tqt3/src/kernel/qscriptengine.cpp

/****************************************************************************
**
** Copyright (C) 2003-2008 Trolltech ASA.  All rights reserved.
**
** This file is part of the kernel module of the TQt GUI Toolkit.
**
** This file may be used under the terms of the GNU General
** Public License versions 2.0 or 3.0 as published by the Free
** Software Foundation and appearing in the files LICENSE.GPL2
** and LICENSE.GPL3 included in the packaging of this file.
** Alternatively you may (at your option) use any later version
** of the GNU General Public License if such license has been
** publicly approved by Trolltech ASA (or its successors, if any)
** and the KDE Free TQt Foundation.
**
** Please review the following information to ensure GNU General
** Public Licensing requirements will be met:
** http://trolltech.com/products/qt/licenses/licensing/opensource/.
** If you are unsure which license is appropriate for your use, please
** review the following information:
** http://trolltech.com/products/qt/licenses/licensing/licensingoverview
** or contact the sales department at sales@trolltech.com.
**
** This file may be used under the terms of the Q Public License as
** defined by Trolltech ASA and appearing in the file LICENSE.TQPL
** included in the packaging of this file.  Licensees holding valid TQt
** Commercial licenses may use this file in accordance with the TQt
** Commercial License Agreement provided with the Software.
**
** This file is provided "AS IS" with NO WARRANTY OF ANY KIND,
** INCLUDING THE WARRANTIES OF DESIGN, MERCHANTABILITY AND FITNESS FOR
** A PARTICULAR PURPOSE. Trolltech reserves all rights not granted
** herein.
**
**********************************************************************/

#include "qscriptengine_p.h"

#include "ntqstring.h"
#include "ntqrect.h"
#include "ntqfont.h"
#include <private/qunicodetables_p.h>
#include "qtextengine_p.h"
#include "qfontengine_p.h"
#include <stdlib.h>

#undef None
#undef Pre
#undef Above
#undef Below

const int Prealloc = 256;
template<class T>
class TQVarLengthArray
{
public:
    inline explicit TQVarLengthArray(int size = 0);
    inline ~TQVarLengthArray() {
        if (ptr != reinterpret_cast<T *>(array))
            free(ptr);
    }

    inline int size() const { return s; }
    inline int count() const { return s; }
    inline bool isEmpty() const { return (s == 0); }
    inline void resize(int size);
    inline void clear() { resize(0); }

    inline int capacity() const { return a; }
    inline void reserve(int size);

    inline T &operator[](int idx) {
        Q_ASSERT(idx >= 0 && idx < s);
        return ptr[idx];
    }
    inline const T &operator[](int idx) const {
        Q_ASSERT(idx >= 0 && idx < s);
        return ptr[idx];
    }

    inline void append(const T &t) {
        const int idx = s;
        resize(idx + 1);
        ptr[idx] = t;
    }

    inline T *data() { return ptr; }
    inline const T *data() const { return ptr; }
    inline const T * constData() const { return ptr; }

private:
    void realloc(int size, int alloc);

    int a;
    int s;
    T *ptr;
    TQ_UINT64 array[((Prealloc * sizeof(T)) / sizeof(TQ_UINT64)) + 1];
};

template <class T>
Q_INLINE_TEMPLATES TQVarLengthArray<T>::TQVarLengthArray(int asize)
    : s(asize) {
    if (s > Prealloc) {
        ptr = reinterpret_cast<T *>(malloc(s * sizeof(T)));
        a = s;
    } else {
        ptr = reinterpret_cast<T *>(array);
        a = Prealloc;
    }
}


// --------------------------------------------------------------------------------------------------------------------------------------------
//
// Basic processing
//
// --------------------------------------------------------------------------------------------------------------------------------------------

static inline void positionCluster(TQShaperItem *item, int gfrom,  int glast)
{
    int nmarks = glast - gfrom;
    if (nmarks <= 0) {
        tqWarning("positionCluster: no marks to position!");
        return;
    }

    TQFontEngine *f = item->font;

    glyph_metrics_t baseInfo = f->boundingBox(item->glyphs[gfrom]);

    if (item->script == TQFont::Hebrew)
        // we need to attach below the baseline, because of the hebrew iud.
        baseInfo.height = TQMAX(baseInfo.height, -baseInfo.y);

    TQRect baseRect(baseInfo.x, baseInfo.y, baseInfo.width, baseInfo.height);

//      tqDebug("---> positionCluster: cluster from %d to %d", gfrom, glast);
//     tqDebug("baseInfo: %f/%f (%f/%f) off=%f/%f", baseInfo.x, baseInfo.y, baseInfo.width, baseInfo.height, baseInfo.xoff, baseInfo.yoff);

    int size = (f->ascent()/10);
    int offsetBase = (size - 4) / 4 + TQMIN(size, 4) + 1;
//     tqDebug("offset = %f", offsetBase);

    bool rightToLeft = item->flags & TQTextEngine::RightToLeft;

    int i;
    unsigned char lastCmb = 0;
    TQRect attachmentRect;

    for(i = 1; i <= nmarks; i++) {
        glyph_t mark = item->glyphs[gfrom+i];
        TQPoint p;
        glyph_metrics_t markInfo = f->boundingBox(mark);
        TQRect markRect(markInfo.x, markInfo.y, markInfo.width, markInfo.height);
//          tqDebug("markInfo: %f/%f (%f/%f) off=%f/%f", markInfo.x, markInfo.y, markInfo.width, markInfo.height, markInfo.xoff, markInfo.yoff);

        int offset = offsetBase;
        unsigned char cmb = item->attributes[gfrom+i].combiningClass;

        // ### maybe the whole position determination should move down to heuristicSetGlyphAttributes. Would save some
        // bits  in the glyphAttributes structure.
        if (cmb < 200) {
            // fixed position classes. We approximate by mapping to one of the others.
            // currently I added only the ones for arabic, hebrew, lao and thai.

            // for Lao and Thai marks with class 0, see below (heuristicSetGlyphAttributes)

            // add a bit more offset to arabic, a bit hacky
            if (cmb >= 27 && cmb <= 36 && offset < 3)
                offset +=1;
            // below
            if ((cmb >= 10 && cmb <= 18) ||
                 cmb == 20 || cmb == 22 ||
                 cmb == 29 || cmb == 32)
                cmb = TQChar::Combining_Below;
            // above
            else if (cmb == 23 || cmb == 27 || cmb == 28 ||
                      cmb == 30 || cmb == 31 || (cmb >= 33 && cmb <= 36))
                cmb = TQChar::Combining_Above;
            //below-right
            else if (cmb == 9 || cmb == 103 || cmb == 118)
                cmb = TQChar::Combining_BelowRight;
            // above-right
            else if (cmb == 24 || cmb == 107 || cmb == 122)
                cmb = TQChar::Combining_AboveRight;
            else if (cmb == 25)
                cmb = TQChar::Combining_AboveLeft;
            // fixed:
            //  19 21

