path: root/src/qjson_p.h

/****************************************************************************
**
** Copyright (C) 2012 Nokia Corporation and/or its subsidiary(-ies).
** All rights reserved.
** Contact: Nokia Corporation (qt-info@nokia.com)
**
** This file is part of the QtCore module of the Qt Toolkit.
**
** $QT_BEGIN_LICENSE:LGPL$
** GNU Lesser General Public License Usage
** This file may be used under the terms of the GNU Lesser General Public
** License version 2.1 as published by the Free Software Foundation and
** appearing in the file LICENSE.LGPL included in the packaging of this
** file. Please review the following information to ensure the GNU Lesser
** General Public License version 2.1 requirements will be met:
** http://www.gnu.org/licenses/old-licenses/lgpl-2.1.html.
**
** In addition, as a special exception, Nokia gives you certain additional
** rights. These rights are described in the Nokia Qt LGPL Exception
** version 1.1, included in the file LGPL_EXCEPTION.txt in this package.
**
** GNU General Public License Usage
** Alternatively, this file may be used under the terms of the GNU General
** Public License version 3.0 as published by the Free Software Foundation
** and appearing in the file LICENSE.GPL included in the packaging of this
** file. Please review the following information to ensure the GNU General
** Public License version 3.0 requirements will be met:
** http://www.gnu.org/copyleft/gpl.html.
**
** Other Usage
** Alternatively, this file may be used in accordance with the terms and
** conditions contained in a signed written agreement between you and Nokia.
**
**
**
**
**
** $QT_END_LICENSE$
**
****************************************************************************/

#ifndef QJSON_P_H
#define QJSON_P_H

//
//  W A R N I N G
//  -------------
//
// This file is not part of the Qt API.  It exists purely as an
// implementation detail.  This header file may change from version to
// version without notice, or even be removed.
//
// We mean it.
//

#include <qjsonglobal.h>
#include <qjsonobject.h>
#include <qjsonvalue.h>
#include <qjsondocument.h>
#include <qjsonarray.h>
#include <qoldbasicatomic.h>
#include <qstring.h>
#include <qendian.h>

/*
  This defines a binary data structure for Json data. The data structure is optimised for fast reading and minimum allocations.
  The whole data structure can be mmap'ed and used directly.

  In most cases the binary structure is not as space efficient as a utf8 encoded text representation, but much faster to access.

  The size requirements are:

  String: 16 bytes header + 2*(string.length())

  Values: 4 bytes + size of data (size can be 0 for some data)
    bool: 0 bytes
    number: 8 bytes
    string: 4 + 2*length of string
    array: size of array
    object: size of object
  Array: 12 bytes + 4*length + size of Value data
  Object: 12 bytes + 8*length + size of Key Strings + size of Value data

  For an example such as

    {                                           // object: 12 + 5*8                = 52
         "firstName": "John",                   // key 4+12, value 8          = 24
         "lastName" : "Smith",                  // key 4+8, value 8         = 20
         "age"      : 25,                       // key 4+4, value 0             = 8
         "address"  :                           // key 4+8, object below = 154
         {                                      // object: 12 + 4*8
             "streetAddress": "21 2nd Street",  // key 4+16, value 16
             "city"         : "New York",       // key 4+4, value 12
             "state"        : "NY",             // key 4+8, value 4
             "postalCode"   : "10021"           // key 4+12, value 8
         },                                     // object total: 148
         "phoneNumber":                         // key: 4+12, value array below = 208
         [                                      // array: 12 + 2*4 + values below = 176
             {                                  // object 12 + 2*8
               "type"  : "home",                // key 4+4, value 8
               "number": "212 555-1234"         // key 4+8, value 16
             },                                 // object total: 84
             {                                  // object 12 + 2*8
               "type"  : "fax",                 // key 4+4, value 8
               "number": "646 555-4567"         // key 4+8, value 16
             }                                  // object total: 84
         ]                                      // array total: 248
    }                                           // great total:                       416 bytes

