tesseract::LigatureTable Class Reference

#include <ligature_table.h>

Public Member Functions
string	AddLigatures (const string &str, const PangoFontInfo *font) const
string	RemoveLigatures (const string &str) const
string	RemoveCustomLigatures (const string &str) const
const LigHash &	norm_to_lig_table () const
const LigHash &	lig_to_norm_table () const

Static Public Member Functions
static LigatureTable *	Get ()

Protected Member Functions
	LigatureTable ()
void	Init ()

Protected Attributes
LigHash	norm_to_lig_table_
LigHash	lig_to_norm_table_
int	min_lig_length_
int	max_lig_length_
int	min_norm_length_
int	max_norm_length_

Static Protected Attributes
static std::unique_ptr< LigatureTable >	instance_

Detailed Description

Definition at line 38 of file ligature_table.h.

Constructor & Destructor Documentation

LigatureTable()

tesseract::LigatureTable::LigatureTable ( )

protected

Definition at line 60 of file ligature_table.cpp.

                             : min_lig_length_(0), max_lig_length_(0),
                                 min_norm_length_(0), max_norm_length_(0) {}

Member Function Documentation

AddLigatures()

string tesseract::LigatureTable::AddLigatures	(	const string &	str,
		const PangoFontInfo *	font
	)		const

Definition at line 156 of file ligature_table.cpp.

                                                                    {
  string result;
  int len = str.size();
  int step = 0;
  int i = 0;
  for (i = 0; i < len - min_norm_length_ + 1; i += step) {
    step = 0;
    for (int liglen = max_norm_length_; liglen >= min_norm_length_; --liglen) {
      if (i + liglen <= len) {
        string lig_cand = str.substr(i, liglen);
        LigHash::const_iterator it = norm_to_lig_table_.find(lig_cand);
        if (it != norm_to_lig_table_.end()) {
          tlog(3, "Considering %s -> %s\n", lig_cand.c_str(),
               it->second.c_str());
          if (font) {
            // Test for renderability.
            if (!font->CanRenderString(it->second.data(), it->second.length()))
              continue;  // Not renderable
          }
          // Found a match so convert it.
          step = liglen;
          result += it->second;
          tlog(2, "Substituted %s -> %s\n", lig_cand.c_str(),
               it->second.c_str());
          break;
        }
      }
    }
    if (step == 0) {
      result += str[i];
      step = 1;
    }
  }
  result += str.substr(i, len - i);
  return result;
}

Get()

LigatureTable * tesseract::LigatureTable::Get ( )

static

Definition at line 52 of file ligature_table.cpp.

                                  {
  if (instance_ == nullptr) {
    instance_.reset(new LigatureTable());
    instance_->Init();
  }
  return instance_.get();
}

Init()

void tesseract::LigatureTable::Init ( )

protected

Definition at line 63 of file ligature_table.cpp.

                         {
  if (norm_to_lig_table_.empty()) {
    for (char32 lig = kMinLigature; lig <= kMaxLigature; ++lig) {
      // For each char in the range, convert to utf8, nfkc normalize, and if
      // the strings are different put the both mappings in the hash_maps.
      string lig8 = EncodeAsUTF8(lig);
      icu::UnicodeString unicode_lig8(static_cast<UChar32>(lig));
      icu::UnicodeString normed8_result;
      icu::ErrorCode status;
      icu::Normalizer::normalize(unicode_lig8, UNORM_NFKC, 0, normed8_result,
                                 status);
      string normed8;
      normed8_result.toUTF8String(normed8);
      // The icu::Normalizer maps the "LONG S T" ligature to "st". Correct that
      // here manually so that AddLigatures() will work as desired.
      if (lig8 == "\uFB05")
        normed8 = "ſt";
      int lig_length = lig8.length();
      int norm_length = normed8.size();
      if (normed8 != lig8 && lig_length > 1 && norm_length > 1) {
        norm_to_lig_table_[normed8] = lig8;
        lig_to_norm_table_[lig8] = normed8;
        if (min_lig_length_ == 0 || lig_length < min_lig_length_)
          min_lig_length_ = lig_length;
        if (lig_length > max_lig_length_)
          max_lig_length_ = lig_length;
        if (min_norm_length_ == 0 || norm_length < min_norm_length_)
          min_norm_length_ = norm_length;
        if (norm_length > max_norm_length_)
          max_norm_length_ = norm_length;
      }
    }
    // Add custom extra ligatures.
    for (int i = 0; UNICHARSET::kCustomLigatures[i][0] != nullptr; ++i) {
      norm_to_lig_table_[UNICHARSET::kCustomLigatures[i][0]] =
          UNICHARSET::kCustomLigatures[i][1];
      int norm_length = strlen(UNICHARSET::kCustomLigatures[i][0]);
      if (min_norm_length_ == 0 || norm_length < min_norm_length_)
        min_norm_length_ = norm_length;
      if (norm_length > max_norm_length_)
        max_norm_length_ = norm_length;