        }

        // combining marks of different class don't interact. Reset the rectangle.
        if (cmb != lastCmb) {
            //tqDebug("resetting rect");
            attachmentRect = baseRect;
        }

        switch(cmb) {
        case TQChar::Combining_DoubleBelow:
                // ### wrong in rtl context!
        case TQChar::Combining_BelowLeft:
            p += TQPoint(0, offset);
        case TQChar::Combining_BelowLeftAttached:
            p += attachmentRect.bottomLeft() - markRect.topLeft();
            break;
        case TQChar::Combining_Below:
            p += TQPoint(0, offset);
        case TQChar::Combining_BelowAttached:
            p += attachmentRect.bottomLeft() - markRect.topLeft();
            p += TQPoint((attachmentRect.width() - markRect.width())/2 , 0);
            break;
            case TQChar::Combining_BelowRight:
            p += TQPoint(0, offset);
        case TQChar::Combining_BelowRightAttached:
            p += attachmentRect.bottomRight() - markRect.topRight();
            break;
            case TQChar::Combining_Left:
            p += TQPoint(-offset, 0);
        case TQChar::Combining_LeftAttached:
            break;
            case TQChar::Combining_Right:
            p += TQPoint(offset, 0);
        case TQChar::Combining_RightAttached:
            break;
        case TQChar::Combining_DoubleAbove:
            // ### wrong in RTL context!
        case TQChar::Combining_AboveLeft:
            p += TQPoint(0, -offset);
        case TQChar::Combining_AboveLeftAttached:
            p += attachmentRect.topLeft() - markRect.bottomLeft();
            break;
        case TQChar::Combining_Above:
            p += TQPoint(0, -offset);
        case TQChar::Combining_AboveAttached:
            p += attachmentRect.topLeft() - markRect.bottomLeft();
            p += TQPoint((attachmentRect.width() - markRect.width())/2 , 0);
            break;
        case TQChar::Combining_AboveRight:
            p += TQPoint(0, -offset);
        case TQChar::Combining_AboveRightAttached:
            p += attachmentRect.topRight() - markRect.bottomRight();
            break;

        case TQChar::Combining_IotaSubscript:
            default:
                break;
        }
//         tqDebug("char=%x combiningClass = %d offset=%d/%d", mark, cmb, p.x(), p.y());
        markRect.moveBy(p.x(), p.y());
        attachmentRect |= markRect;
        lastCmb = cmb;
        if (rightToLeft) {
            item->offsets[gfrom+i].x = p.x();
            item->offsets[gfrom+i].y = p.y();
        } else {
            item->offsets[gfrom+i].x = p.x() - baseInfo.xoff;
            item->offsets[gfrom+i].y = p.y() - baseInfo.yoff;
        }
        item->advances[gfrom+i] = 0;
    }
    item->has_positioning = TRUE;
}


void qt_heuristicPosition(TQShaperItem *item)
{
    int cEnd = -1;
    int i = item->num_glyphs;
    while (i--) {
        if (cEnd == -1 && item->attributes[i].mark) {
            cEnd = i;
        } else if (cEnd != -1 && !item->attributes[i].mark) {
            positionCluster(item, i, cEnd);
            cEnd = -1;
        }
    }
}



// set the glyph attributes heuristically. Assumes a 1 to 1 relationship between chars and glyphs
// and no reordering.
// also computes logClusters heuristically
static void heuristicSetGlyphAttributes(TQShaperItem *item, const TQChar *uc, int length)
{
    // justification is missing here!!!!!

    if ( item->num_glyphs != length )
       tqWarning("TQScriptEngine::heuristicSetGlyphAttributes: char length and num glyphs disagree" );

    unsigned short *logClusters = item->log_clusters;

    int i;
    for (i = 0; i < length; ++i)
        logClusters[i] = i;

    // first char in a run is never (treated as) a mark
    int cStart = 0;
    item->attributes[0].mark = FALSE;
    item->attributes[0].clusterStart = TRUE;
    item->attributes[0].combiningClass = 0;
    if (qIsZeroWidthChar(uc[0].unicode())) {
        item->attributes[0].zeroWidth = TRUE;
        item->advances[0] = 0;
        item->has_positioning = TRUE;
    } else {
        item->attributes[0].zeroWidth = FALSE;
    }

    int lastCat = ::category(uc[0]);
    for (i = 1; i < length; ++i) {
        int cat = ::category(uc[i]);
        if (qIsZeroWidthChar(uc[i].unicode())) {
	    item->attributes[i].mark = FALSE;
	    item->attributes[i].clusterStart = TRUE;
	    item->attributes[i].zeroWidth = TRUE;
	    item->attributes[i].combiningClass = 0;
	    cStart = i;
            item->advances[i] = 0;
            item->has_positioning = TRUE;
        } else if (cat != TQChar::Mark_NonSpacing) {
            item->attributes[i].mark = FALSE;
            item->attributes[i].clusterStart = TRUE;
            item->attributes[i].combiningClass = 0;
            cStart = i;
        } else {
            int cmb = ::combiningClass(uc[i]);

            if (cmb == 0) {
                // Fix 0 combining classes
                if ((uc[i].unicode() & 0xff00) == 0x0e00) {
                    // thai or lao
                    unsigned char col = uc[i].cell();
                    if (col == 0x31 ||
                         col == 0x34 ||
                         col == 0x35 ||
                         col == 0x36 ||
                         col == 0x37 ||
                         col == 0x47 ||
                         col == 0x4c ||
                         col == 0x4d ||
                         col == 0x4e) {
                        cmb = TQChar::Combining_AboveRight;
                    } else if (col == 0xb1 ||
                                col == 0xb4 ||
                                col == 0xb5 ||
                                col == 0xb6 ||
                                col == 0xb7 ||
                                col == 0xbb ||
                                col == 0xcc ||
                                col == 0xcd) {
                        cmb = TQChar::Combining_Above;
                    } else if (col == 0xbc) {
                        cmb = TQChar::Combining_Below;
                    }
                }
            }

            item->attributes[i].mark = TRUE;
            item->attributes[i].clusterStart = FALSE;
            item->attributes[i].combiningClass = cmb;
            logClusters[i] = cStart;
            item->advances[i] = 0;
            item->has_positioning = TRUE;
        }

        if (lastCat == TQChar::Separator_Space)
            item->attributes[i-1].justification = GlyphAttributes::Space;
        else if (cat != TQChar::Mark_NonSpacing)
            item->attributes[i-1].justification = GlyphAttributes::Character;
        else
            item->attributes[i-1].justification = GlyphAttributes::NoJustification;

        lastCat = cat;
    }
}

static void heuristicSetGlyphAttributes(TQShaperItem *item)
{
    heuristicSetGlyphAttributes(item, item->string->unicode() + item->from, item->length);
}


static bool basic_shape(TQShaperItem *item)
{
    if (item->font->stringToCMap(item->string->unicode()+item->from, item->length, item->glyphs, item->advances,
                                 &item->num_glyphs, item->flags & TQTextEngine::RightToLeft) != TQFontEngine::NoError)
        return FALSE;

    heuristicSetGlyphAttributes(item);
    qt_heuristicPosition(item);
    return TRUE;
}

// --------------------------------------------------------------------------------------------------------------------------------------------
//
// Middle eastern languages
//
// --------------------------------------------------------------------------------------------------------------------------------------------

// Uniscribe also defines dlig for Hebrew, but we leave this out for now, as it's mostly
// ligatures one does not want in modern Hebrew (as lam-alef ligatures).
enum {
    CcmpProperty = 0x1
};
#if defined(Q_WS_X11) && !defined(TQT_NO_XFTFREETYPE)
static const TQOpenType::Features hebrew_features[] = {
    { FT_MAKE_TAG('c', 'c', 'm', 'p'), CcmpProperty },
    {0, 0}
};
#endif
/* Hebrew shaping. In the non opentype case we try to use the
   presentation forms specified for Hebrew. Especially for the
   ligatures with Dagesh this gives much better results than we could
   achieve manually.
*/
static bool hebrew_shape(TQShaperItem *item)
{
    Q_ASSERT(item->script == TQFont::Hebrew);

#if defined(Q_WS_X11) && !defined(TQT_NO_XFTFREETYPE)
    TQOpenType *openType = item->font->openType();

    if (openType && openType->supportsScript(item->script)) {
        openType->selectScript(item->script, hebrew_features);

        if (item->font->stringToCMap(item->string->unicode()+item->from, item->length, item->glyphs, item->advances,
                                     &item->num_glyphs, item->flags & TQTextEngine::RightToLeft) != TQFontEngine::NoError)
            return FALSE;

        heuristicSetGlyphAttributes(item);
        openType->shape(item);
        return openType->positionAndAdd(item);
    }
#endif

    enum {
        Dagesh = 0x5bc,
        ShinDot = 0x5c1,
        SinDot = 0x5c2,
        Patah = 0x5b7,
        Qamats = 0x5b8,
        Holam = 0x5b9,
        Rafe = 0x5bf
    };
    unsigned short chars[512];
    TQChar *shapedChars = item->length > 256 ? (TQChar *)::malloc(2*item->length * sizeof(TQChar)) : (TQChar *)chars;

    const TQChar *uc = item->string->unicode() + item->from;
    unsigned short *logClusters = item->log_clusters;

