#include <dawg.h>

Inheritance diagram for tesseract::Dawg:

Public Member Functions
DawgType	type () const

const STRING &	lang () const

PermuterType	permuter () const

virtual	~Dawg ()

bool	word_in_dawg (const WERD_CHOICE &word) const
	Returns true if the given word is in the Dawg. More...

bool	prefix_in_dawg (const WERD_CHOICE &prefix, bool requires_complete) const

int	check_for_words (const char *filename, const UNICHARSET &unicharset, bool enable_wildcard) const

void	iterate_words (const UNICHARSET &unicharset, TessCallback1< const WERD_CHOICE * > *cb) const

void	iterate_words (const UNICHARSET &unicharset, TessCallback1< const char * > *cb) const

virtual EDGE_REF	edge_char_of (NODE_REF node, UNICHAR_ID unichar_id, bool word_end) const =0
	Returns the edge that corresponds to the letter out of this node. More...

virtual void	unichar_ids_of (NODE_REF node, NodeChildVector *vec, bool word_end) const =0

virtual NODE_REF	next_node (EDGE_REF edge_ref) const =0

virtual bool	end_of_word (EDGE_REF edge_ref) const =0

virtual UNICHAR_ID	edge_letter (EDGE_REF edge_ref) const =0
	Returns UNICHAR_ID stored in the edge indicated by the given EDGE_REF. More...

virtual void	print_node (NODE_REF node, int max_num_edges) const =0

virtual void	unichar_id_to_patterns (UNICHAR_ID unichar_id, const UNICHARSET &unicharset, GenericVector< UNICHAR_ID > *vec) const

virtual EDGE_REF	pattern_loop_edge (EDGE_REF edge_ref, UNICHAR_ID unichar_id, bool word_end) const

Static Public Attributes
static const inT16	kDawgMagicNumber = 42
	Magic number to determine endianness when reading the Dawg from file. More...

static const UNICHAR_ID	kPatternUnicharID = 0

Protected Member Functions
	Dawg ()

NODE_REF	next_node_from_edge_rec (const EDGE_RECORD &edge_rec) const
	Returns the next node visited by following this edge. More...

bool	marker_flag_from_edge_rec (const EDGE_RECORD &edge_rec) const
	Returns the marker flag of this edge. More...

int	direction_from_edge_rec (const EDGE_RECORD &edge_rec) const
	Returns the direction flag of this edge. More...

bool	end_of_word_from_edge_rec (const EDGE_RECORD &edge_rec) const
	Returns true if this edge marks the end of a word. More...

UNICHAR_ID	unichar_id_from_edge_rec (const EDGE_RECORD &edge_rec) const
	Returns UNICHAR_ID recorded in this edge. More...

void	set_next_node_in_edge_rec (EDGE_RECORD *edge_rec, EDGE_REF value)
	Sets the next node link for this edge in the Dawg. More...

void	set_marker_flag_in_edge_rec (EDGE_RECORD *edge_rec)
	Sets this edge record to be the last one in a sequence of edges. More...

int	given_greater_than_edge_rec (NODE_REF next_node, bool word_end, UNICHAR_ID unichar_id, const EDGE_RECORD &edge_rec) const

bool	edge_rec_match (NODE_REF next_node, bool word_end, UNICHAR_ID unichar_id, NODE_REF other_next_node, bool other_word_end, UNICHAR_ID other_unichar_id) const

void	init (DawgType type, const STRING &lang, PermuterType perm, int unicharset_size, int debug_level)

bool	match_words (WERD_CHOICE *word, inT32 index, NODE_REF node, UNICHAR_ID wildcard) const

void	iterate_words_rec (const WERD_CHOICE &word_so_far, NODE_REF to_explore, TessCallback1< const WERD_CHOICE * > *cb) const

Protected Attributes
DawgType	type_

STRING	lang_

PermuterType	perm_
	Permuter code that should be used if the word is found in this Dawg. More...

int	unicharset_size_

int	flag_start_bit_

int	next_node_start_bit_

uinT64	next_node_mask_

uinT64	flags_mask_

uinT64	letter_mask_

int	debug_level_

Detailed Description

Abstract class (an interface) that declares methods needed by the various tesseract classes to operate on SquishedDawg and Trie objects.

