgCat == NULL )
            return wxEmptyString;

        trans = pMsgCat->GetString(wxEmptyString, (size_t)-1);
    }
    else
    {
        // search in all domains
        for ( pMsgCat = m_pMsgCat; pMsgCat != NULL; pMsgCat = pMsgCat->m_pNext )
        {
            trans = pMsgCat->GetString(wxEmptyString, (size_t)-1);
            if ( trans != NULL )   // take the first found
                break;
        }
    }

    if ( !trans || trans->empty() )
        return wxEmptyString;

    size_t found = trans->find(header);
    if ( found == wxString::npos )
        return wxEmptyString;

    found += header.length() + 2 /* ': ' */;

    // Every header is separated by \n

    size_t endLine = trans->find(wxS('\n'), found);
    size_t len = (endLine == wxString::npos) ?
                wxString::npos : (endLine - found);

    return trans->substr(found, len);
}


// find catalog by name in a linked list, return NULL if !found
wxMsgCatalog *wxTranslations::FindCatalog(const wxString& domain) const
{
    // linear search in the linked list
    wxMsgCatalog *pMsgCat;
    for ( pMsgCat = m_pMsgCat; pMsgCat != NULL; pMsgCat = pMsgCat->m_pNext )
    {
        if ( pMsgCat->GetDomain() == domain )
            return pMsgCat;
    }

    return NULL;
}

// ----------------------------------------------------------------------------
// wxFileTranslationsLoader
// ----------------------------------------------------------------------------

namespace
{

// the list of the directories to search for message catalog files
wxArrayString gs_searchPrefixes;

// return the directories to search for message catalogs under the given
// prefix, separated by wxPATH_SEP
wxString GetMsgCatalogSubdirs(const wxString& prefix, const wxString& lang)
{
    // Search first in Unix-standard prefix/lang/LC_MESSAGES, then in
    // prefix/lang and finally in just prefix.
    //
    // Note that we use LC_MESSAGES on all platforms and not just Unix, because
    // it doesn't cost much to look into one more directory and doing it this
    // way has two important benefits:
    // a) we don't break compatibility with wx-2.6 and older by stopping to
    //    look in a directory where the catalogs used to be and thus silently
    //    breaking apps after they are recompiled against the latest wx
    // b) it makes it possible to package app's support files in the same
    //    way on all target platforms
    const wxString pathPrefix = wxFileName(prefix, lang).GetFullPath();

    wxString searchPath;
    searchPath.reserve(4*pathPrefix.length());
    searchPath << pathPrefix << wxFILE_SEP_PATH << "LC_MESSAGES" << wxPATH_SEP
            << prefix << wxFILE_SEP_PATH << wxPATH_SEP
            << pathPrefix;

    return searchPath;
}

// construct the search path for the given language
static wxString GetFullSearchPath(const wxString& lang)
{
    // first take the entries explicitly added by the program
    wxArrayString paths;
    paths.reserve(gs_searchPrefixes.size() + 1);
    size_t n,
        count = gs_searchPrefixes.size();
    for ( n = 0; n < count; n++ )
    {
        paths.Add(GetMsgCatalogSubdirs(gs_searchPrefixes[n], lang));
    }


#if wxUSE_STDPATHS
    // then look in the standard location
    const wxString stdp = wxStandardPaths::Get().
        GetLocalizedResourcesDir(lang, wxStandardPaths::ResourceCat_Messages);

    if ( paths.Index(stdp) == wxNOT_FOUND )
        paths.Add(stdp);
#endif // wxUSE_STDPATHS

    // last look in default locations
#ifdef __UNIX__
    // LC_PATH is a standard env var containing the search path for the .mo
    // files
    const char *pszLcPath = wxGetenv("LC_PATH");
    if ( pszLcPath )
    {
        const wxString lcp = GetMsgCatalogSubdirs(pszLcPath, lang);
        if ( paths.Index(lcp) == wxNOT_FOUND )
            paths.Add(lcp);
    }