    *shapedChars = *uc;
    logClusters[0] = 0;
    int slen = 1;
    int cluster_start = 0;
    int i;
    for (i = 1; i < item->length; ++i) {
        ushort base = shapedChars[slen-1].unicode();
        ushort shaped = 0;
        bool invalid = FALSE;
        if (uc[i].unicode() == Dagesh) {
            if (base >= 0x5d0
                && base <= 0x5ea
                && base != 0x5d7
                && base != 0x5dd
                && base != 0x5df
                && base != 0x5e2
                && base != 0x5e5) {
                shaped = base - 0x5d0 + 0xfb30;
            } else if (base == 0xfb2a || base == 0xfb2b /* Shin with Shin or Sin dot */) {
                shaped = base + 2;
            } else {
                invalid = TRUE;
            }
        } else if (uc[i].unicode() == ShinDot) {
            if (base == 0x05e9)
                shaped = 0xfb2a;
            else if (base == 0xfb49)
                shaped = 0xfb2c;
            else
                invalid = TRUE;
        } else if (uc[i].unicode() == SinDot) {
            if (base == 0x05e9)
                shaped = 0xfb2b;
            else if (base == 0xfb49)
                shaped = 0xfb2d;
            else
                invalid = TRUE;
        } else if (uc[i].unicode() == Patah) {
            if (base == 0x5d0)
                shaped = 0xfb2e;
        } else if (uc[i].unicode() == Qamats) {
            if (base == 0x5d0)
                shaped = 0xfb2f;
        } else if (uc[i].unicode() == Holam) {
            if (base == 0x5d5)
                shaped = 0xfb4b;
        } else if (uc[i].unicode() == Rafe) {
            if (base == 0x5d1)
                shaped = 0xfb4c;
            else if (base == 0x5db)
                shaped = 0xfb4d;
            else if (base == 0x5e4)
                shaped = 0xfb4e;
        }

        if (invalid) {
            shapedChars[slen] = 0x25cc;
            item->attributes[slen].clusterStart = TRUE;
            item->attributes[slen].mark = FALSE;
            item->attributes[slen].combiningClass = 0;
            cluster_start = slen;
            ++slen;
        }
        if (shaped) {
            if (item->font->canRender((TQChar *)&shaped, 1)) {
                shapedChars[slen-1] = TQChar(shaped);
            } else
                shaped = 0;
        }
        if (!shaped) {
            shapedChars[slen] = uc[i];
            if (::category(uc[i]) != TQChar::Mark_NonSpacing) {
                item->attributes[slen].clusterStart = TRUE;
                item->attributes[slen].mark = FALSE;
                item->attributes[slen].combiningClass = 0;
                cluster_start = slen;
            } else {
                item->attributes[slen].clusterStart = FALSE;
                item->attributes[slen].mark = TRUE;
                item->attributes[slen].combiningClass = ::combiningClass(uc[i]);
            }
            ++slen;
        }
        logClusters[i] = cluster_start;
    }

    if (item->font->stringToCMap(shapedChars, slen, item->glyphs, item->advances,
                                 &item->num_glyphs, item->flags & TQTextEngine::RightToLeft) != TQFontEngine::NoError)
        return FALSE;
    for (i = 0; i < item->num_glyphs; ++i) {
        if (item->attributes[i].mark)
            item->advances[i] = 0;
    }
    qt_heuristicPosition(item);

    if (item->length > 256)
        ::free(shapedChars);
    return TRUE;
}

// these groups correspond to the groups defined in the Unicode standard.
// Some of these groups are equal whith regards to both joining and line breaking behaviour,
// and thus have the same enum value
//
// I'm not sure the mapping of syriac to arabic enums is correct with regards to justification, but as
// I couldn't find any better document I'll hope for the best.
enum ArabicGroup {
    // NonJoining
    ArabicNone,
    ArabicSpace,
    // Transparent
    Transparent,
    // Causing
    Center,
    Kashida,

    // Arabic
    // Dual
    Beh,
    Noon,
    Meem = Noon,
    Heh = Noon,
    KnottedHeh = Noon,
    HehGoal = Noon,
    SwashKaf = Noon,
    Yeh,
    Hah,
    Seen,
    Sad = Seen,
    Tah,
    Kaf = Tah,
    Gaf = Tah,
    Lam = Tah,
    Ain,
    Feh = Ain,
    Qaf = Ain,
    // Right
    Alef,
    Waw,
    Dal,
    TehMarbuta = Dal,
    Reh,
    HamzaOnHehGoal,
    YehWithTail = HamzaOnHehGoal,
    YehBarre = HamzaOnHehGoal,

    // Syriac
    // Dual
    Beth = Beh,
    Gamal = Ain,
    Heth = Noon,
    Teth = Hah,
    Yudh = Noon,
    Kaph = Noon,
    Lamadh = Lam,
    Mim = Noon,
    Nun = Noon,
    Semakh = Noon,
    FinalSemakh = Noon,
    SyriacE = Ain,
    Pe = Ain,
    ReversedPe = Hah,
    Qaph = Noon,
    Shin = Noon,
    Fe = Ain,

    // Right
    Alaph = Alef,
    Dalath = Dal,
    He = Dal,
    SyriacWaw = Waw,
    Zain = Alef,
    YudhHe = Waw,
    Sadhe = HamzaOnHehGoal,
    Taw = Dal,

    // Compiler bug? Otherwise ArabicGroupsEnd would be equal to Dal + 1.
    Dummy = HamzaOnHehGoal,
    ArabicGroupsEnd
};

static const unsigned char arabic_group[0x150] = {
    ArabicNone, ArabicNone, ArabicNone, ArabicNone,
    ArabicNone, ArabicNone, ArabicNone, ArabicNone,
    ArabicNone, ArabicNone, ArabicNone, ArabicNone,
    ArabicNone, ArabicNone, ArabicNone, ArabicNone,

    Transparent, Transparent, Transparent, Transparent,
    Transparent, Transparent, ArabicNone, ArabicNone,
    ArabicNone, ArabicNone, ArabicNone, ArabicNone,
    ArabicNone, ArabicNone, ArabicNone, ArabicNone,

    ArabicNone, ArabicNone, Alef, Alef,
    Waw, Alef, Yeh, Alef,
    Beh, TehMarbuta, Beh, Beh,
    Hah, Hah, Hah, Dal,

    Dal, Reh, Reh, Seen,
    Seen, Sad, Sad, Tah,
    Tah, Ain, Ain, ArabicNone,
    ArabicNone, ArabicNone, ArabicNone, ArabicNone,

    // 0x640
    Kashida, Feh, Qaf, Kaf,
    Lam, Meem, Noon, Heh,
    Waw, Yeh, Yeh, Transparent,
    Transparent, Transparent, Transparent, Transparent,

    Transparent, Transparent, Transparent, Transparent,
    Transparent, Transparent, Transparent, Transparent,
    Transparent, ArabicNone, ArabicNone, ArabicNone,
    ArabicNone, ArabicNone, ArabicNone, ArabicNone,

    ArabicNone, ArabicNone, ArabicNone, ArabicNone,
    ArabicNone, ArabicNone, ArabicNone, ArabicNone,
    ArabicNone, ArabicNone, ArabicNone, ArabicNone,
    ArabicNone, ArabicNone, Beh, Qaf,

    Transparent, Alef, Alef, Alef,
    ArabicNone, Alef, Waw, Waw,
    Yeh, Beh, Beh, Beh,
    Beh, Beh, Beh, Beh,