This class initializes all the edge masks (since their usage by SquishedDawg and Trie is identical) and implements simple accessors for each of the fields encoded in an EDGE_RECORD. This class also implements word_in_dawg() and check_for_words() (since they use only the public methods of SquishedDawg and Trie classes that are inherited from the Dawg base class).

Definition at line 118 of file dawg.h.

Constructor & Destructor Documentation

virtual tesseract::Dawg::~Dawg ( )

inlinevirtual

Definition at line 131 of file dawg.h.

131 {};

tesseract::Dawg::Dawg ( )

inlineprotected

Definition at line 197 of file dawg.h.

197 {}

Member Function Documentation

int tesseract::Dawg::check_for_words	(	const char *	filename,
		const UNICHARSET &	unicharset,
		bool	enable_wildcard
	)		const

Checks the Dawg for the words that are listed in the requested file. Returns the number of words in the given file missing from the Dawg.

Definition at line 74 of file dawg.cpp.

                                                       {
   if (filename == NULL) return 0;
 
   FILE       *word_file;
   char       string [CHARS_PER_LINE];
   int misses = 0;
   UNICHAR_ID wildcard = unicharset.unichar_to_id(kWildcard);
 
   word_file = open_file (filename, "r");
 
   while (fgets (string, CHARS_PER_LINE, word_file) != NULL) {
     chomp_string(string);  // remove newline
     WERD_CHOICE word(string, unicharset);
     if (word.length() > 0 &&
         !word.contains_unichar_id(INVALID_UNICHAR_ID)) {
       if (!match_words(&word, 0, 0,
                        enable_wildcard ? wildcard : INVALID_UNICHAR_ID)) {
         tprintf("Missing word: %s\n", string);
         ++misses;
       }
     } else {
       tprintf("Failed to create a valid word from %s\n", string);
     }
   }
   fclose (word_file);
   // Make sure the user sees this with fprintf instead of tprintf.
   if (debug_level_) tprintf("Number of lost words=%d\n", misses);
   return misses;
 }

int tesseract::Dawg::direction_from_edge_rec ( const EDGE_RECORD & edge_rec ) const

inlineprotected

Returns the direction flag of this edge.

Definition at line 208 of file dawg.h.

                                                                         {
     return ((edge_rec & (DIRECTION_FLAG << flag_start_bit_))) ?
       BACKWARD_EDGE : FORWARD_EDGE;
   }

virtual EDGE_REF tesseract::Dawg::edge_char_of	(	NODE_REF	node,
		UNICHAR_ID	unichar_id,
		bool	word_end
	)		const

pure virtual

Returns the edge that corresponds to the letter out of this node.

Implemented in tesseract::SquishedDawg, and tesseract::Trie.

virtual UNICHAR_ID tesseract::Dawg::edge_letter ( EDGE_REF edge_ref ) const

pure virtual

Returns UNICHAR_ID stored in the edge indicated by the given EDGE_REF.

Implemented in tesseract::SquishedDawg, and tesseract::Trie.

bool tesseract::Dawg::edge_rec_match	(	NODE_REF	next_node,
		bool	word_end,
		UNICHAR_ID	unichar_id,
		NODE_REF	other_next_node,
		bool	other_word_end,
		UNICHAR_ID	other_unichar_id
	)		const

inlineprotected

Returns true if all the values are equal (any value matches next_node if next_node == NO_EDGE, any value matches word_end if word_end is false).

Definition at line 259 of file dawg.h.

                                                                 {
     return ((unichar_id == other_unichar_id) &&
             (next_node == NO_EDGE || next_node == other_next_node) &&
             (!word_end || (word_end == other_word_end)));
   }

virtual bool tesseract::Dawg::end_of_word ( EDGE_REF edge_ref ) const

pure virtual

Returns true if the edge indicated by the given EDGE_REF marks the end of a word.

Implemented in tesseract::SquishedDawg, and tesseract::Trie.

bool tesseract::Dawg::end_of_word_from_edge_rec ( const EDGE_RECORD & edge_rec ) const

inlineprotected

Returns true if this edge marks the end of a word.

Definition at line 213 of file dawg.h.

