michael@0: /* -*- Mode: C; tab-width: 2; indent-tabs-mode: nil; c-basic-offset: 2 -*-
michael@0:  *
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 nsBidi_h__
michael@0: #define nsBidi_h__
michael@0: 
michael@0: #include "nsBidiUtils.h"
michael@0: 
michael@0: // Bidi reordering engine from ICU
michael@0: /*
michael@0:  * javadoc-style comments are intended to be transformed into HTML
michael@0:  * using DOC++ - see
michael@0:  * http://www.zib.de/Visual/software/doc++/index.html .
michael@0:  *
michael@0:  * The HTML documentation is created with
michael@0:  *  doc++ -H nsIBidi.h
michael@0:  */
michael@0: 
michael@0: /**
michael@0:  * @mainpage BIDI algorithm for Mozilla (from ICU)
michael@0:  *
michael@0:  * <h2>BIDI algorithm for Mozilla</h2>
michael@0:  *
michael@0:  * This is an implementation of the Unicode Bidirectional algorithm.
michael@0:  * The algorithm is defined in the
michael@0:  * <a href="http://www.unicode.org/unicode/reports/tr9/">Unicode Technical Report 9</a>,
michael@0:  * version 5, also described in The Unicode Standard, Version 3.0 .<p>
michael@0:  *
michael@0:  * <h3>General remarks about the API:</h3>
michael@0:  *
michael@0:  * The <quote>limit</quote> of a sequence of characters is the position just after their
michael@0:  * last character, i.e., one more than that position.<p>
michael@0:  *
michael@0:  * Some of the API functions provide access to <quote>runs</quote>.
michael@0:  * Such a <quote>run</quote> is defined as a sequence of characters
michael@0:  * that are at the same embedding level
michael@0:  * after performing the BIDI algorithm.<p>
michael@0:  *
michael@0:  * @author Markus W. Scherer. Ported to Mozilla by Simon Montagu
michael@0:  * @version 1.0
michael@0:  */
michael@0: 
michael@0: /**
michael@0:  * nsBidiLevel is the type of the level values in this
michael@0:  * Bidi implementation.
michael@0:  * It holds an embedding level and indicates the visual direction
michael@0:  * by its bit 0 (even/odd value).<p>
michael@0:  *
michael@0:  * It can also hold non-level values for the
michael@0:  * <code>aParaLevel</code> and <code>aEmbeddingLevels</code>
michael@0:  * arguments of <code>SetPara</code>; there:
michael@0:  * <ul>
michael@0:  * <li>bit 7 of an <code>aEmbeddingLevels[]</code>
michael@0:  * value indicates whether the using application is
michael@0:  * specifying the level of a character to <i>override</i> whatever the
michael@0:  * Bidi implementation would resolve it to.</li>
michael@0:  * <li><code>aParaLevel</code> can be set to the
michael@0:  * pseudo-level values <code>NSBIDI_DEFAULT_LTR</code>
michael@0:  * and <code>NSBIDI_DEFAULT_RTL</code>.</li></ul>
michael@0:  *
michael@0:  * @see nsIBidi::SetPara
michael@0:  *
michael@0:  * <p>The related constants are not real, valid level values.
michael@0:  * <code>NSBIDI_DEFAULT_XXX</code> can be used to specify
michael@0:  * a default for the paragraph level for
michael@0:  * when the <code>SetPara</code> function
michael@0:  * shall determine it but there is no
michael@0:  * strongly typed character in the input.<p>
michael@0:  *
michael@0:  * Note that the value for <code>NSBIDI_DEFAULT_LTR</code> is even
michael@0:  * and the one for <code>NSBIDI_DEFAULT_RTL</code> is odd,
michael@0:  * just like with normal LTR and RTL level values -
michael@0:  * these special values are designed that way. Also, the implementation
michael@0:  * assumes that NSBIDI_MAX_EXPLICIT_LEVEL is odd.
michael@0:  *
michael@0:  * @see NSBIDI_DEFAULT_LTR
michael@0:  * @see NSBIDI_DEFAULT_RTL
michael@0:  * @see NSBIDI_LEVEL_OVERRIDE
michael@0:  * @see NSBIDI_MAX_EXPLICIT_LEVEL
michael@0:  */
michael@0: typedef uint8_t nsBidiLevel;
michael@0: 
michael@0: /** Paragraph level setting.