    // 0x680
    Beh, Hah, Hah, Hah,
    Hah, Hah, Hah, Hah,
    Dal, Dal, Dal, Dal,
    Dal, Dal, Dal, Dal,

    Dal, Reh, Reh, Reh,
    Reh, Reh, Reh, Reh,
    Reh, Reh, Seen, Seen,
    Seen, Sad, Sad, Tah,

    Ain, Feh, Feh, Feh,
    Feh, Feh, Feh, Qaf,
    Qaf, Gaf, SwashKaf, Gaf,
    Kaf, Kaf, Kaf, Gaf,

    Gaf, Gaf, Gaf, Gaf,
    Gaf, Lam, Lam, Lam,
    Lam, Noon, Noon, Noon,
    Noon, Noon, KnottedHeh, Hah,

    // 0x6c0
    TehMarbuta, HehGoal, HamzaOnHehGoal, HamzaOnHehGoal,
    Waw, Waw, Waw, Waw,
    Waw, Waw, Waw, Waw,
    Yeh, YehWithTail, Yeh, Waw,

    Yeh, Yeh, YehBarre, YehBarre,
    ArabicNone, TehMarbuta, Transparent, Transparent,
    Transparent, Transparent, Transparent, Transparent,
    Transparent, ArabicNone, ArabicNone, Transparent,

    Transparent, Transparent, Transparent, Transparent,
    Transparent, ArabicNone, ArabicNone, Transparent,
    Transparent, ArabicNone, Transparent, Transparent,
    Transparent, Transparent, Dal, Reh,

    ArabicNone, ArabicNone, ArabicNone, ArabicNone,
    ArabicNone, ArabicNone, ArabicNone, ArabicNone,
    ArabicNone, ArabicNone, Seen, Sad,
    Ain, ArabicNone, ArabicNone, KnottedHeh,

    // 0x700
    ArabicNone, ArabicNone, ArabicNone, ArabicNone,
    ArabicNone, ArabicNone, ArabicNone, ArabicNone,
    ArabicNone, ArabicNone, ArabicNone, ArabicNone,
    ArabicNone, ArabicNone, ArabicNone, ArabicNone,

    Alaph, Transparent, Beth, Gamal,
    Gamal, Dalath, Dalath, He,
    SyriacWaw, Zain, Heth, Teth,
    Teth, Yudh, YudhHe, Kaph,

    Lamadh, Mim, Nun, Semakh,
    FinalSemakh, SyriacE, Pe, ReversedPe,
    Sadhe, Qaph, Dalath, Shin,
    Taw, Beth, Gamal, Dalath,

    Transparent, Transparent, Transparent, Transparent,
    Transparent, Transparent, Transparent, Transparent,
    Transparent, Transparent, Transparent, Transparent,
    Transparent, Transparent, Transparent, Transparent,

    Transparent, Transparent, Transparent, Transparent,
    Transparent, Transparent, Transparent, Transparent,
    Transparent, Transparent, Transparent, ArabicNone,
    ArabicNone, Zain, Kaph, Fe,
};

static inline ArabicGroup arabicGroup(unsigned short uc)
{
    if (uc >= 0x0600 && uc < 0x750)
        return (ArabicGroup) arabic_group[uc-0x600];
    else if (uc == 0x200d)
        return Center;
    else if (::category(uc) == TQChar::Separator_Space)
        return ArabicSpace;
    else
        return ArabicNone;
}


/*
   Arabic shaping obeys a number of rules according to the joining classes (see Unicode book, section on
   arabic).

   Each unicode char has a joining class (right, dual (left&right), center (joincausing) or transparent).
   transparent joining is not encoded in TQChar::joining(), but applies to all combining marks and format marks.

   Right join-causing: dual + center
   Left join-causing: dual + right + center

   Rules are as follows (for a string already in visual order, as we have it here):

   R1 Transparent characters do not affect joining behaviour.
   R2 A right joining character, that has a right join-causing char on the right will get form XRight
   (R3 A left joining character, that has a left join-causing char on the left will get form XLeft)
   Note: the above rule is meaningless, as there are no pure left joining characters defined in Unicode
   R4 A dual joining character, that has a left join-causing char on the left and a right join-causing char on
             the right will get form XMedial
   R5  A dual joining character, that has a right join causing char on the right, and no left join causing char on the left
         will get form XRight
   R6 A dual joining character, that has a  left join causing char on the left, and no right join causing char on the right
         will get form XLeft
   R7 Otherwise the character will get form XIsolated

   Additionally we have to do the minimal ligature support for lam-alef ligatures:

   L1 Transparent characters do not affect ligature behaviour.
   L2 Any sequence of Alef(XRight) + Lam(XMedial) will form the ligature Alef.Lam(XLeft)
   L3 Any sequence of Alef(XRight) + Lam(XLeft) will form the ligature Alef.Lam(XIsolated)

   The state table below handles rules R1-R7.
*/

enum Shape {
    XIsolated,
    XFinal,
    XInitial,
    XMedial,
    // intermediate state
    XCausing
};


enum Joining {
    JNone,
    JCausing,
    JDual,
    JRight,
    JTransparent
};


static const Joining joining_for_group[ArabicGroupsEnd] = {
    // NonJoining
    JNone, // ArabicNone
    JNone, // ArabicSpace
    // Transparent
    JTransparent, // Transparent
    // Causing
    JCausing, // Center
    JCausing, // Kashida
    // Dual
    JDual, // Beh
    JDual, // Noon
    JDual, // Yeh
    JDual, // Hah
    JDual, // Seen
    JDual, // Tah
    JDual, // Ain
    // Right
    JRight, // Alef
    JRight, // Waw
    JRight, // Dal
    JRight, // Reh
    JRight  // HamzaOnHehGoal
};


struct JoiningPair {
    Shape form1;
    Shape form2;
};

static const JoiningPair joining_table[5][4] =
// None, Causing, Dual, Right
{
    { { XIsolated, XIsolated }, { XIsolated, XCausing }, { XIsolated, XInitial }, { XIsolated, XIsolated } }, // XIsolated
    { { XFinal, XIsolated }, { XFinal, XCausing }, { XFinal, XInitial }, { XFinal, XIsolated } }, // XFinal
    { { XIsolated, XIsolated }, { XInitial, XCausing }, { XInitial, XMedial }, { XInitial, XFinal } }, // XInitial
    { { XFinal, XIsolated }, { XMedial, XCausing }, { XMedial, XMedial }, { XMedial, XFinal } }, // XMedial
    { { XIsolated, XIsolated }, { XIsolated, XCausing }, { XIsolated, XMedial }, { XIsolated, XFinal } }, // XCausing
};


/*
According to http://www.microsoft.com/middleeast/Arabicdev/IE6/KBase.asp

1. Find the priority of the connecting opportunities in each word
2. Add expansion at the highest priority connection opportunity
3. If more than one connection opportunity have the same highest value,
   use the opportunity closest to the end of the word.

Following is a chart that provides the priority for connection
opportunities and where expansion occurs. The character group names
are those in table 6.6 of the UNICODE 2.0 book.


PrioritY        Glyph                   Condition                                       Kashida Location

Arabic_Kashida        User inserted Kashida   The user entered a Kashida in a position.       After the user
                (Shift+j or Shift+[E with hat])    Thus, it is the highest priority to insert an   inserted kashida
                                        automatic kashida.

Arabic_Seen        Seen, Sad               Connecting to the next character.               After the character.
                                        (Initial or medial form).

Arabic_HaaDal        Teh Marbutah, Haa, Dal  Connecting to previous character.               Before the final form
                                                                                        of these characters.

Arabic_Alef     Alef, Tah, Lam,         Connecting to previous character.               Before the final form
                Kaf and Gaf                                                             of these characters.

Arabic_BaRa     Reh, Yeh                Connected to medial Beh                         Before preceding medial Baa

Arabic_Waw        Waw, Ain, Qaf, Feh      Connecting to previous character.               Before the final form of
                                                                                        these characters.