                                                                            {
     return (edge_rec & (WERD_END_FLAG << flag_start_bit_)) != 0;
   }

int tesseract::Dawg::given_greater_than_edge_rec	(	NODE_REF	next_node,
		bool	word_end,
		UNICHAR_ID	unichar_id,
		const EDGE_RECORD &	edge_rec
	)		const

inlineprotected

Sequentially compares the given values of unichar ID, next node and word end marker with the values in the given EDGE_RECORD. Returns: 1 if at any step the given input value exceeds that of edge_rec (and all the values already checked are the same) 0 if edge_rec_match() returns true -1 otherwise

Definition at line 238 of file dawg.h.

                                                                             {
     UNICHAR_ID curr_unichar_id = unichar_id_from_edge_rec(edge_rec);
     NODE_REF curr_next_node = next_node_from_edge_rec(edge_rec);
     bool curr_word_end = end_of_word_from_edge_rec(edge_rec);
     if (edge_rec_match(next_node, word_end, unichar_id, curr_next_node,
                        curr_word_end, curr_unichar_id)) return 0;
     if (unichar_id > curr_unichar_id) return 1;
     if (unichar_id == curr_unichar_id) {
       if (next_node > curr_next_node) return 1;
       if (next_node == curr_next_node) {
         if (word_end > curr_word_end) return 1;
       }
     }
     return -1;
   }

void tesseract::Dawg::init	(	DawgType	type,
		const STRING &	lang,
		PermuterType	perm,
		int	unicharset_size,
		int	debug_level
	)

protected

Sets type_, lang_, perm_, unicharset_size_. Initializes the values of various masks from unicharset_size_.

Definition at line 177 of file dawg.cpp.

                                                                          {
   type_ = type;
   lang_ = lang;
   perm_ = perm;
   ASSERT_HOST(unicharset_size > 0);
   unicharset_size_ = unicharset_size;
   // Set bit masks. We will use the value unicharset_size_ as a null char, so
   // the actual number of unichars is unicharset_size_ + 1.
   flag_start_bit_ = ceil(log(unicharset_size_ + 1.0) / log(2.0));
   next_node_start_bit_ = flag_start_bit_ + NUM_FLAG_BITS;
   letter_mask_ = ~(~0ull << flag_start_bit_);
   next_node_mask_ = ~0ull << (flag_start_bit_ + NUM_FLAG_BITS);
   flags_mask_ = ~(letter_mask_ | next_node_mask_);
 
   debug_level_ = debug_level;
 }

void tesseract::Dawg::iterate_words	(	const UNICHARSET &	unicharset,
		TessCallback1< const WERD_CHOICE * > *	cb
	)		const

Definition at line 106 of file dawg.cpp.

                                                                        {
   WERD_CHOICE word(&unicharset);
   iterate_words_rec(word, 0, cb);
 }

void tesseract::Dawg::iterate_words	(	const UNICHARSET &	unicharset,
		TessCallback1< const char * > *	cb
	)		const

Definition at line 118 of file dawg.cpp.

                                                                 {
   TessCallback1<const WERD_CHOICE *> *shim =
       NewPermanentTessCallback(CallWithUTF8, cb);
   WERD_CHOICE word(&unicharset);
   iterate_words_rec(word, 0, shim);
   delete shim;
 }

void tesseract::Dawg::iterate_words_rec	(	const WERD_CHOICE &	word_so_far,
		NODE_REF	to_explore,
		TessCallback1< const WERD_CHOICE * > *	cb
	)		const

protected

Definition at line 127 of file dawg.cpp.

                                                                            {
   NodeChildVector children;
   this->unichar_ids_of(to_explore, &children, false);
   for (int i = 0; i < children.size(); i++) {
     WERD_CHOICE next_word(word_so_far);
     next_word.append_unichar_id(children[i].unichar_id, 1, 0.0, 0.0);
     if (this->end_of_word(children[i].edge_ref)) {
       cb->Run(&next_word);
     }
     NODE_REF next = next_node(children[i].edge_ref);
     if (next != 0) {
       iterate_words_rec(next_word, next, cb);
     }
   }
 }

const STRING& tesseract::Dawg::lang ( ) const

inline

Definition at line 128 of file dawg.h.

