#include <validate_khmer.h>

Inheritance diagram for tesseract::ValidateKhmer:

Public Member Functions
	ValidateKhmer (ViramaScript script, bool report_errors)

	~ValidateKhmer () override=default

Public Member Functions inherited from tesseract::Validator
virtual	~Validator ()

Protected Member Functions
bool	ConsumeGraphemeIfValid () override

CharClass	UnicodeToCharClass (char32 ch) const override

Protected Member Functions inherited from tesseract::Validator
	Validator (ViramaScript script, bool report_errors)

bool	ValidateCleanAndSegmentInternal (GraphemeNormMode g_mode, const std::vector< char32 > &src, std::vector< std::vector< char32 > > *dest)

void	MoveResultsToDest (GraphemeNormMode g_mode, std::vector< std::vector< char32 > > *dest)

bool	IsSubscriptScript () const

bool	CodeOnlyToOutput ()

void	MultiCodePart (unsigned length)

bool	UseMultiCode (unsigned length)

virtual bool	ConsumeGraphemeIfValid ()=0

void	ComputeClassCodes (const std::vector< char32 > &text)

virtual CharClass	UnicodeToCharClass (char32 ch) const =0

void	Clear ()

Additional Inherited Members
Static Public Member Functions inherited from tesseract::Validator
static bool	ValidateCleanAndSegment (GraphemeNormMode g_mode, bool report_errors, const std::vector< char32 > &src, std::vector< std::vector< char32 > > *dest)

static bool	IsZeroWidthMark (char32 ch)

Static Public Attributes inherited from tesseract::Validator
static const char32	kZeroWidthSpace = 0x200B

static const char32	kZeroWidthNonJoiner = 0x200C

static const char32	kZeroWidthJoiner = 0x200D

static const char32	kLeftToRightMark = 0x200E

static const char32	kRightToLeftMark = 0x200F

static const char32	kInvalid = 0xfffd

Protected Types inherited from tesseract::Validator
enum class	CharClass { kConsonant = 'C' , kVowel = 'V' , kVirama = 'H' , kMatra = 'M' , kMatraPiece = 'P' , kVowelModifier = 'D' , kZeroWidthNonJoiner = 'z' , kZeroWidthJoiner = 'Z' , kVedicMark = 'v' , kNukta = 'N' , kRobat = 'R' , kOther = 'O' , kWhitespace = ' ' , kCombiner = 'c' }

using	IndicPair = std::pair< CharClass, char32 >

Static Protected Member Functions inherited from tesseract::Validator
static std::unique_ptr< Validator >	ScriptValidator (ViramaScript script, bool report_errors)

static ViramaScript	MostFrequentViramaScript (const std::vector< char32 > &utf32)

static bool	IsVirama (char32 unicode)

static bool	IsVedicAccent (char32 unicode)

Protected Attributes inherited from tesseract::Validator
ViramaScript	script_

std::vector< IndicPair >	codes_

std::vector< std::vector< char32 > >	parts_

std::vector< char32 >	output_

unsigned	codes_used_

unsigned	output_used_

bool	report_errors_

Static Protected Attributes inherited from tesseract::Validator
static const int	kIndicCodePageSize = 128

static const char32	kMinIndicUnicode = 0x900

static const char32	kMaxSinhalaUnicode = 0xdff

static const char32	kMaxViramaScriptUnicode = 0x17ff

static const char32	kSinhalaVirama = 0xdca

static const char32	kMyanmarVirama = 0x1039

static const char32	kKhmerVirama = 0x17d2

static const char32	kJavaneseVirama = 0xa9c0

static const char32	kMaxJavaneseUnicode = 0xa9df

Detailed Description

Definition at line 9 of file validate_khmer.h.

Constructor & Destructor Documentation

◆ ValidateKhmer()

tesseract::ValidateKhmer::ValidateKhmer	(	ViramaScript	script,
		bool	report_errors
	)

inline

Definition at line 11 of file validate_khmer.h.

11: Validator(script, report_errors) {}

tesseract::Validator::Validator

Validator(ViramaScript script, bool report_errors)

Definition: validator.h:137

◆ ~ValidateKhmer()

tesseract::ValidateKhmer::~ValidateKhmer ( )

overridedefault

Member Function Documentation

◆ ConsumeGraphemeIfValid()

bool tesseract::ValidateKhmer::ConsumeGraphemeIfValid ( )

overrideprotectedvirtual

Implements tesseract::Validator.

Definition at line 20 of file validate_khmer.cpp.