Arabic_Normal   Other connecting        Connecting to previous character.               Before the final form
                characters                                                              of these characters.



This seems to imply that we have at most one kashida point per arabic word.

*/

struct TQArabicProperties {
    unsigned char shape;
    unsigned char justification;
};


static void getArabicProperties(const unsigned short *chars, int len, TQArabicProperties *properties)
{
//     tqDebug("arabicSyriacOpenTypeShape: properties:");
    int lastPos = 0;
    int lastGroup = ArabicNone;

    ArabicGroup group = arabicGroup(chars[0]);
    Joining j = joining_for_group[group];
    Shape shape = joining_table[XIsolated][j].form2;
    properties[0].justification = GlyphAttributes::NoJustification;

    for (int i = 1; i < len; ++i) {
        // #### fix handling for spaces and punktuation
        properties[i].justification = GlyphAttributes::NoJustification;

        group = arabicGroup(chars[i]);
        j = joining_for_group[group];

        if (j == JTransparent) {
            properties[i].shape = XIsolated;
            continue;
        }

        properties[lastPos].shape = joining_table[shape][j].form1;
        shape = joining_table[shape][j].form2;

        switch(lastGroup) {
        case Seen:
            if (properties[lastPos].shape == XInitial || properties[lastPos].shape == XMedial)
                properties[i-1].justification = GlyphAttributes::Arabic_Seen;
            break;
        case Hah:
            if (properties[lastPos].shape == XFinal)
                properties[lastPos-1].justification = GlyphAttributes::Arabic_HaaDal;
            break;
        case Alef:
            if (properties[lastPos].shape == XFinal)
                properties[lastPos-1].justification = GlyphAttributes::Arabic_Alef;
            break;
        case Ain:
            if (properties[lastPos].shape == XFinal)
                properties[lastPos-1].justification = GlyphAttributes::Arabic_Waw;
            break;
        case Noon:
            if (properties[lastPos].shape == XFinal)
                properties[lastPos-1].justification = GlyphAttributes::Arabic_Normal;
            break;
        case ArabicNone:
            break;

        default:
            Q_ASSERT(FALSE);
        }

        lastGroup = ArabicNone;

        switch(group) {
        case ArabicNone:
        case Transparent:
        // ### Center should probably be treated as transparent when it comes to justification.
        case Center:
            break;
        case ArabicSpace:
            properties[i].justification = GlyphAttributes::Arabic_Space;
            break;
        case Kashida:
            properties[i].justification = GlyphAttributes::Arabic_Kashida;
            break;
        case Seen:
            lastGroup = Seen;
            break;

        case Hah:
        case Dal:
            lastGroup = Hah;
            break;

        case Alef:
        case Tah:
            lastGroup = Alef;
            break;

        case Yeh:
        case Reh:
            if (properties[lastPos].shape == XMedial && arabicGroup(chars[lastPos]) == Beh)
                properties[lastPos-1].justification = GlyphAttributes::Arabic_BaRa;
            break;

        case Ain:
        case Waw:
            lastGroup = Ain;
            break;

        case Noon:
        case Beh:
        case HamzaOnHehGoal:
            lastGroup = Noon;
            break;
        case ArabicGroupsEnd:
            Q_ASSERT(FALSE);
        }

        lastPos = i;
    }
    properties[lastPos].shape = joining_table[shape][JNone].form1;


//     for (int i = 0; i < len; ++i)
//         tqDebug("arabic properties(%d): uc=%x shape=%d, justification=%d", i, chars[i], properties[i].shape, properties[i].justification);
}






// The unicode to unicode shaping codec.
// does only presentation forms B at the moment, but that should be enough for
// simple display
static const ushort arabicUnicodeMapping[256][2] = {
    // base of shaped forms, and number-1 of them (0 for non shaping,
    // 1 for right binding and 3 for dual binding

    // These are just the glyphs available in Unicode,
    // some characters are in R class, but have no glyphs in Unicode.

    { 0x0600, 0 }, // 0x0600
    { 0x0601, 0 }, // 0x0601
    { 0x0602, 0 }, // 0x0602
    { 0x0603, 0 }, // 0x0603
    { 0x0604, 0 }, // 0x0604
    { 0x0605, 0 }, // 0x0605
    { 0x0606, 0 }, // 0x0606
    { 0x0607, 0 }, // 0x0607
    { 0x0608, 0 }, // 0x0608
    { 0x0609, 0 }, // 0x0609
    { 0x060A, 0 }, // 0x060A
    { 0x060B, 0 }, // 0x060B
    { 0x060C, 0 }, // 0x060C
    { 0x060D, 0 }, // 0x060D
    { 0x060E, 0 }, // 0x060E
    { 0x060F, 0 }, // 0x060F

    { 0x0610, 0 }, // 0x0610
    { 0x0611, 0 }, // 0x0611
    { 0x0612, 0 }, // 0x0612
    { 0x0613, 0 }, // 0x0613
    { 0x0614, 0 }, // 0x0614
    { 0x0615, 0 }, // 0x0615
    { 0x0616, 0 }, // 0x0616
    { 0x0617, 0 }, // 0x0617
    { 0x0618, 0 }, // 0x0618
    { 0x0619, 0 }, // 0x0619
    { 0x061A, 0 }, // 0x061A
    { 0x061B, 0 }, // 0x061B
    { 0x061C, 0 }, // 0x061C
    { 0x061D, 0 }, // 0x061D
    { 0x061E, 0 }, // 0x061E
    { 0x061F, 0 }, // 0x061F

    { 0x0620, 0 }, // 0x0620
    { 0xFE80, 0 }, // 0x0621            HAMZA
    { 0xFE81, 1 }, // 0x0622    R       ALEF WITH MADDA ABOVE
    { 0xFE83, 1 }, // 0x0623    R       ALEF WITH HAMZA ABOVE
    { 0xFE85, 1 }, // 0x0624    R       WAW WITH HAMZA ABOVE
    { 0xFE87, 1 }, // 0x0625    R       ALEF WITH HAMZA BELOW
    { 0xFE89, 3 }, // 0x0626    D       YEH WITH HAMZA ABOVE
    { 0xFE8D, 1 }, // 0x0627    R       ALEF
    { 0xFE8F, 3 }, // 0x0628    D       BEH
    { 0xFE93, 1 }, // 0x0629    R       TEH MARBUTA
    { 0xFE95, 3 }, // 0x062A    D       TEH
    { 0xFE99, 3 }, // 0x062B    D       THEH
    { 0xFE9D, 3 }, // 0x062C    D       JEEM
    { 0xFEA1, 3 }, // 0x062D    D       HAH
    { 0xFEA5, 3 }, // 0x062E    D       KHAH
    { 0xFEA9, 1 }, // 0x062F    R       DAL

    { 0xFEAB, 1 }, // 0x0630    R       THAL
    { 0xFEAD, 1 }, // 0x0631    R       REH
    { 0xFEAF, 1 }, // 0x0632    R       ZAIN
    { 0xFEB1, 3 }, // 0x0633    D       SEEN
    { 0xFEB5, 3 }, // 0x0634    D       SHEEN
    { 0xFEB9, 3 }, // 0x0635    D       SAD
    { 0xFEBD, 3 }, // 0x0636    D       DAD
    { 0xFEC1, 3 }, // 0x0637    D       TAH
    { 0xFEC5, 3 }, // 0x0638    D       ZAH
    { 0xFEC9, 3 }, // 0x0639    D       AIN
    { 0xFECD, 3 }, // 0x063A    D       GHAIN
    { 0x063B, 0 }, // 0x063B
    { 0x063C, 0 }, // 0x063C
    { 0x063D, 0 }, // 0x063D
    { 0x063E, 0 }, // 0x063E
    { 0x063F, 0 }, // 0x063F