128 { return lang_; }

tesseract::Dawg::lang_

STRING lang_

Definition: dawg.h:290

bool tesseract::Dawg::marker_flag_from_edge_rec ( const EDGE_RECORD & edge_rec ) const

inlineprotected

Returns the marker flag of this edge.

Definition at line 204 of file dawg.h.

                                                                            {
     return (edge_rec & (MARKER_FLAG << flag_start_bit_)) != 0;
   }

bool tesseract::Dawg::match_words	(	WERD_CHOICE *	word,
		inT32	index,
		NODE_REF	node,
		UNICHAR_ID	wildcard
	)		const

protected

Matches all of the words that are represented by this string. If wilcard is set to something other than INVALID_UNICHAR_ID, the *'s in this string are interpreted as wildcards. WERD_CHOICE param is not passed by const so that wildcard searches can modify it and work without having to copy WERD_CHOICEs.

Definition at line 145 of file dawg.cpp.

                                                                  {
   EDGE_REF edge;
   inT32 word_end;
 
   if (wildcard != INVALID_UNICHAR_ID && word->unichar_id(index) == wildcard) {
     bool any_matched = false;
     NodeChildVector vec;
     this->unichar_ids_of(node, &vec, false);
     for (int i = 0; i < vec.size(); ++i) {
       word->set_unichar_id(vec[i].unichar_id, index);
       if (match_words(word, index, node, wildcard))
         any_matched = true;
     }
     word->set_unichar_id(wildcard, index);
     return any_matched;
   } else {
     word_end = index == word->length() - 1;
     edge = edge_char_of(node, word->unichar_id(index), word_end);
     if (edge != NO_EDGE) {  // normal edge in DAWG
       node = next_node(edge);
       if (word_end) {
         if (debug_level_ > 1) word->print("match_words() found: ");
         return true;
       } else if (node != 0) {
         return match_words(word, index+1, node, wildcard);
       }
     }
   }
   return false;
 }

virtual NODE_REF tesseract::Dawg::next_node ( EDGE_REF edge_ref ) const

pure virtual

Returns the next node visited by following the edge indicated by the given EDGE_REF.

Implemented in tesseract::SquishedDawg, and tesseract::Trie.

NODE_REF tesseract::Dawg::next_node_from_edge_rec ( const EDGE_RECORD & edge_rec ) const

inlineprotected

Returns the next node visited by following this edge.

Definition at line 200 of file dawg.h.

                                                                              {
     return ((edge_rec & next_node_mask_) >> next_node_start_bit_);
   }

virtual EDGE_REF tesseract::Dawg::pattern_loop_edge	(	EDGE_REF	edge_ref,
		UNICHAR_ID	unichar_id,
		bool	word_end
	)		const

inlinevirtual

Returns the given EDGE_REF if the EDGE_RECORD that it points to has a self loop and the given unichar_id matches the unichar_id stored in the EDGE_RECORD, returns NO_EDGE otherwise.

Reimplemented in tesseract::Trie.

Definition at line 191 of file dawg.h.

                                                                      {
     return false;
   }

PermuterType tesseract::Dawg::permuter ( ) const

inline

Definition at line 129 of file dawg.h.

129 { return perm_; }

tesseract::Dawg::perm_

PermuterType perm_

Permuter code that should be used if the word is found in this Dawg.

Definition: dawg.h:292

bool tesseract::Dawg::prefix_in_dawg	(	const WERD_CHOICE &	prefix,
		bool	requires_complete
	)		const

Definition at line 49 of file dawg.cpp.

                                                         {
   if (word.length() == 0) return !requires_complete;
   NODE_REF node = 0;
   int end_index = word.length() - 1;
   for (int i = 0; i < end_index; i++) {
     EDGE_REF edge = edge_char_of(node, word.unichar_id(i), false);
     if (edge == NO_EDGE) {
       return false;
     }
     if ((node = next_node(edge)) == 0) {
       // This only happens if all words following this edge terminate --
       // there are no larger words.  See Trie::add_word_to_dawg()
       return false;
     }
   }
   // Now check the last character.
   return edge_char_of(node, word.unichar_id(end_index), requires_complete) !=
       NO_EDGE;
 }

virtual void tesseract::Dawg::print_node	(	NODE_REF	node,
		int	max_num_edges
	)		const

pure virtual

Prints the contents of the node indicated by the given NODE_REF. At most max_num_edges will be printed.