michael@0:  *  If there is no strong character, then set the paragraph level to 0 (left-to-right).
michael@0:  */
michael@0: #define NSBIDI_DEFAULT_LTR 0xfe
michael@0: 
michael@0: /** Paragraph level setting.
michael@0:  *  If there is no strong character, then set the paragraph level to 1 (right-to-left).
michael@0:  */
michael@0: #define NSBIDI_DEFAULT_RTL 0xff
michael@0: 
michael@0: /**
michael@0:  * Maximum explicit embedding level.
michael@0:  * (The maximum resolved level can be up to <code>NSBIDI_MAX_EXPLICIT_LEVEL+1</code>).
michael@0:  *
michael@0:  */
michael@0: #define NSBIDI_MAX_EXPLICIT_LEVEL 61
michael@0: 
michael@0: /** Bit flag for level input. 
michael@0:  *  Overrides directional properties. 
michael@0:  */
michael@0: #define NSBIDI_LEVEL_OVERRIDE 0x80
michael@0: 
michael@0: /**
michael@0:  * <code>nsBidiDirection</code> values indicate the text direction.
michael@0:  */
michael@0: enum nsBidiDirection {
michael@0:   /** All left-to-right text This is a 0 value. */
michael@0:   NSBIDI_LTR,
michael@0:   /** All right-to-left text This is a 1 value. */
michael@0:   NSBIDI_RTL,
michael@0:   /** Mixed-directional text. */
michael@0:   NSBIDI_MIXED
michael@0: };
michael@0: 
michael@0: typedef enum nsBidiDirection nsBidiDirection;
michael@0: 
michael@0: /* miscellaneous definitions ------------------------------------------------ */
michael@0: /** option flags for WriteReverse() */
michael@0: /**
michael@0:  * option bit for WriteReverse():
michael@0:  * keep combining characters after their base characters in RTL runs
michael@0:  *
michael@0:  * @see WriteReverse
michael@0:  */
michael@0: #define NSBIDI_KEEP_BASE_COMBINING       1
michael@0: 
michael@0: /**
michael@0:  * option bit for WriteReverse():
michael@0:  * replace characters with the "mirrored" property in RTL runs
michael@0:  * by their mirror-image mappings
michael@0:  *
michael@0:  * @see WriteReverse
michael@0:  */
michael@0: #define NSBIDI_DO_MIRRORING              2
michael@0: 
michael@0: /**
michael@0:  * option bit for WriteReverse():
michael@0:  * remove Bidi control characters
michael@0:  *
michael@0:  * @see WriteReverse
michael@0:  */
michael@0: #define NSBIDI_REMOVE_BIDI_CONTROLS      8
michael@0: 
michael@0: /* helper macros for each allocated array member */
michael@0: #define GETDIRPROPSMEMORY(length) \
michael@0:                                   GetMemory((void **)&mDirPropsMemory, &mDirPropsSize, \
michael@0:                                   mMayAllocateText, (length))
michael@0: 
michael@0: #define GETLEVELSMEMORY(length) \
michael@0:                                 GetMemory((void **)&mLevelsMemory, &mLevelsSize, \
michael@0:                                 mMayAllocateText, (length))
michael@0: 
michael@0: #define GETRUNSMEMORY(length) \
michael@0:                               GetMemory((void **)&mRunsMemory, &mRunsSize, \
michael@0:                               mMayAllocateRuns, (length)*sizeof(Run))
michael@0: 
michael@0: /* additional macros used by constructor - always allow allocation */
michael@0: #define GETINITIALDIRPROPSMEMORY(length) \
michael@0:                                          GetMemory((void **)&mDirPropsMemory, &mDirPropsSize, \
michael@0:                                          true, (length))
michael@0: 
michael@0: #define GETINITIALLEVELSMEMORY(length) \
michael@0:                                        GetMemory((void **)&mLevelsMemory, &mLevelsSize, \
michael@0:                                        true, (length))
michael@0: 
michael@0: #define GETINITIALRUNSMEMORY(length) \
michael@0:                                      GetMemory((void **)&mRunsMemory, &mRunsSize, \
michael@0:                                      true, (length)*sizeof(Run))
michael@0: 
michael@0: /*
michael@0:  * Sometimes, bit values are more appropriate
michael@0:  * to deal with directionality properties.
michael@0:  * Abbreviations in these macro names refer to names
michael@0:  * used in the Bidi algorithm.