    { 0x0640, 0 }, // 0x0640    C       TATWEEL // ### Join Causing, only one glyph
    { 0xFED1, 3 }, // 0x0641    D       FEH
    { 0xFED5, 3 }, // 0x0642    D       TQAF
    { 0xFED9, 3 }, // 0x0643    D       KAF
    { 0xFEDD, 3 }, // 0x0644    D       LAM
    { 0xFEE1, 3 }, // 0x0645    D       MEEM
    { 0xFEE5, 3 }, // 0x0646    D       NOON
    { 0xFEE9, 3 }, // 0x0647    D       HEH
    { 0xFEED, 1 }, // 0x0648    R       WAW
    { 0x0649, 3 }, // 0x0649            ALEF MAKSURA // ### Dual, glyphs not consecutive, handle in code.
    { 0xFEF1, 3 }, // 0x064A    D       YEH
    { 0x064B, 0 }, // 0x064B
    { 0x064C, 0 }, // 0x064C
    { 0x064D, 0 }, // 0x064D
    { 0x064E, 0 }, // 0x064E
    { 0x064F, 0 }, // 0x064F

    { 0x0650, 0 }, // 0x0650
    { 0x0651, 0 }, // 0x0651
    { 0x0652, 0 }, // 0x0652
    { 0x0653, 0 }, // 0x0653
    { 0x0654, 0 }, // 0x0654
    { 0x0655, 0 }, // 0x0655
    { 0x0656, 0 }, // 0x0656
    { 0x0657, 0 }, // 0x0657
    { 0x0658, 0 }, // 0x0658
    { 0x0659, 0 }, // 0x0659
    { 0x065A, 0 }, // 0x065A
    { 0x065B, 0 }, // 0x065B
    { 0x065C, 0 }, // 0x065C
    { 0x065D, 0 }, // 0x065D
    { 0x065E, 0 }, // 0x065E
    { 0x065F, 0 }, // 0x065F

    { 0x0660, 0 }, // 0x0660
    { 0x0661, 0 }, // 0x0661
    { 0x0662, 0 }, // 0x0662
    { 0x0663, 0 }, // 0x0663
    { 0x0664, 0 }, // 0x0664
    { 0x0665, 0 }, // 0x0665
    { 0x0666, 0 }, // 0x0666
    { 0x0667, 0 }, // 0x0667
    { 0x0668, 0 }, // 0x0668
    { 0x0669, 0 }, // 0x0669
    { 0x066A, 0 }, // 0x066A
    { 0x066B, 0 }, // 0x066B
    { 0x066C, 0 }, // 0x066C
    { 0x066D, 0 }, // 0x066D
    { 0x066E, 0 }, // 0x066E
    { 0x066F, 0 }, // 0x066F

    { 0x0670, 0 }, // 0x0670
    { 0xFB50, 1 }, // 0x0671    R       ALEF WASLA
    { 0x0672, 0 }, // 0x0672
    { 0x0673, 0 }, // 0x0673
    { 0x0674, 0 }, // 0x0674
    { 0x0675, 0 }, // 0x0675
    { 0x0676, 0 }, // 0x0676
    { 0x0677, 0 }, // 0x0677
    { 0x0678, 0 }, // 0x0678
    { 0xFB66, 3 }, // 0x0679    D       TTEH
    { 0xFB5E, 3 }, // 0x067A    D       TTEHEH
    { 0xFB52, 3 }, // 0x067B    D       BEEH
    { 0x067C, 0 }, // 0x067C
    { 0x067D, 0 }, // 0x067D
    { 0xFB56, 3 }, // 0x067E    D       PEH
    { 0xFB62, 3 }, // 0x067F    D       TEHEH

    { 0xFB5A, 3 }, // 0x0680    D       BEHEH
    { 0x0681, 0 }, // 0x0681
    { 0x0682, 0 }, // 0x0682
    { 0xFB76, 3 }, // 0x0683    D       NYEH
    { 0xFB72, 3 }, // 0x0684    D       DYEH
    { 0x0685, 0 }, // 0x0685
    { 0xFB7A, 3 }, // 0x0686    D       TCHEH
    { 0xFB7E, 3 }, // 0x0687    D       TCHEHEH
    { 0xFB88, 1 }, // 0x0688    R       DDAL
    { 0x0689, 0 }, // 0x0689
    { 0x068A, 0 }, // 0x068A
    { 0x068B, 0 }, // 0x068B
    { 0xFB84, 1 }, // 0x068C    R       DAHAL
    { 0xFB82, 1 }, // 0x068D    R       DDAHAL
    { 0xFB86, 1 }, // 0x068E    R       DUL
    { 0x068F, 0 }, // 0x068F

    { 0x0690, 0 }, // 0x0690
    { 0xFB8C, 1 }, // 0x0691    R       RREH
    { 0x0692, 0 }, // 0x0692
    { 0x0693, 0 }, // 0x0693
    { 0x0694, 0 }, // 0x0694
    { 0x0695, 0 }, // 0x0695
    { 0x0696, 0 }, // 0x0696
    { 0x0697, 0 }, // 0x0697
    { 0xFB8A, 1 }, // 0x0698    R       JEH
    { 0x0699, 0 }, // 0x0699
    { 0x069A, 0 }, // 0x069A
    { 0x069B, 0 }, // 0x069B
    { 0x069C, 0 }, // 0x069C
    { 0x069D, 0 }, // 0x069D
    { 0x069E, 0 }, // 0x069E
    { 0x069F, 0 }, // 0x069F

    { 0x06A0, 0 }, // 0x06A0
    { 0x06A1, 0 }, // 0x06A1
    { 0x06A2, 0 }, // 0x06A2
    { 0x06A3, 0 }, // 0x06A3
    { 0xFB6A, 3 }, // 0x06A4    D       VEH
    { 0x06A5, 0 }, // 0x06A5
    { 0xFB6E, 3 }, // 0x06A6    D       PEHEH
    { 0x06A7, 0 }, // 0x06A7
    { 0x06A8, 0 }, // 0x06A8
    { 0xFB8E, 3 }, // 0x06A9    D       KEHEH
    { 0x06AA, 0 }, // 0x06AA
    { 0x06AB, 0 }, // 0x06AB
    { 0x06AC, 0 }, // 0x06AC
    { 0xFBD3, 3 }, // 0x06AD    D       NG
    { 0x06AE, 0 }, // 0x06AE
    { 0xFB92, 3 }, // 0x06AF    D       GAF

    { 0x06B0, 0 }, // 0x06B0
    { 0xFB9A, 3 }, // 0x06B1    D       NGOEH
    { 0x06B2, 0 }, // 0x06B2
    { 0xFB96, 3 }, // 0x06B3    D       GUEH
    { 0x06B4, 0 }, // 0x06B4
    { 0x06B5, 0 }, // 0x06B5
    { 0x06B6, 0 }, // 0x06B6
    { 0x06B7, 0 }, // 0x06B7
    { 0x06B8, 0 }, // 0x06B8
    { 0x06B9, 0 }, // 0x06B9
    { 0xFB9E, 1 }, // 0x06BA    R       NOON GHUNNA
    { 0xFBA0, 3 }, // 0x06BB    D       RNOON
    { 0x06BC, 0 }, // 0x06BC
    { 0x06BD, 0 }, // 0x06BD
    { 0xFBAA, 3 }, // 0x06BE    D       HEH DOACHASHMEE
    { 0x06BF, 0 }, // 0x06BF

    { 0xFBA4, 1 }, // 0x06C0    R       HEH WITH YEH ABOVE
    { 0xFBA6, 3 }, // 0x06C1    D       HEH GOAL
    { 0x06C2, 0 }, // 0x06C2
    { 0x06C3, 0 }, // 0x06C3
    { 0x06C4, 0 }, // 0x06C4
    { 0xFBE0, 1 }, // 0x06C5    R       KIRGHIZ OE
    { 0xFBD9, 1 }, // 0x06C6    R       OE
    { 0xFBD7, 1 }, // 0x06C7    R       U
    { 0xFBDB, 1 }, // 0x06C8    R       YU
    { 0xFBE2, 1 }, // 0x06C9    R       KIRGHIZ YU
    { 0x06CA, 0 }, // 0x06CA
    { 0xFBDE, 1 }, // 0x06CB    R       VE
    { 0xFBFC, 3 }, // 0x06CC    D       FARSI YEH
    { 0x06CD, 0 }, // 0x06CD
    { 0x06CE, 0 }, // 0x06CE
    { 0x06CF, 0 }, // 0x06CF