Implemented in tesseract::SquishedDawg, and tesseract::Trie.

void tesseract::Dawg::set_marker_flag_in_edge_rec ( EDGE_RECORD * edge_rec )

inlineprotected

Sets this edge record to be the last one in a sequence of edges.

Definition at line 228 of file dawg.h.

                                                                  {
     *edge_rec |= (MARKER_FLAG << flag_start_bit_);
   }

void tesseract::Dawg::set_next_node_in_edge_rec	(	EDGE_RECORD *	edge_rec,
		EDGE_REF	value
	)

inlineprotected

Sets the next node link for this edge in the Dawg.

Definition at line 222 of file dawg.h.

                                              {
     *edge_rec &= (~next_node_mask_);
     *edge_rec |= ((value << next_node_start_bit_) & next_node_mask_);
   }

DawgType tesseract::Dawg::type ( ) const

inline

Definition at line 127 of file dawg.h.

127 { return type_; }

tesseract::Dawg::type_

DawgType type_

Definition: dawg.h:289

UNICHAR_ID tesseract::Dawg::unichar_id_from_edge_rec ( const EDGE_RECORD & edge_rec ) const

inlineprotected

Returns UNICHAR_ID recorded in this edge.

Definition at line 217 of file dawg.h.

                                          {
     return ((edge_rec & letter_mask_) >> LETTER_START_BIT);
   }

virtual void tesseract::Dawg::unichar_id_to_patterns	(	UNICHAR_ID	unichar_id,
		const UNICHARSET &	unicharset,
		GenericVector< UNICHAR_ID > *	vec
	)		const

inlinevirtual

Fills vec with unichar ids that represent the character classes of the given unichar_id.

Reimplemented in tesseract::Trie.

Definition at line 184 of file dawg.h.

186 {};

virtual void tesseract::Dawg::unichar_ids_of	(	NODE_REF	node,
		NodeChildVector *	vec,
		bool	word_end
	)		const

pure virtual

Fills the given NodeChildVector with all the unichar ids (and the corresponding EDGE_REFs) for which there is an edge out of this node.

Implemented in tesseract::SquishedDawg, and tesseract::Trie.

bool tesseract::Dawg::word_in_dawg ( const WERD_CHOICE & word ) const

Returns true if the given word is in the Dawg.

Definition at line 70 of file dawg.cpp.

                                                      {
   return prefix_in_dawg(word, true);
 }

Member Data Documentation

int tesseract::Dawg::debug_level_

protected

Definition at line 304 of file dawg.h.

int tesseract::Dawg::flag_start_bit_

protected

Definition at line 298 of file dawg.h.

uinT64 tesseract::Dawg::flags_mask_

protected

Definition at line 301 of file dawg.h.

const inT16 tesseract::Dawg::kDawgMagicNumber = 42

static

Magic number to determine endianness when reading the Dawg from file.

Definition at line 121 of file dawg.h.

const UNICHAR_ID tesseract::Dawg::kPatternUnicharID = 0

static

A special unichar id that indicates that any appropriate pattern (e.g.dicitonary word, 0-9 digit, etc) can be inserted instead Used for expressing patterns in punctuation and number Dawgs.

Definition at line 125 of file dawg.h.

STRING tesseract::Dawg::lang_

protected

Definition at line 290 of file dawg.h.

uinT64 tesseract::Dawg::letter_mask_

protected

Definition at line 302 of file dawg.h.

uinT64 tesseract::Dawg::next_node_mask_

protected

Definition at line 300 of file dawg.h.

int tesseract::Dawg::next_node_start_bit_

protected

Definition at line 299 of file dawg.h.

PermuterType tesseract::Dawg::perm_

protected

Permuter code that should be used if the word is found in this Dawg.

Definition at line 292 of file dawg.h.

DawgType tesseract::Dawg::type_

protected

Definition at line 289 of file dawg.h.

int tesseract::Dawg::unicharset_size_

protected

Definition at line 297 of file dawg.h.

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

dict/dawg.h
dict/dawg.cpp

Public Member Functions

Static Public Attributes

Protected Member Functions

Protected Attributes

Detailed Description

Constructor & Destructor Documentation

Member Function Documentation

Member Data Documentation