    The uncompressed text file used roughly 500 bytes, so we end up using about the same space
    as the text representation.
*/
namespace QtJson
{

namespace Private {

struct Array;
struct Object;
struct Value;
struct Entry;

template<typename T>
struct q_littleendian
{
    T val;

    q_littleendian &operator =(T i) { val = qToLittleEndian(i); return *this; }
    operator T() const { return qFromLittleEndian(val); }

    bool operator ==(T i) { return qFromLittleEndian(val) == i; }
    bool operator !=(T i) { return qFromLittleEndian(val) != i; }
    bool operator <(T i) { return qFromLittleEndian(val) < i; }
    bool operator >(T i) { return qFromLittleEndian(val) > i; }
    bool operator <=(T i) { return qFromLittleEndian(val) <= i; }
    bool operator >=(T i) { return qFromLittleEndian(val) >= i; }
    q_littleendian &operator +=(T i) {
        val = qToLittleEndian(qFromLittleEndian(val) + i);
        return *this;
    }
};

typedef q_littleendian<short> qle_short;
typedef q_littleendian<unsigned short> qle_ushort;
typedef q_littleendian<int> qle_int;
typedef q_littleendian<unsigned int> qle_uint;

template<int pos, int width>
struct qle_bitfield
{
    uint val;

    enum {
        mask = ((1u << width) - 1) << pos
    };

    void operator =(uint t) {
        uint i = qFromLittleEndian(val);
        i &= ~mask;
        i |= t << pos;
        val = qToLittleEndian(i);
    }
    operator uint() const {
        uint t = qFromLittleEndian(val);
        t &= mask;
        t >>= pos;
        return t;
    }
    bool operator !() const {
        return !operator uint();
    }

    bool operator ==(uint t) { return uint(*this) == t; }
    bool operator !=(uint t) { return uint(*this) != t; }
    bool operator <(uint t) { return uint(*this) < t; }
    bool operator >(uint t) { return uint(*this) > t; }
    bool operator <=(uint t) { return uint(*this) <= t; }
    bool operator >=(uint t) { return uint(*this) >= t; }
    qle_bitfield &operator +=(uint i) {
        *this = (uint(*this) + i);
        return *this;
    }
    qle_bitfield &operator -=(uint i) {
        *this = (uint(*this) - i);
        return *this;
    }
};

template<int pos, int width>
struct qle_signedbitfield
{
    uint val;

    enum {
        mask = ((1u << width) - 1) << pos
    };

    void operator =(int t) {
        uint i = qFromLittleEndian(val);
        i &= ~mask;
        i |= t << pos;
        val = qToLittleEndian(i);
    }
    operator int() const {
        uint i = qFromLittleEndian(val);
        i <<= 32 - width - pos;
        int t = (int) i;
        t >>= pos;
        return t;
    }
    bool operator !() const {
        return !operator int();
    }

    bool operator ==(int t) { return int(*this) == t; }
    bool operator !=(int t) { return int(*this) != t; }
    bool operator <(int t) { return int(*this) < t; }
    bool operator >(int t) { return int(*this) > t; }
    bool operator <=(int t) { return int(*this) <= t; }
    bool operator >=(int t) { return int(*this) >= t; }
    qle_signedbitfield &operator +=(int i) {
        *this = (int(*this) + i);
        return *this;
    }
    qle_signedbitfield &operator -=(int i) {
        *this = (int(*this) - i);
        return *this;
    }
};

typedef qle_uint offset;

static inline int alignedSize(int size) { return (size + 3) & ~3; }

static inline bool useCompressed(const QString &s)
{
    if (s.length() >= 0x8000)
        return false;
    const ushort *uc = (const ushort *)s.constData();
    const ushort *e = uc + s.length();
    while (uc < e) {
        if (*uc > 0xff)
            return false;
        ++uc;
    }
    return true;
}

static inline int qStringSize(const QString &string, bool compress)
{
    int l = 2 + string.length();
    if (!compress)
        l *= 2;
    return alignedSize(l);
}


// returns INT_MAX if it can't compress it into 28 bits
static inline int compressedNumber(double d)
{
    // this relies on details of how ieee floats are represented
    const int exponent_off = 52;
    const quint64 fraction_mask = 0x000fffffffffffffull;
    const quint64 exponent_mask = 0x7ff0000000000000ull;

    union {
        quint64 val;
        double dbl;
    };
    dbl = d;
    int exp = (int)((val & exponent_mask) >> exponent_off) - 1023;
    if (exp < 0 || exp > 25)
        return INT_MAX;

    quint64 non_int = val & (fraction_mask >> exp);
    if (non_int)
        return INT_MAX;

    bool neg = (val >> 63);
    val &= fraction_mask;
    val |= ((quint64)1 << 52);
    int res = (int)(val >> (52 - exp));
    return neg ? -res : res;
}

struct Latin1String;

struct String
{
    String(const char *data) { d = (Data *)data; }

    struct Data {
        qle_int length;
        qle_ushort utf16[1];
    };