    { 0xFBE4, 3 }, // 0x06D0    D       E
    { 0x06D1, 0 }, // 0x06D1
    { 0xFBAE, 1 }, // 0x06D2    R       YEH BARREE
    { 0xFBB0, 1 }, // 0x06D3    R       YEH BARREE WITH HAMZA ABOVE
    { 0x06D4, 0 }, // 0x06D4
    { 0x06D5, 0 }, // 0x06D5
    { 0x06D6, 0 }, // 0x06D6
    { 0x06D7, 0 }, // 0x06D7
    { 0x06D8, 0 }, // 0x06D8
    { 0x06D9, 0 }, // 0x06D9
    { 0x06DA, 0 }, // 0x06DA
    { 0x06DB, 0 }, // 0x06DB
    { 0x06DC, 0 }, // 0x06DC
    { 0x06DD, 0 }, // 0x06DD
    { 0x06DE, 0 }, // 0x06DE
    { 0x06DF, 0 }, // 0x06DF

    { 0x06E0, 0 }, // 0x06E0
    { 0x06E1, 0 }, // 0x06E1
    { 0x06E2, 0 }, // 0x06E2
    { 0x06E3, 0 }, // 0x06E3
    { 0x06E4, 0 }, // 0x06E4
    { 0x06E5, 0 }, // 0x06E5
    { 0x06E6, 0 }, // 0x06E6
    { 0x06E7, 0 }, // 0x06E7
    { 0x06E8, 0 }, // 0x06E8
    { 0x06E9, 0 }, // 0x06E9
    { 0x06EA, 0 }, // 0x06EA
    { 0x06EB, 0 }, // 0x06EB
    { 0x06EC, 0 }, // 0x06EC
    { 0x06ED, 0 }, // 0x06ED
    { 0x06EE, 0 }, // 0x06EE
    { 0x06EF, 0 }, // 0x06EF

    { 0x06F0, 0 }, // 0x06F0
    { 0x06F1, 0 }, // 0x06F1
    { 0x06F2, 0 }, // 0x06F2
    { 0x06F3, 0 }, // 0x06F3
    { 0x06F4, 0 }, // 0x06F4
    { 0x06F5, 0 }, // 0x06F5
    { 0x06F6, 0 }, // 0x06F6
    { 0x06F7, 0 }, // 0x06F7
    { 0x06F8, 0 }, // 0x06F8
    { 0x06F9, 0 }, // 0x06F9
    { 0x06FA, 0 }, // 0x06FA
    { 0x06FB, 0 }, // 0x06FB
    { 0x06FC, 0 }, // 0x06FC
    { 0x06FD, 0 }, // 0x06FD
    { 0x06FE, 0 }, // 0x06FE
    { 0x06FF, 0 }  // 0x06FF
};

// the arabicUnicodeMapping does not work for U+0649 ALEF MAKSURA, this table does
static const ushort alefMaksura[4] = {0xFEEF, 0xFEF0, 0xFBE8, 0xFBE9};

// this is a bit tricky. Alef always binds to the right, so the second parameter descibing the shape
// of the lam can be either initial of medial. So initial maps to the isolated form of the ligature,
// medial to the final form
static const ushort arabicUnicodeLamAlefMapping[6][4] = {
    { 0xfffd, 0xfffd, 0xfef5, 0xfef6 }, // 0x622        R       Alef with Madda above
    { 0xfffd, 0xfffd, 0xfef7, 0xfef8 }, // 0x623        R       Alef with Hamza above
    { 0xfffd, 0xfffd, 0xfffd, 0xfffd }, // 0x624        // Just to fill the table ;-)
    { 0xfffd, 0xfffd, 0xfef9, 0xfefa }, // 0x625        R       Alef with Hamza below
    { 0xfffd, 0xfffd, 0xfffd, 0xfffd }, // 0x626        // Just to fill the table ;-)
    { 0xfffd, 0xfffd, 0xfefb, 0xfefc }  // 0x627        R       Alef
};

static inline int getShape(uchar cell, int shape)
{
    // the arabicUnicodeMapping does not work for U+0649 ALEF MAKSURA, handle this here
    uint ch = (cell != 0x49)
              ? (shape ? arabicUnicodeMapping[cell][0] + shape : 0x600+cell)
              : alefMaksura[shape] ;
    return ch;
}


/*
  Two small helper functions for arabic shaping.
*/
static inline const TQChar prevChar(const TQString *str, int pos)
{
    //tqDebug("leftChar: pos=%d", pos);
    pos--;
    const TQChar *ch = str->unicode() + pos;
    while(pos > -1) {
        if(::category(*ch) != TQChar::Mark_NonSpacing)
            return *ch;
        pos--;
        ch--;
    }
    return TQChar::replacement;
}

static inline const TQChar nextChar(const TQString *str, int pos)
{
    pos++;
    int len = str->length();
    const TQChar *ch = str->unicode() + pos;
    while(pos < len) {
        //tqDebug("rightChar: %d isLetter=%d, joining=%d", pos, ch.isLetter(), ch.joining());
        if(::category(*ch) != TQChar::Mark_NonSpacing)
            return *ch;
        // assume it's a transparent char, this might not be 100% correct
        pos++;
        ch++;
    }
    return TQChar::replacement;
}


static void shapedString(const TQString *uc, int from, int len, TQChar *shapeBuffer, int *shapedLength,
                         bool reverse, GlyphAttributes *attributes, unsigned short *logClusters)
{
    Q_ASSERT((int)uc->length() >= from + len);

    if(len == 0) {
        *shapedLength = 0;
        return;
    }

    TQVarLengthArray<TQArabicProperties> props(len + 2);
    TQArabicProperties *properties = props.data();
    int f = from;
    int l = len;
    if (from > 0) {
        --f;
        ++l;
        ++properties;
    }
    if (f + l < (int)uc->length()) {
        ++l;
    }
    getArabicProperties((const unsigned short *)(uc->unicode()+f), l, props.data());

    const TQChar *ch = uc->unicode() + from;
    TQChar *data = shapeBuffer;
    int clusterStart = 0;

    for (int i = 0; i < len; i++) {
        uchar r = ch->row();
        int gpos = data - shapeBuffer;

        if (r != 0x06) {
            if (r == 0x20) {
                uchar c = ch->cell();
                if (c == 0x0c || c == 0x0d)
                    // remove ZWJ and ZWNJ
                    goto skip;
            }
            if (reverse)
                *data = mirroredChar(*ch);
            else
                *data = *ch;
        } else {
            uchar c = ch->cell();
            int pos = i + from;
            int shape = properties[i].shape;
//            tqDebug("mapping U+%x to shape %d glyph=0x%x", ch->unicode(), shape, getShape(c, shape));
            // take care of lam-alef ligatures (lam right of alef)
            ushort map;
            switch (c) {
                case 0x44: { // lam
                    const TQChar pch = nextChar(uc, pos);
                    if (pch.row() == 0x06) {
                        switch (pch.cell()) {
                            case 0x22:
                            case 0x23:
                            case 0x25:
                            case 0x27:
//                                 tqDebug(" lam of lam-alef ligature");
                                map = arabicUnicodeLamAlefMapping[pch.cell() - 0x22][shape];
                                goto next;
                            default:
                                break;
                        }
                    }
                    break;
                }
                case 0x22: // alef with madda
                case 0x23: // alef with hamza above
                case 0x25: // alef with hamza below
                case 0x27: // alef
                    if (prevChar(uc, pos).unicode() == 0x0644) {
                        // have a lam alef ligature
                        //tqDebug(" alef of lam-alef ligature");
                        goto skip;
                    }
                default:
                    break;
            }
            map = getShape(c, shape);
        next:
            *data = map;
        }
        // ##### Fixme
        //attributes[gpos].zeroWidth = zeroWidth;
        if (::category(*ch) == TQChar::Mark_NonSpacing) {
            attributes[gpos].mark = TRUE;
//             tqDebug("glyph %d (char %d) is mark!", gpos, i);
        } else {
            attributes[gpos].mark = FALSE;
            clusterStart = data - shapeBuffer;
        }
        attributes[gpos].clusterStart = !attributes[gpos].mark;
        attributes[gpos].combiningClass = combiningClass(*ch);
        attributes[gpos].justification = properties[i].justification;
//         tqDebug("data[%d] = %x (from %x)", gpos, (uint)data->unicode(), ch->unicode());
        data++;
    skip:
        ch++;
        logClusters[i] = clusterStart;
    }
    *shapedLength = data - shapeBuffer;
}