    // also add the one from where wxWin was installed:
    wxString wxp = wxGetInstallPrefix();
    if ( !wxp.empty() )
    {
        wxp = GetMsgCatalogSubdirs(wxp + wxS("/share/locale"), lang);
        if ( paths.Index(wxp) == wxNOT_FOUND )
            paths.Add(wxp);
    }
#endif // __UNIX__


    // finally construct the full search path
    wxString searchPath;
    searchPath.reserve(500);
    count = paths.size();
    for ( n = 0; n < count; n++ )
    {
        searchPath += paths[n];
        if ( n != count - 1 )
            searchPath += wxPATH_SEP;
    }

    return searchPath;
}

} // anonymous namespace


void wxFileTranslationsLoader::AddCatalogLookupPathPrefix(const wxString& prefix)
{
    if ( gs_searchPrefixes.Index(prefix) == wxNOT_FOUND )
    {
        gs_searchPrefixes.Add(prefix);
    }
    //else: already have it
}


bool wxFileTranslationsLoader::LoadCatalog(wxTranslations *translations,
                                           const wxString& domain,
                                           const wxString& lang)
{
    wxCHECK_MSG( lang.length() >= LEN_LANG, false,
                 "invalid language specification" );

    wxString searchPath;

#if wxUSE_FONTMAP
    // first look for the catalog for this language and the current locale:
    // notice that we don't use the system name for the locale as this would
    // force us to install catalogs in different locations depending on the
    // system but always use the canonical name
    wxFontEncoding encSys = wxLocale::GetSystemEncoding();
    if ( encSys != wxFONTENCODING_SYSTEM )
    {
        wxString fullname(lang);
        fullname << wxS('.') << wxFontMapperBase::GetEncodingName(encSys);
        searchPath << GetFullSearchPath(fullname) << wxPATH_SEP;
    }
#endif // wxUSE_FONTMAP

    searchPath += GetFullSearchPath(lang);
    if ( lang.length() > LEN_LANG && lang[LEN_LANG] == wxS('_') )
    {
        // also add just base locale name: for things like "fr_BE" (Belgium
        // French) we should use fall back on plain "fr" if no Belgium-specific
        // message catalogs exist
        searchPath << wxPATH_SEP
                    << GetFullSearchPath(ExtractLang(lang));
    }

    wxLogTrace(TRACE_I18N, wxS("Looking for \"%s.mo\" in search path \"%s\""),
                domain, searchPath);

    wxFileName fn(domain);
    fn.SetExt(wxS("mo"));

    wxString strFullName;
    if ( !wxFindFileInPath(&strFullName, searchPath, fn.GetFullPath()) )
    {
        wxLogVerbose(_("catalog file for domain '%s' not found."), domain);
        wxLogTrace(TRACE_I18N, wxS("Catalog \"%s.mo\" not found"), domain);
        return false;
    }

    // open file and read its data
    wxLogVerbose(_("using catalog '%s' from '%s'."), domain, strFullName.c_str());
    wxLogTrace(TRACE_I18N, wxS("Using catalog \"%s\"."), strFullName.c_str());

    return translations->LoadCatalogFile(strFullName, domain);
}

// ----------------------------------------------------------------------------
// wxTranslationsModule module (for destruction of gs_translations)
// ----------------------------------------------------------------------------

class wxTranslationsModule: public wxModule
{
    DECLARE_DYNAMIC_CLASS(wxTranslationsModule)
    public:
        wxTranslationsModule() {}

        bool OnInit()
        {
            return true;
        }

        void OnExit()
        {
            if ( gs_translationsOwned )
                delete gs_translations;
            gs_translations = NULL;
            gs_translationsOwned = true;
        }
};

IMPLEMENT_DYNAMIC_CLASS(wxTranslationsModule, wxModule)

#endif // wxUSE_INTL