/*  -*- C++ -*-

    SPDX-FileCopyrightText: 1998 Netscape Communications Corporation <developer@mozilla.org>

    SPDX-License-Identifier: MIT

*/

#ifndef CharDistribution_h__

#define CharDistribution_h__

#include "kcodecs_export.h"

#define ENOUGH_DATA_THRESHOLD 256

namespace kencodingprober

{

class KCODECS_NO_EXPORT CharDistributionAnalysis

{

public:

    CharDistributionAnalysis() = default;

    virtual ~CharDistributionAnalysis() = default;

    // Feed a character with known length

    void HandleOneChar(const char *aStr, unsigned int aCharLen)

    {

        // we only care about 2-bytes character in our distribution analysis

        const int code = (aCharLen == 2) ? GetCode(aStr) : -1;

        if (code >= 0) {

            mTotalChars++;

            // code is valid

            if ((unsigned int)code < mTableSize) {

                if (mCharToFreqOrder[code] < 512) {

                    mFreqChars++;

                }

            }

        }

    }

    // return confidence base on existing data

    float GetConfidence();

    // It is not necessary to receive all data to draw conclusion. For charset detection,

    // certain amount of data is enough

    bool GotEnoughData()

    {

        return mTotalChars > ENOUGH_DATA_THRESHOLD;

    }

protected:

    // Characters are not handled based on its original encoded value, but

    // converted to an encoding specific unique code.

    // This allows multiple encoding formats (e.g. SJIS and EUCJP) of an

    // encoding (like JIS X 213) to share one frequency table, mapping this

    // code to its frequency.

    virtual int GetCode(const char * /* str */) = 0;

    // The number of characters whose frequency order is less than 512

    unsigned int mFreqChars = 0;

    // Total character encountered.

    unsigned int mTotalChars = 0;

    // Mapping table to get frequency order from code (from GetCode())

    const short *mCharToFreqOrder = nullptr;

    // Size of above table

    unsigned int mTableSize = 0;

    // This is a constant value varies from language to language, it is used in

    // calculating confidence. See my paper for further detail.

    float mTypicalDistributionRatio = 0.0f;

};

class KCODECS_NO_EXPORT EUCKRDistributionAnalysis : public CharDistributionAnalysis

{

public:

    EUCKRDistributionAnalysis();

protected:

    // for euc-KR encoding, we are interested

    //  first  byte range: 0xb0 -- 0xfe

    //  second byte range: 0xa1 -- 0xfe

    // no validation needed here. State machine has done that

    int GetCode(const char *str) override

    {

        if ((unsigned char)*str >= (unsigned char)0xb0) {

            return 94 * ((unsigned char)str[0] - (unsigned char)0xb0) + (unsigned char)str[1] - (unsigned char)0xa1;

        } else {

            return -1;

        }

    }

};

class KCODECS_NO_EXPORT GB2312DistributionAnalysis : public CharDistributionAnalysis

{

public:

    GB2312DistributionAnalysis();

protected:

    // for GB2312 encoding, we are interested

    //  first  byte range: 0xb0 -- 0xfe

    //  second byte range: 0xa1 -- 0xfe

    // no validation needed here. State machine has done that

    int GetCode(const char *str) override

    {

        if ((unsigned char)*str >= (unsigned char)0xb0 && (unsigned char)str[1] >= (unsigned char)0xa1) {

            return 94 * ((unsigned char)str[0] - (unsigned char)0xb0) + (unsigned char)str[1] - (unsigned char)0xa1;

        } else {

            return -1;

        }

    }

};

class KCODECS_NO_EXPORT Big5DistributionAnalysis : public CharDistributionAnalysis

{

public:

    Big5DistributionAnalysis();

protected:

    // for big5 encoding, we are interested

    //  first  byte range: 0xa4 -- 0xfe

    //  second byte range: 0x40 -- 0x7e , 0xa1 -- 0xfe

    // no validation needed here. State machine has done that

    int GetCode(const char *str) override

    {

        if ((unsigned char)*str >= (unsigned char)0xa4)

            if ((unsigned char)str[1] >= (unsigned char)0xa1) {

                return 157 * ((unsigned char)str[0] - (unsigned char)0xa4) + (unsigned char)str[1] - (unsigned char)0xa1 + 63;

            } else {

                return 157 * ((unsigned char)str[0] - (unsigned char)0xa4) + (unsigned char)str[1] - (unsigned char)0x40;

            }

        else {

            return -1;

        }

    }

};

class KCODECS_NO_EXPORT SJISDistributionAnalysis : public CharDistributionAnalysis

{

public:

    SJISDistributionAnalysis();

protected:

    // for sjis encoding, we are interested

    //  first  byte range: 0x81 -- 0x9f , 0xe0 -- 0xfe

    //  second byte range: 0x40 -- 0x7e,  0x81 -- oxfe

    // no validation needed here. State machine has done that

    int GetCode(const char *str) override

    {

        int code;

        if ((unsigned char)*str >= (unsigned char)0x81 && (unsigned char)*str <= (unsigned char)0x9f) {

            code = 188 * ((unsigned char)str[0] - (unsigned char)0x81);

        } else if ((unsigned char)*str >= (unsigned char)0xe0 && (unsigned char)*str <= (unsigned char)0xef) {

            code = 188 * ((unsigned char)str[0] - (unsigned char)0xe0 + 31);

        } else {

            return -1;

        }

        code += (unsigned char)*(str + 1) - 0x40;

        if ((unsigned char)str[1] > (unsigned char)0x7f) {

            code--;

        }

        return code;

    }

};

class KCODECS_NO_EXPORT EUCJPDistributionAnalysis : public CharDistributionAnalysis

{

public:

    EUCJPDistributionAnalysis();

protected:

    // for euc-JP encoding, we are interested

    //  first  byte range: 0xa0 -- 0xfe

    //  second byte range: 0xa1 -- 0xfe

    // no validation needed here. State machine has done that

    int GetCode(const char *str) override

    {

        if ((unsigned char)*str >= (unsigned char)0xa0) {

            return 94 * ((unsigned char)str[0] - (unsigned char)0xa1) + (unsigned char)str[1] - (unsigned char)0xa1;

        } else {

            return -1;

        }

    }

};

}

#endif // CharDistribution_h__