    Data *d;

    inline String &operator=(const QString &str)
    {
        d->length = str.length();
#if Q_BYTE_ORDER == Q_BIG_ENDIAN
        for (int i = 0; i < str.length(); ++i)
            d->utf16[i] = uc[i];
#else
        memcpy(d->utf16, str.unicode(), str.length()*sizeof(ushort));
#endif
        if (str.length() & 1)
            d->utf16[str.length()] = 0;
        return *this;
    }

    bool operator ==(const QString &str) {
        int slen = str.length();
        const ushort *s = (const ushort *)str.constData();
        int l = d->length;
        if (slen != l)
            return false;
        const qle_ushort *a = d->utf16;
        const ushort *b = s;
        while (l-- && *a == *b)
            a++,b++;
        return (l == -1);
    }
    bool operator !=(const QString &str) {
        return !operator ==(str);
    }
    QString toString() const {
#if Q_BYTE_ORDER == Q_LITTLE_ENDIAN
        return QString((QChar *)d->utf16, d->length);
#else
        int l = d->length;
        QString str(l, QChar());
        QChar *ch = str.data();
        for (int i = 0; i < l; ++i)
            ch[i] = d->utf16[i];
        return str;
#endif
    }

};

struct Latin1String
{
    Latin1String(const char *data) { d = (Data *)data; }

    struct Data {
        qle_short length;
        char latin1[1];
    };
    Data *d;

    inline Latin1String &operator=(const QString &str)
    {
        d->length = str.length();
        uchar *l = (uchar *)d->latin1;
        const ushort *uc = (const ushort *)str.unicode();
        for (int i = 0; i < str.length(); ++i)
            *l++ = uc[i];
        while ((quintptr)l & 0x3)
            *l++ = 0;
        return *this;
    }

    bool operator ==(const QString &str) {
        int slen = str.length();
        const ushort *s = (const ushort *)str.constData();
        if (slen != d->length)
            return false;
        int l = d->length;
        const uchar *a = (uchar *)d->latin1;
        const ushort *b = s;
        while (l-- && *a == *b)
            a++,b++;
        return (l == -1);
    }
    bool operator !=(const QString &str) {
        return !operator ==(str);
    }

    QString toString() const {
        return QString::fromLatin1(d->latin1, d->length);
    }
};


static inline void copyString(char *dest, const QString &str, bool compress)
{
    if (compress) {
        Latin1String string(dest);
        string = str;
    } else {
        String string(dest);
        string = str;
    }
}



struct Base
{
    qle_uint size;
    union {
        uint _dummy;
        qle_bitfield<0, 1> is_object;
        qle_bitfield<1, 31> length;
    };
    offset tableOffset;
    // content follows here

    inline bool isObject() const { return is_object; }
    inline bool isArray() const { return !isObject(); }

    inline offset *table() const { return (offset *) (((char *) this) + tableOffset); }

    int reserveSpace(uint dataSize, int posInTable, uint numItems, bool replace);
    void removeItems(int pos, int numItems);
};

struct Object : public Base
{

    Entry *entryAt(int i) const {
        return reinterpret_cast<Entry *>(((char *)this) + table()[i]);
    }
    int indexOf(const QString &key);

    bool isValid() const;
};


struct Array : public Base
{
    inline Value at(int i) const;
    inline Value &operator [](int i);

    bool isValid() const;
};


struct Value
{
    // unfortunately some compilers can't handle mixed types in bitfields.
    // this works around the problem
    union {
        uint _dummy;
        qle_bitfield<0, 3> type;
        qle_bitfield<3, 1> latinOrIntValue;
        qle_bitfield<4, 1> latinKey;
        qle_bitfield<5, 27> val;
        qle_signedbitfield<5, 27> int_val;
    };

    inline char *data(const Base *b) const { return ((char *)b) + val; }
    int usedStorage(const Base *b) const;

    bool toBoolean() const;
    double toDouble(const Base *b) const;
    QString toString(const Base *b) const;
    String asString(const Base *b) const;
    Latin1String asLatin1String(const Base *b) const;
    Base *objectOrArray(const Base *b) const;

    bool isValid(const Base *b) const;
};

inline Value Array::at(int i) const
{
    return * (Value *) (((char *) this) + tableOffset + i*sizeof(Value));
}

inline Value &Array::operator [](int i)
{
    return * (Value *) (((char *) this) + tableOffset + i*sizeof(Value));
}



struct Entry {
    Value value;
    // key
    // value data follows key

    int size() const {
        int s = sizeof(Entry);
        if (value.latinKey)
            s += sizeof(ushort) + *(ushort *) ((const char *)this + sizeof(Entry));
        else
            s += sizeof(uint) + *(int *) ((const char *)this + sizeof(Entry));
        return alignedSize(s);
    }

    int usedStorage(Base *b) const {
        return size() + value.usedStorage(b);
    }