michael@0:  */
michael@0: typedef uint8_t DirProp;
michael@0: 
michael@0: #define DIRPROP_FLAG(dir) (1UL<<(dir))
michael@0: 
michael@0: /* special flag for multiple runs from explicit embedding codes */
michael@0: #define DIRPROP_FLAG_MULTI_RUNS (1UL<<31)
michael@0: 
michael@0: /* are there any characters that are LTR or RTL? */
michael@0: #define MASK_LTR (DIRPROP_FLAG(L)|DIRPROP_FLAG(EN)|DIRPROP_FLAG(AN)|DIRPROP_FLAG(LRE)|DIRPROP_FLAG(LRO))
michael@0: #define MASK_RTL (DIRPROP_FLAG(R)|DIRPROP_FLAG(AL)|DIRPROP_FLAG(RLE)|DIRPROP_FLAG(RLO))
michael@0: 
michael@0: /* explicit embedding codes */
michael@0: #define MASK_LRX (DIRPROP_FLAG(LRE)|DIRPROP_FLAG(LRO))
michael@0: #define MASK_RLX (DIRPROP_FLAG(RLE)|DIRPROP_FLAG(RLO))
michael@0: #define MASK_OVERRIDE (DIRPROP_FLAG(LRO)|DIRPROP_FLAG(RLO))
michael@0: 
michael@0: #define MASK_EXPLICIT (MASK_LRX|MASK_RLX|DIRPROP_FLAG(PDF))
michael@0: #define MASK_BN_EXPLICIT (DIRPROP_FLAG(BN)|MASK_EXPLICIT)
michael@0: 
michael@0: /* paragraph and segment separators */
michael@0: #define MASK_B_S (DIRPROP_FLAG(B)|DIRPROP_FLAG(S))
michael@0: 
michael@0: /* all types that are counted as White Space or Neutral in some steps */
michael@0: #define MASK_WS (MASK_B_S|DIRPROP_FLAG(WS)|MASK_BN_EXPLICIT)
michael@0: #define MASK_N (DIRPROP_FLAG(O_N)|MASK_WS)
michael@0: 
michael@0: /* all types that are included in a sequence of European Terminators for (W5) */
michael@0: #define MASK_ET_NSM_BN (DIRPROP_FLAG(ET)|DIRPROP_FLAG(NSM)|MASK_BN_EXPLICIT)
michael@0: 
michael@0: /* types that are neutrals or could becomes neutrals in (Wn) */
michael@0: #define MASK_POSSIBLE_N (DIRPROP_FLAG(CS)|DIRPROP_FLAG(ES)|DIRPROP_FLAG(ET)|MASK_N)
michael@0: 
michael@0: /*
michael@0:  * These types may be changed to "e",
michael@0:  * the embedding type (L or R) of the run,
michael@0:  * in the Bidi algorithm (N2)
michael@0:  */
michael@0: #define MASK_EMBEDDING (DIRPROP_FLAG(NSM)|MASK_POSSIBLE_N)
michael@0: 
michael@0: /* the dirProp's L and R are defined to 0 and 1 values in nsCharType */
michael@0: #define GET_LR_FROM_LEVEL(level) ((DirProp)((level)&1))
michael@0: 
michael@0: #define IS_DEFAULT_LEVEL(level) (((level)&0xfe)==0xfe)
michael@0: 
michael@0: /* handle surrogate pairs --------------------------------------------------- */
michael@0: 
michael@0: #define IS_FIRST_SURROGATE(uchar) (((uchar)&0xfc00)==0xd800)
michael@0: #define IS_SECOND_SURROGATE(uchar) (((uchar)&0xfc00)==0xdc00)
michael@0: 
michael@0: /* get the UTF-32 value directly from the surrogate pseudo-characters */
michael@0: #define SURROGATE_OFFSET ((0xd800<<10UL)+0xdc00-0x10000)
michael@0: #define GET_UTF_32(first, second) (((first)<<10UL)+(second)-SURROGATE_OFFSET)
michael@0: 
michael@0: 
michael@0: #define UTF_ERROR_VALUE 0xffff
michael@0: /* definitions with forward iteration --------------------------------------- */
michael@0: 
michael@0: /*
michael@0:  * all the macros that go forward assume that
michael@0:  * the initial offset is 0<=i<length;
michael@0:  * they update the offset
michael@0:  */
michael@0: 
michael@0: /* fast versions, no error-checking */
michael@0: 
michael@0: #define UTF16_APPEND_CHAR_UNSAFE(s, i, c){ \
michael@0:                                          if((uint32_t)(c)<=0xffff) { \
michael@0:                                          (s)[(i)++]=(char16_t)(c); \
michael@0:                                          } else { \
michael@0:                                          (s)[(i)++]=(char16_t)((c)>>10)+0xd7c0; \
michael@0:                                          (s)[(i)++]=(char16_t)(c)&0x3ff|0xdc00; \
michael@0:                                          } \
michael@0: }
michael@0: 
michael@0: /* safe versions with error-checking and optional regularity-checking */
michael@0: 
michael@0: #define UTF16_APPEND_CHAR_SAFE(s, i, length, c) { \
michael@0:                                                 if((PRUInt32)(c)<=0xffff) { \
michael@0:                                                 (s)[(i)++]=(char16_t)(c); \