                                           {
  const unsigned num_codes = codes_.size();
  if (codes_used_ == num_codes) {
    return false;
  }
  if (codes_[codes_used_].first == CharClass::kOther) {
    UseMultiCode(1);
    return true;
  }
  if (codes_[codes_used_].first != CharClass::kConsonant) {
    if (report_errors_) {
      tprintf("Invalid start of Khmer syllable:0x%x\n", codes_[codes_used_].second);
    }
    return false;
  }
  if (UseMultiCode(1)) {
    return true;
  }
  if (codes_[codes_used_].first == CharClass::kRobat ||
      codes_[codes_used_].first == CharClass::kNukta) {
    if (UseMultiCode(1)) {
      return true;
    }
  }
  while (codes_used_ + 1 < num_codes && codes_[codes_used_].first == CharClass::kVirama &&
         codes_[codes_used_ + 1].first == CharClass::kConsonant) {
    ASSERT_HOST(!CodeOnlyToOutput());
    if (UseMultiCode(2)) {
      return true;
    }
    if (codes_[codes_used_].first == CharClass::kRobat) {
      if (UseMultiCode(1)) {
        return true;
      }
    }
  }
  unsigned num_matra_parts = 0;
  if (codes_[codes_used_].second == kZeroWidthJoiner ||
      codes_[codes_used_].second == kZeroWidthNonJoiner) {
    if (CodeOnlyToOutput()) {
      if (report_errors_) {
        tprintf("Unterminated joiner: 0x%x\n", output_.back());
      }
      return false;
    }
    ++num_matra_parts;
  }
  // Not quite as shown by the BNF, the matra piece is allowed as a matra on its
  // own or as an addition to other matras.
  if (codes_[codes_used_].first == CharClass::kMatra ||
      codes_[codes_used_].first == CharClass::kMatraPiece) {
    ++num_matra_parts;
    if (UseMultiCode(num_matra_parts)) {
      return true;
    }
  } else if (num_matra_parts) {
    if (report_errors_) {
      tprintf("Joiner with non-dependent vowel after it!:0x%x 0x%x\n", output_.back(),
              codes_[codes_used_].second);
    }
    return false;
  }
  if (codes_[codes_used_].first == CharClass::kMatraPiece &&
      codes_[codes_used_ - 1].first != CharClass::kMatraPiece) {
    if (UseMultiCode(1)) {
      return true;
    }
  }
  if (codes_[codes_used_].first == CharClass::kVowelModifier) {
    if (UseMultiCode(1)) {
      return true;
    }
  }
  if (codes_used_ + 1 < num_codes && codes_[codes_used_].first == CharClass::kVirama &&
      codes_[codes_used_ + 1].first == CharClass::kConsonant) {
    ASSERT_HOST(!CodeOnlyToOutput());
    if (UseMultiCode(2)) {
      return true;
    }
  }
  return true;
}

◆ UnicodeToCharClass()

Validator::CharClass tesseract::ValidateKhmer::UnicodeToCharClass ( char32 ch ) const

overrideprotectedvirtual

Implements tesseract::Validator.

Definition at line 103 of file validate_khmer.cpp.

                                                                    {
  if (IsVedicAccent(ch)) {
    return CharClass::kVedicMark;
  }
  if (ch == kZeroWidthNonJoiner) {
    return CharClass::kZeroWidthNonJoiner;
  }
  if (ch == kZeroWidthJoiner) {
    return CharClass::kZeroWidthJoiner;
  }
  // Offset from the start of the relevant unicode code block aka code page.
  int off = ch - static_cast<char32>(script_);
  // Anything in another code block is other.
  if (off < 0 || off >= kIndicCodePageSize) {
    return CharClass::kOther;
  }
  if (off <= 0x33) {
    return CharClass::kConsonant;
  }
  if (off <= 0x45) {
    return CharClass::kMatra;
  }
  if (off == 0x46) {
    return CharClass::kMatraPiece;
  }
  if (off == 0x4c) {
    return CharClass::kRobat;
  }
  if (off == 0x49 || off == 0x4a) {
    return CharClass::kNukta;
  }
  if (off <= 0x51) {
    return CharClass::kVowelModifier;
  }
  if (off == 0x52) {
    return CharClass::kVirama;
  }
  return CharClass::kOther;
}

The documentation for this class was generated from the following files:

/media/home/debian/src/github/tesseract-ocr/tesseract/src/training/unicharset/validate_khmer.h
/media/home/debian/src/github/tesseract-ocr/tesseract/src/training/unicharset/validate_khmer.cpp

Public Member Functions

Protected Member Functions