#if defined(Q_WS_X11) && !defined(TQT_NO_XFTFREETYPE)

enum {
    InitProperty = 0x2,
    IsolProperty = 0x4,
    FinaProperty = 0x8,
    MediProperty = 0x10,
    RligProperty = 0x20,
    CaltProperty = 0x40,
    LigaProperty = 0x80,
    DligProperty = 0x100,
    CswhProperty = 0x200,
    MsetProperty = 0x400
};

static const TQOpenType::Features arabic_features[] = {
    { FT_MAKE_TAG('c', 'c', 'm', 'p'), CcmpProperty },
    { FT_MAKE_TAG('i', 's', 'o', 'l'), IsolProperty },
    { FT_MAKE_TAG('f', 'i', 'n', 'a'), FinaProperty },
    { FT_MAKE_TAG('m', 'e', 'd', 'i'), MediProperty },
    { FT_MAKE_TAG('i', 'n', 'i', 't'), InitProperty },
    { FT_MAKE_TAG('r', 'l', 'i', 'g'), RligProperty },
    { FT_MAKE_TAG('c', 'a', 'l', 't'), CaltProperty },
    { FT_MAKE_TAG('l', 'i', 'g', 'a'), LigaProperty },
    { FT_MAKE_TAG('d', 'l', 'i', 'g'), DligProperty },
    { FT_MAKE_TAG('c', 's', 'w', 'h'), CswhProperty },
    // mset is used in old Win95 fonts that don't have a 'mark' positioning table.
    { FT_MAKE_TAG('m', 's', 'e', 't'), MsetProperty },
    {0, 0}
};

static const TQOpenType::Features syriac_features[] = {
    { FT_MAKE_TAG('c', 'c', 'm', 'p'), CcmpProperty },
    { FT_MAKE_TAG('i', 's', 'o', 'l'), IsolProperty },
    { FT_MAKE_TAG('f', 'i', 'n', 'a'), FinaProperty },
    { FT_MAKE_TAG('f', 'i', 'n', '2'), FinaProperty },
    { FT_MAKE_TAG('f', 'i', 'n', '3'), FinaProperty },
    { FT_MAKE_TAG('m', 'e', 'd', 'i'), MediProperty },
    { FT_MAKE_TAG('m', 'e', 'd', '2'), MediProperty },
    { FT_MAKE_TAG('i', 'n', 'i', 't'), InitProperty },
    { FT_MAKE_TAG('r', 'l', 'i', 'g'), RligProperty },
    { FT_MAKE_TAG('c', 'a', 'l', 't'), CaltProperty },
    { FT_MAKE_TAG('l', 'i', 'g', 'a'), LigaProperty },
    { FT_MAKE_TAG('d', 'l', 'i', 'g'), DligProperty },
    {0, 0}
};

static bool arabicSyriacOpenTypeShape(TQOpenType *openType, TQShaperItem *item, bool *ot_ok)
{
    *ot_ok = true;

    openType->selectScript(item->script, item->script == TQFont::Arabic ? arabic_features : syriac_features);
    int nglyphs = item->num_glyphs;
    if (item->font->stringToCMap(item->string->unicode()+item->from, item->length, item->glyphs, item->advances,
                                 &item->num_glyphs, item->flags & TQTextEngine::RightToLeft) != TQFontEngine::NoError)
        return FALSE;
    heuristicSetGlyphAttributes(item);

    unsigned short *logClusters = item->log_clusters;
    const unsigned short *uc = (const unsigned short *)item->string->unicode() + item->from;

    TQVarLengthArray<TQArabicProperties> props(item->length+2);
    TQArabicProperties *properties = props.data();
    int f = 0;
    int l = item->length;
    if (item->from > 0) {
        --f;
        ++l;
        ++properties;
    }
    if (f + l < (int)item->string->length()) {
        ++l;
    }
    getArabicProperties((const unsigned short *)(uc+f), l, props.data());

    TQVarLengthArray<uint> apply(item->num_glyphs);


    // Hack to remove ZWJ and ZWNJ from rendered output.
    int j = 0;
    for (int i = 0; i < item->num_glyphs; i++) {
        if (uc[i] == 0x200c || uc[i] == 0x200d)
            continue;
        item->glyphs[j] = item->glyphs[i];
        item->attributes[j] = item->attributes[i];
        item->advances[j] = item->advances[i];
        item->offsets[j] = item->offsets[i];
        properties[j] = properties[i];
        item->attributes[j].justification = properties[i].justification;
        logClusters[i] = logClusters[j];
        ++j;
    }
    item->num_glyphs = j;

    for (int i = 0; i < item->num_glyphs; i++) {
        apply[i] = 0;

        if (properties[i].shape == XIsolated)
            apply[i] |= MediProperty|FinaProperty|InitProperty;
        else if (properties[i].shape == XMedial)
            apply[i] |= IsolProperty|FinaProperty|InitProperty;
        else if (properties[i].shape == XFinal)
            apply[i] |= IsolProperty|MediProperty|InitProperty;
        else if (properties[i].shape == XInitial)
            apply[i] |= IsolProperty|MediProperty|FinaProperty;
    }

    if (!openType->shape(item, apply.data())) {
        *ot_ok = false;
        return false;
    }
    item->num_glyphs = nglyphs;
    return openType->positionAndAdd(item);
}

#endif

// #### stil missing: identify invalid character combinations
static bool arabic_shape(TQShaperItem *item)
{
    Q_ASSERT(item->script == TQFont::Arabic);

#if defined(Q_WS_X11) && !defined(TQT_NO_XFTFREETYPE)
    TQOpenType *openType = item->font->openType();

    if (openType && openType->supportsScript(TQFont::Arabic)) {
        bool ot_ok;
        if (arabicSyriacOpenTypeShape(openType, item, &ot_ok))
            return true;
        if (ot_ok)
            return false;
        // fall through to the non OT code
    }
#endif

    TQVarLengthArray<ushort> shapedChars(item->length);

    int slen;
    shapedString(item->string, item->from, item->length, (TQChar *)shapedChars.data(), &slen,
                  item->flags & TQTextEngine::RightToLeft,
                  item->attributes, item->log_clusters);

    if (item->font->stringToCMap((TQChar *)shapedChars.data(), slen, item->glyphs, item->advances,
                                 &item->num_glyphs, item->flags & TQTextEngine::RightToLeft) != TQFontEngine::NoError)
        return FALSE;

    for (int i = 0; i < slen; ++i)
        if (item->attributes[i].mark)
            item->advances[i] = 0;
    qt_heuristicPosition(item);
    return TRUE;
}

#if defined(Q_WS_X11)
# include "qscriptengine_x11.cpp"
#elif defined(Q_WS_WIN)
# include "qscriptengine_win.cpp"
#elif defined(Q_WS_MAC)
# include "qscriptengine_mac.cpp"
#elif defined(Q_WS_QWS)
# include "qscriptengine_qws.cpp"
#endif