michael@0:                                                 } else if((PRUInt32)(c)<=0x10ffff) { \
michael@0:                                                 if((i)+1<(length)) { \
michael@0:                                                 (s)[(i)++]=(char16_t)((c)>>10)+0xd7c0; \
michael@0:                                                 (s)[(i)++]=(char16_t)(c)&0x3ff|0xdc00; \
michael@0:                                                 } else /* not enough space */ { \
michael@0:                                                 (s)[(i)++]=UTF_ERROR_VALUE; \
michael@0:                                                 } \
michael@0:                                                 } else /* c>0x10ffff, write error value */ { \
michael@0:                                                 (s)[(i)++]=UTF_ERROR_VALUE; \
michael@0:                                                 } \
michael@0: }
michael@0: 
michael@0: /* definitions with backward iteration -------------------------------------- */
michael@0: 
michael@0: /*
michael@0:  * all the macros that go backward assume that
michael@0:  * the valid buffer range starts at offset 0
michael@0:  * and that the initial offset is 0<i<=length;
michael@0:  * they update the offset
michael@0:  */
michael@0: 
michael@0: /* fast versions, no error-checking */
michael@0: 
michael@0: /*
michael@0:  * Get a single code point from an offset that points behind the last
michael@0:  * of the code units that belong to that code point.
michael@0:  * Assume 0<=i<length.
michael@0:  */
michael@0: #define UTF16_PREV_CHAR_UNSAFE(s, i, c) { \
michael@0:                                         (c)=(s)[--(i)]; \
michael@0:                                         if(IS_SECOND_SURROGATE(c)) { \
michael@0:                                         (c)=GET_UTF_32((s)[--(i)], (c)); \
michael@0:                                         } \
michael@0: }
michael@0: 
michael@0: #define UTF16_BACK_1_UNSAFE(s, i) { \
michael@0:                                   if(IS_SECOND_SURROGATE((s)[--(i)])) { \
michael@0:                                   --(i); \
michael@0:                                   } \
michael@0: }
michael@0: 
michael@0: #define UTF16_BACK_N_UNSAFE(s, i, n) { \
michael@0:                                      int32_t __N=(n); \
michael@0:                                      while(__N>0) { \
michael@0:                                      UTF16_BACK_1_UNSAFE(s, i); \
michael@0:                                      --__N; \
michael@0:                                      } \
michael@0: }
michael@0: 
michael@0: /* safe versions with error-checking and optional regularity-checking */
michael@0: 
michael@0: #define UTF16_PREV_CHAR_SAFE(s, start, i, c, strict) { \
michael@0:                                                      (c)=(s)[--(i)]; \
michael@0:                                                      if(IS_SECOND_SURROGATE(c)) { \
michael@0:                                                      char16_t __c2; \
michael@0:                                                      if((i)>(start) && IS_FIRST_SURROGATE(__c2=(s)[(i)-1])) { \
michael@0:                                                      --(i); \
michael@0:                                                      (c)=GET_UTF_32(__c2, (c)); \
michael@0:       /* strict: ((c)&0xfffe)==0xfffe is caught by UTF_IS_ERROR() */ \
michael@0:                                                      } else if(strict) {\
michael@0:       /* unmatched second surrogate */ \
michael@0:                                                      (c)=UTF_ERROR_VALUE; \
michael@0:                                                      } \
michael@0:                                                      } else if(strict && IS_FIRST_SURROGATE(c)) { \
michael@0:       /* unmatched first surrogate */ \
michael@0:                                                      (c)=UTF_ERROR_VALUE; \
