qscriptengine.cpp

qt-x11-opensource-src-4.1.4.tar.gz源码

        const int availableGlyphs = item->num_glyphs;
        if (!item->font->stringToCMap(item->string->unicode()+item->from, item->length, item->glyphs, &item->num_glyphs, QFlag(item->flags)))
            return false;

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

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

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

    *shapedChars = *uc;
    logClusters[0] = 0;
    int slen = 1;
    int cluster_start = 0;
    for (int 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;
            glyphs[slen].attributes.clusterStart = true;
            glyphs[slen].attributes.mark = false;
            glyphs[slen].attributes.combiningClass = 0;
            cluster_start = slen;
            ++slen;
        }
        if (shaped) {
            if (item->font->canRender((QChar *)&shaped, 1)) {
                shapedChars[slen-1] = QChar(shaped);
            } else
                shaped = 0;
        }
        if (!shaped) {
            shapedChars[slen] = uc[i];
            if (::category(uc[i]) != QChar::Mark_NonSpacing) {
                glyphs[slen].attributes.clusterStart = true;
                glyphs[slen].attributes.mark = false;
                glyphs[slen].attributes.combiningClass = 0;
                glyphs[slen].attributes.dontPrint = qIsControlChar(uc[i].unicode());
                cluster_start = slen;
            } else {
                glyphs[slen].attributes.clusterStart = false;
                glyphs[slen].attributes.mark = true;
                glyphs[slen].attributes.combiningClass = ::combiningClass(uc[i]);
            }
            ++slen;
        }
        logClusters[i] = cluster_start;
    }

    if (!item->font->stringToCMap(shapedChars, slen, glyphs, &item->num_glyphs, QFlag(item->flags))) {
        if (item->length > 256)
            ::free(shapedChars);
        return false;
    }
    for (int i = 0; i < item->num_glyphs; ++i) {
        if (glyphs[i].attributes.mark) {
            glyphs[i].advance.x = 0;
        }
    }
    qt_heuristicPosition(item);

    if (item->length > 256)
        ::free(shapedChars);
    return true;
}
#endif

// 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) == QChar::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 QChar::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 Joining {
    JNone,
    JCausing,
    JDual,
    JRight,
    JTransparent
};

static const Joining joining_for_group[ArabicGroupsEnd] = {
    // NonJoining
    JNone, // ArabicNone
    JNone, // ArabicSpace
    // Transparent
    JTransparent, // Transparent
    // Causing