    String shallowKey() const
    {
        Q_ASSERT(!value.latinKey);
        return String((const char *)this + sizeof(Entry));
    }
    Latin1String shallowLatin1Key() const
    {
        Q_ASSERT(value.latinKey);
        return Latin1String((const char *)this + sizeof(Entry));
    }
    QString key() const
    {
        if (value.latinKey) {
            return shallowLatin1Key().toString();
        }
        return shallowKey().toString();
    }

    bool matchesKey(const QString &key);
};

struct Header {
    qle_uint tag; // 'qbjs'
    qle_uint version; // 1
    Base *root() { return (Base *)(this + 1); }
};


inline bool Value::toBoolean() const
{
    Q_ASSERT(type == QJsonValue::Bool);
    return val != 0;
}

inline double Value::toDouble(const Base *b) const
{
    Q_ASSERT(type == QJsonValue::Double);
    if (latinOrIntValue)
        return int_val;

    union {
        quint64 i;
        double d;
    };
    i = qFromLittleEndian<quint64>((const uchar *)b + val);
    return d;
}

inline QString Value::toString(const Base *b) const
{
    char *d = data(b);
    if (latinOrIntValue) {
        int l = *(ushort *)d;
        return QString::fromLatin1(d + sizeof(ushort), l);
    }
    int l = *(int *)d;
    const QChar *c = (const QChar *)(d + sizeof(int));
    return QString(c, l);
}

inline String Value::asString(const Base *b) const
{
    Q_ASSERT(type == QJsonValue::String && !latinOrIntValue);
    return String(data(b));
}

inline Latin1String Value::asLatin1String(const Base *b) const
{
    Q_ASSERT(type == QJsonValue::String && latinOrIntValue);
    return Latin1String(data(b));
}

inline Base *Value::objectOrArray(const Base *b) const
{
    Q_ASSERT(type == QJsonValue::Array || type == QJsonValue::Object);
    return reinterpret_cast<Base *>(data(b));
}

struct Data {
    enum Validation {
        Unchecked,
        Validated,
        Invalid
    };

    inline Data(char *raw, int a)
        : alloc(a), compactionCounter(0), valid(Unchecked), ownsData(true), rawData(raw)
    {
        ref.store(0);
    }
    inline Data(int reserved, QJsonValue::Type valueType)
        : compactionCounter(0), valid(Validated), ownsData(true), rawData(0)
    {
        ref.store(0);

        alloc = sizeof(Header) + sizeof(Base) + reserved + sizeof(offset);
        header = (Header *)malloc(alloc);
        header->tag = QJsonDocument::BinaryFormatTag;
        header->version = 1;
        Base *b = header->root();
        b->size = sizeof(Base);
        b->is_object = (valueType == QJsonValue::Object);
        b->tableOffset = sizeof(Base);
        b->length = 0;
    }
    inline ~Data()
    { if (ownsData) free(rawData); }

    QBasicAtomicInt ref;
    int alloc;
    int compactionCounter;
    Validation valid;
    bool ownsData;
    union {
        char *rawData;
        Header *header;
    };

    uint offsetOf(const void *ptr) const { return ((char *)ptr - rawData); }

    Object *object(int offset) const { return reinterpret_cast<Object *>(rawData + offset); }
    Entry *entry(int offset) const { return reinterpret_cast<Entry *>(rawData + offset); }
    Value *value(int offset) const { return reinterpret_cast<Value *>(rawData + offset); }
    Array *array(int offset) const { return reinterpret_cast<Array *>(rawData + offset); }


    QJsonObject toObject(Object *o) const
    {
        return QJsonObject(const_cast<Data *>(this), o);
    }

    QJsonArray toArray(Array *a) const
    {
        return QJsonArray(const_cast<Data *>(this), a);
    }

    Data *detach(Base *b, int reserve = 0)
    {
        int size = sizeof(Header) + b->size + reserve;
        char *raw = (char *)malloc(size);
        memcpy(raw + sizeof(Header), b, b->size);
        Header *h = (Header *)raw;
        h->tag = QJsonDocument::BinaryFormatTag;
        h->version = 1;
        Data *d = new Data(raw, size);
        d->compactionCounter = compactionCounter;
        return d;
    }

    void compact();
    void validate();
};

}

}

#endif // QJSON_P_H