michael@0:   /* else strict: (c)==0xfffe is caught by UTF_IS_ERROR() */ \
michael@0:                                                      } \
michael@0: }
michael@0: 
michael@0: #define UTF16_BACK_1_SAFE(s, start, i) { \
michael@0:                                        if(IS_SECOND_SURROGATE((s)[--(i)]) && (i)>(start) && IS_FIRST_SURROGATE((s)[(i)-1])) { \
michael@0:                                        --(i); \
michael@0:                                        } \
michael@0: }
michael@0: 
michael@0: #define UTF16_BACK_N_SAFE(s, start, i, n) { \
michael@0:                                           int32_t __N=(n); \
michael@0:                                           while(__N>0 && (i)>(start)) { \
michael@0:                                           UTF16_BACK_1_SAFE(s, start, i); \
michael@0:                                           --__N; \
michael@0:                                           } \
michael@0: }
michael@0: 
michael@0: #define UTF_PREV_CHAR_UNSAFE(s, i, c)                UTF16_PREV_CHAR_UNSAFE(s, i, c)
michael@0: #define UTF_PREV_CHAR_SAFE(s, start, i, c, strict)   UTF16_PREV_CHAR_SAFE(s, start, i, c, strict)
michael@0: #define UTF_BACK_1_UNSAFE(s, i)                      UTF16_BACK_1_UNSAFE(s, i)
michael@0: #define UTF_BACK_1_SAFE(s, start, i)                 UTF16_BACK_1_SAFE(s, start, i)
michael@0: #define UTF_BACK_N_UNSAFE(s, i, n)                   UTF16_BACK_N_UNSAFE(s, i, n)
michael@0: #define UTF_BACK_N_SAFE(s, start, i, n)              UTF16_BACK_N_SAFE(s, start, i, n)
michael@0: #define UTF_APPEND_CHAR_UNSAFE(s, i, c)              UTF16_APPEND_CHAR_UNSAFE(s, i, c)
michael@0: #define UTF_APPEND_CHAR_SAFE(s, i, length, c)        UTF16_APPEND_CHAR_SAFE(s, i, length, c)
michael@0: 
michael@0: #define UTF_PREV_CHAR(s, start, i, c)                UTF_PREV_CHAR_SAFE(s, start, i, c, false)
michael@0: #define UTF_BACK_1(s, start, i)                      UTF_BACK_1_SAFE(s, start, i)
michael@0: #define UTF_BACK_N(s, start, i, n)                   UTF_BACK_N_SAFE(s, start, i, n)
michael@0: #define UTF_APPEND_CHAR(s, i, length, c)             UTF_APPEND_CHAR_SAFE(s, i, length, c)
michael@0: 
michael@0: /* Run structure for reordering --------------------------------------------- */
michael@0: 
michael@0: typedef struct Run {
michael@0:   int32_t logicalStart,  /* first character of the run; b31 indicates even/odd level */
michael@0:   visualLimit;  /* last visual position of the run +1 */
michael@0: } Run;
michael@0: 
michael@0: /* in a Run, logicalStart will get this bit set if the run level is odd */
michael@0: #define INDEX_ODD_BIT (1UL<<31)
michael@0: 
michael@0: #define MAKE_INDEX_ODD_PAIR(index, level) (index|((uint32_t)level<<31))
michael@0: #define ADD_ODD_BIT_FROM_LEVEL(x, level)  ((x)|=((uint32_t)level<<31))
michael@0: #define REMOVE_ODD_BIT(x)          ((x)&=~INDEX_ODD_BIT)
michael@0: 
michael@0: #define GET_INDEX(x)   (x&~INDEX_ODD_BIT)
michael@0: #define GET_ODD_BIT(x) ((uint32_t)x>>31)
michael@0: #define IS_ODD_RUN(x)  ((x&INDEX_ODD_BIT)!=0)
michael@0: #define IS_EVEN_RUN(x) ((x&INDEX_ODD_BIT)==0)
michael@0: 
michael@0: typedef uint32_t Flags;
michael@0: 
michael@0: /**
michael@0:  * This class holds information about a paragraph of text
michael@0:  * with Bidi-algorithm-related details, or about one line of
michael@0:  * such a paragraph.<p>
michael@0:  * Reordering can be done on a line, or on a paragraph which is
michael@0:  * then interpreted as one single line.<p>
michael@0:  *
michael@0:  * On construction, the class is initially empty. It is assigned
michael@0:  * the Bidi properties of a paragraph by <code>SetPara</code>
michael@0:  * or the Bidi properties of a line of a paragraph by
michael@0:  * <code>SetLine</code>.<p>
michael@0:  * A Bidi class can be reused for as long as it is not deallocated
michael@0:  * by calling its destructor.<p>
michael@0:  * <code>SetPara</code> will allocate additional memory for
michael@0:  * internal structures as necessary.
