michael@0: /* -*- Mode: C; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
michael@0: /* This Source Code Form is subject to the terms of the Mozilla Public
michael@0:  * License, v. 2.0. If a copy of the MPL was not distributed with this
michael@0:  * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
michael@0: 
michael@0: #ifndef nsHebrewProber_h__
michael@0: #define nsHebrewProber_h__
michael@0: 
michael@0: #include "nsSBCharSetProber.h"
michael@0: 
michael@0: // This prober doesn't actually recognize a language or a charset.
michael@0: // It is a helper prober for the use of the Hebrew model probers
michael@0: class nsHebrewProber: public nsCharSetProber
michael@0: {
michael@0: public:
michael@0:   nsHebrewProber(void) :mLogicalProb(0), mVisualProb(0) { Reset(); }
michael@0: 
michael@0:   virtual ~nsHebrewProber(void) {}
michael@0:   virtual nsProbingState HandleData(const char* aBuf, uint32_t aLen);
michael@0:   virtual const char* GetCharSetName();
michael@0:   virtual void Reset(void);
michael@0: 
michael@0:   virtual nsProbingState GetState(void);
michael@0: 
michael@0:   virtual float     GetConfidence(void) { return (float)0.0; }
michael@0: 
michael@0:   void SetModelProbers(nsCharSetProber *logicalPrb, nsCharSetProber *visualPrb) 
michael@0:   { mLogicalProb = logicalPrb; mVisualProb = visualPrb; }
michael@0: 
michael@0: #ifdef DEBUG_chardet
michael@0:   virtual void  DumpStatus();
michael@0: #endif
michael@0: 
michael@0: protected:
michael@0:   static bool isFinal(char c);
michael@0:   static bool isNonFinal(char c);
michael@0: 
michael@0:   int32_t mFinalCharLogicalScore, mFinalCharVisualScore;
michael@0: 
michael@0:   // The two last characters seen in the previous buffer.
michael@0:   char mPrev, mBeforePrev;
michael@0: 
michael@0:   // These probers are owned by the group prober.
michael@0:   nsCharSetProber *mLogicalProb, *mVisualProb;
michael@0: };
michael@0: 
michael@0: /**
michael@0:  * ** General ideas of the Hebrew charset recognition **
michael@0:  *
michael@0:  * Four main charsets exist in Hebrew:
michael@0:  * "ISO-8859-8" - Visual Hebrew
michael@0:  * "windows-1255" - Logical Hebrew 
michael@0:  * "ISO-8859-8-I" - Logical Hebrew
michael@0:  * "x-mac-hebrew" - ?? Logical Hebrew ??
michael@0:  *
michael@0:  * Both "ISO" charsets use a completely identical set of code points, whereas
michael@0:  * "windows-1255" and "x-mac-hebrew" are two different proper supersets of 
michael@0:  * these code points. windows-1255 defines additional characters in the range
michael@0:  * 0x80-0x9F as some misc punctuation marks as well as some Hebrew-specific 
michael@0:  * diacritics and additional 'Yiddish' ligature letters in the range 0xc0-0xd6.
michael@0:  * x-mac-hebrew defines similar additional code points but with a different 
michael@0:  * mapping.
michael@0:  *
michael@0:  * As far as an average Hebrew text with no diacritics is concerned, all four 
michael@0:  * charsets are identical with respect to code points. Meaning that for the 
michael@0:  * main Hebrew alphabet, all four map the same values to all 27 Hebrew letters 
michael@0:  * (including final letters).
michael@0:  *
michael@0:  * The dominant difference between these charsets is their directionality.
michael@0:  * "Visual" directionality means that the text is ordered as if the renderer is
michael@0:  * not aware of a BIDI rendering algorithm. The renderer sees the text and 
michael@0:  * draws it from left to right. The text itself when ordered naturally is read 
michael@0:  * backwards. A buffer of Visual Hebrew generally looks like so:
michael@0:  * "[last word of first line spelled backwards] [whole line ordered backwards
michael@0:  * and spelled backwards] [first word of first line spelled backwards] 
michael@0:  * [end of line] [last word of second line] ... etc' "
michael@0:  * adding punctuation marks, numbers and English text to visual text is
michael@0:  * naturally also "visual" and from left to right.