      lig_to_norm_table_[UNICHARSET::kCustomLigatures[i][1]] =
          UNICHARSET::kCustomLigatures[i][0];
    }
  }
}

lig_to_norm_table()

const LigHash& tesseract::LigatureTable::lig_to_norm_table ( ) const

inline

Definition at line 55 of file ligature_table.h.

                                           {
    return lig_to_norm_table_;
  }

norm_to_lig_table()

const LigHash& tesseract::LigatureTable::norm_to_lig_table ( ) const

inline

Definition at line 52 of file ligature_table.h.

                                           {
    return norm_to_lig_table_;
  }

RemoveCustomLigatures()

string tesseract::LigatureTable::RemoveCustomLigatures

(

const string &

str

)

const

Definition at line 130 of file ligature_table.cpp.

                                                                   {
  string result;
  UNICHAR::const_iterator it_begin = UNICHAR::begin(str.c_str(), str.length());
  UNICHAR::const_iterator it_end = UNICHAR::end(str.c_str(), str.length());
  char tmp[5];
  int len;
  int norm_ind;
  for (UNICHAR::const_iterator it = it_begin; it != it_end; ++it) {
    len = it.get_utf8(tmp);
    tmp[len] = '\0';
    norm_ind = -1;
    for (int i = 0;
         UNICHARSET::kCustomLigatures[i][0] != nullptr && norm_ind < 0; ++i) {
      if (!strcmp(tmp, UNICHARSET::kCustomLigatures[i][1])) {
        norm_ind = i;
      }
    }
    if (norm_ind >= 0) {
      result += UNICHARSET::kCustomLigatures[norm_ind][0];
    } else {
      result += tmp;
    }
  }
  return result;
}

RemoveLigatures()

string tesseract::LigatureTable::RemoveLigatures

(

const string &

str

)

const

Definition at line 111 of file ligature_table.cpp.

                                                             {
  string result;
  UNICHAR::const_iterator it_begin = UNICHAR::begin(str.c_str(), str.length());
  UNICHAR::const_iterator it_end = UNICHAR::end(str.c_str(), str.length());
  char tmp[5];
  int len;
  for (UNICHAR::const_iterator it = it_begin; it != it_end; ++it) {
    len = it.get_utf8(tmp);
    tmp[len] = '\0';
    LigHash::const_iterator lig_it = lig_to_norm_table_.find(tmp);
    if (lig_it != lig_to_norm_table_.end()) {
      result += lig_it->second;
    } else {
      result += tmp;
    }
  }
  return result;
}

Member Data Documentation

instance_

std::unique_ptr< LigatureTable > tesseract::LigatureTable::instance_

staticprotected

Definition at line 65 of file ligature_table.h.

lig_to_norm_table_

LigHash tesseract::LigatureTable::lig_to_norm_table_

protected

Definition at line 67 of file ligature_table.h.

max_lig_length_

int tesseract::LigatureTable::max_lig_length_

protected

Definition at line 69 of file ligature_table.h.

max_norm_length_

int tesseract::LigatureTable::max_norm_length_

protected

Definition at line 71 of file ligature_table.h.

min_lig_length_

int tesseract::LigatureTable::min_lig_length_

protected

Definition at line 68 of file ligature_table.h.

min_norm_length_

int tesseract::LigatureTable::min_norm_length_

protected

Definition at line 70 of file ligature_table.h.

norm_to_lig_table_

LigHash tesseract::LigatureTable::norm_to_lig_table_

protected

Definition at line 66 of file ligature_table.h.

---

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

• /home/stefan/src/github/tesseract-ocr/tesseract/training/ligature_table.h
• /home/stefan/src/github/tesseract-ocr/tesseract/training/ligature_table.cpp