michael@0:  */
michael@0: class nsBidi
michael@0: {
michael@0: public: 
michael@0:   /** @brief Default constructor.
michael@0:    * 
michael@0:    * The nsBidi object is initially empty. It is assigned
michael@0:    * the Bidi properties of a paragraph by <code>SetPara()</code>
michael@0:    * or the Bidi properties of a line of a paragraph by
michael@0:    * <code>GetLine()</code>.<p>
michael@0:    * This object can be reused for as long as it is not destroyed.<p>
michael@0:    * <code>SetPara()</code> will allocate additional memory for
michael@0:    * internal structures as necessary.
michael@0:    *
michael@0:    */
michael@0:   nsBidi();
michael@0: 
michael@0:   /** @brief Destructor. */
michael@0:   virtual ~nsBidi();
michael@0: 
michael@0: 
michael@0:   /**
michael@0:    * Perform the Unicode Bidi algorithm. It is defined in the
michael@0:    * <a href="http://www.unicode.org/unicode/reports/tr9/">Unicode Technical Report 9</a>,
michael@0:    * version 5,
michael@0:    * also described in The Unicode Standard, Version 3.0 .<p>
michael@0:    *
michael@0:    * This function takes a single plain text paragraph with or without
michael@0:    * externally specified embedding levels from <quote>styled</quote> text
michael@0:    * and computes the left-right-directionality of each character.<p>
michael@0:    *
michael@0:    * If the entire paragraph consists of text of only one direction, then
michael@0:    * the function may not perform all the steps described by the algorithm,
michael@0:    * i.e., some levels may not be the same as if all steps were performed.
michael@0:    * This is not relevant for unidirectional text.<br>
michael@0:    * For example, in pure LTR text with numbers the numbers would get
michael@0:    * a resolved level of 2 higher than the surrounding text according to
michael@0:    * the algorithm. This implementation may set all resolved levels to
michael@0:    * the same value in such a case.<p>
michael@0:    *
michael@0:    * The text must be externally split into separate paragraphs (rule P1).
michael@0:    * Paragraph separators (B) should appear at most at the very end.
michael@0:    *
michael@0:    * @param aText is a pointer to the single-paragraph text that the
michael@0:    *      Bidi algorithm will be performed on
michael@0:    *      (step (P1) of the algorithm is performed externally).
michael@0:    *      <strong>The text must be (at least) <code>aLength</code> long.</strong>
michael@0:    *
michael@0:    * @param aLength is the length of the text; if <code>aLength==-1</code> then
michael@0:    *      the text must be zero-terminated.
michael@0:    *
michael@0:    * @param aParaLevel specifies the default level for the paragraph;
michael@0:    *      it is typically 0 (LTR) or 1 (RTL).
michael@0:    *      If the function shall determine the paragraph level from the text,
michael@0:    *      then <code>aParaLevel</code> can be set to
michael@0:    *      either <code>NSBIDI_DEFAULT_LTR</code>
michael@0:    *      or <code>NSBIDI_DEFAULT_RTL</code>;
michael@0:    *      if there is no strongly typed character, then
michael@0:    *      the desired default is used (0 for LTR or 1 for RTL).
michael@0:    *      Any other value between 0 and <code>NSBIDI_MAX_EXPLICIT_LEVEL</code> is also valid,
michael@0:    *      with odd levels indicating RTL.
michael@0:    *
michael@0:    * @param aEmbeddingLevels (in) may be used to preset the embedding and override levels,
michael@0:    *      ignoring characters like LRE and PDF in the text.