michael@0:  * 
michael@0:  * "Logical" directionality means the text is ordered "naturally" according to
michael@0:  * the order it is read. It is the responsibility of the renderer to display 
michael@0:  * the text from right to left. A BIDI algorithm is used to place general 
michael@0:  * punctuation marks, numbers and English text in the text.
michael@0:  *
michael@0:  * Texts in x-mac-hebrew are almost impossible to find on the Internet. From 
michael@0:  * what little evidence I could find, it seems that its general directionality
michael@0:  * is Logical.
michael@0:  *
michael@0:  * To sum up all of the above, the Hebrew probing mechanism knows about two
michael@0:  * charsets:
michael@0:  * Visual Hebrew - "ISO-8859-8" - backwards text - Words and sentences are
michael@0:  *    backwards while line order is natural. For charset recognition purposes
michael@0:  *    the line order is unimportant (In fact, for this implementation, even 
michael@0:  *    word order is unimportant).
michael@0:  * Logical Hebrew - "windows-1255" - normal, naturally ordered text.
michael@0:  *
michael@0:  * "ISO-8859-8-I" is a subset of windows-1255 and doesn't need to be 
michael@0:  *    specifically identified.
michael@0:  * "x-mac-hebrew" is also identified as windows-1255. A text in x-mac-hebrew
michael@0:  *    that contain special punctuation marks or diacritics is displayed with
michael@0:  *    some unconverted characters showing as question marks. This problem might
michael@0:  *    be corrected using another model prober for x-mac-hebrew. Due to the fact
michael@0:  *    that x-mac-hebrew texts are so rare, writing another model prober isn't 
michael@0:  *    worth the effort and performance hit.
michael@0:  *
michael@0:  * *** The Prober ***
michael@0:  *
michael@0:  * The prober is divided between two nsSBCharSetProbers and an nsHebrewProber,
michael@0:  * all of which are managed, created, fed data, inquired and deleted by the
michael@0:  * nsSBCSGroupProber. The two nsSBCharSetProbers identify that the text is in
michael@0:  * fact some kind of Hebrew, Logical or Visual. The final decision about which
michael@0:  * one is it is made by the nsHebrewProber by combining final-letter scores
michael@0:  * with the scores of the two nsSBCharSetProbers to produce a final answer.
michael@0:  *
michael@0:  * The nsSBCSGroupProber is responsible for stripping the original text of HTML
michael@0:  * tags, English characters, numbers, low-ASCII punctuation characters, spaces
michael@0:  * and new lines. It reduces any sequence of such characters to a single space.
michael@0:  * The buffer fed to each prober in the SBCS group prober is pure text in
michael@0:  * high-ASCII.
michael@0:  * The two nsSBCharSetProbers (model probers) share the same language model:
michael@0:  * Win1255Model.
michael@0:  * The first nsSBCharSetProber uses the model normally as any other
michael@0:  * nsSBCharSetProber does, to recognize windows-1255, upon which this model was
michael@0:  * built. The second nsSBCharSetProber is told to make the pair-of-letter
michael@0:  * lookup in the language model backwards. This in practice exactly simulates
michael@0:  * a visual Hebrew model using the windows-1255 logical Hebrew model.
michael@0:  *
michael@0:  * The nsHebrewProber is not using any language model. All it does is look for
michael@0:  * final-letter evidence suggesting the text is either logical Hebrew or visual
michael@0:  * Hebrew. Disjointed from the model probers, the results of the nsHebrewProber
michael@0:  * alone are meaningless. nsHebrewProber always returns 0.00 as confidence
michael@0:  * since it never identifies a charset by itself. Instead, the pointer to the
michael@0:  * nsHebrewProber is passed to the model probers as a helper "Name Prober".
michael@0:  * When the Group prober receives a positive identification from any prober,
michael@0:  * it asks for the name of the charset identified. If the prober queried is a
michael@0:  * Hebrew model prober, the model prober forwards the call to the
michael@0:  * nsHebrewProber to make the final decision. In the nsHebrewProber, the
michael@0:  * decision is made according to the final-letters scores maintained and Both
michael@0:  * model probers scores. The answer is returned in the form of the name of the
michael@0:  * charset identified, either "windows-1255" or "ISO-8859-8".
michael@0:  *
michael@0:  */
michael@0: #endif /* nsHebrewProber_h__ */