michael@0:    *      A level overrides the directional property of its corresponding
michael@0:    *      (same index) character if the level has the
michael@0:    *      <code>NSBIDI_LEVEL_OVERRIDE</code> bit set.<p>
michael@0:    *      Except for that bit, it must be
michael@0:    *      <code>aParaLevel<=aEmbeddingLevels[]<=NSBIDI_MAX_EXPLICIT_LEVEL</code>.<p>
michael@0:    *      <strong>Caution: </strong>A copy of this pointer, not of the levels,
michael@0:    *      will be stored in the <code>nsBidi</code> object;
michael@0:    *      the <code>aEmbeddingLevels</code> array must not be
michael@0:    *      deallocated before the <code>nsBidi</code> object is destroyed or reused,
michael@0:    *      and the <code>aEmbeddingLevels</code>
michael@0:    *      should not be modified to avoid unexpected results on subsequent Bidi operations.
michael@0:    *      However, the <code>SetPara</code> and
michael@0:    *      <code>SetLine</code> functions may modify some or all of the levels.<p>
michael@0:    *      After the <code>nsBidi</code> object is reused or destroyed, the caller
michael@0:    *      must take care of the deallocation of the <code>aEmbeddingLevels</code> array.<p>
michael@0:    *      <strong>The <code>aEmbeddingLevels</code> array must be
michael@0:    *      at least <code>aLength</code> long.</strong>
michael@0:    */
michael@0:   nsresult SetPara(const char16_t *aText, int32_t aLength, nsBidiLevel aParaLevel, nsBidiLevel *aEmbeddingLevels);
michael@0: 
michael@0:   /**
michael@0:    * Get the directionality of the text.
michael@0:    *
michael@0:    * @param aDirection receives a <code>NSBIDI_XXX</code> value that indicates if the entire text
michael@0:    *       represented by this object is unidirectional,
michael@0:    *       and which direction, or if it is mixed-directional.
michael@0:    *
michael@0:    * @see nsBidiDirection
michael@0:    */
michael@0:   nsresult GetDirection(nsBidiDirection* aDirection);
michael@0: 
michael@0:   /**
michael@0:    * Get the paragraph level of the text.
michael@0:    *
michael@0:    * @param aParaLevel receives a <code>NSBIDI_XXX</code> value indicating the paragraph level
michael@0:    *
michael@0:    * @see nsBidiLevel
michael@0:    */
michael@0:   nsresult GetParaLevel(nsBidiLevel* aParaLevel);
michael@0: 
michael@0: #ifdef FULL_BIDI_ENGINE
michael@0:   /**
michael@0:    * <code>SetLine</code> sets an <code>nsBidi</code> to
michael@0:    * contain the reordering information, especially the resolved levels,
michael@0:    * for all the characters in a line of text. This line of text is
michael@0:    * specified by referring to an <code>nsBidi</code> object representing
michael@0:    * this information for a paragraph of text, and by specifying
michael@0:    * a range of indexes in this paragraph.<p>
michael@0:    * In the new line object, the indexes will range from 0 to <code>aLimit-aStart</code>.<p>
michael@0:    *
michael@0:    * This is used after calling <code>SetPara</code>
michael@0:    * for a paragraph, and after line-breaking on that paragraph.
michael@0:    * It is not necessary if the paragraph is treated as a single line.<p>
michael@0:    *
michael@0:    * After line-breaking, rules (L1) and (L2) for the treatment of
michael@0:    * trailing WS and for reordering are performed on
michael@0:    * an <code>nsBidi</code> object that represents a line.<p>
michael@0:    *
michael@0:    * <strong>Important:</strong> the line <code>nsBidi</code> object shares data with
michael@0:    * <code>aParaBidi</code>.
michael@0:    * You must destroy or reuse this object before <code>aParaBidi</code>.
michael@0:    * In other words, you must destroy or reuse the <code>nsBidi</code> object for a line
michael@0:    * before the object for its parent paragraph.
michael@0:    *
michael@0:    * @param aParaBidi is the parent paragraph object.
michael@0:    *
michael@0:    * @param aStart is the line's first index into the paragraph text.
michael@0:    *
michael@0:    * @param aLimit is just behind the line's last index into the paragraph text
michael@0:    *      (its last index +1).<br>
michael@0:    *      It must be <code>0<=aStart<=aLimit<=</code>paragraph length.
michael@0:    *
michael@0:    * @see SetPara
michael@0:    */
michael@0:   nsresult SetLine(nsIBidi* aParaBidi, int32_t aStart, int32_t aLimit);  
michael@0: 
michael@0:   /**
michael@0:    * Get the length of the text.
michael@0:    *
michael@0:    * @param aLength receives the length of the text that the nsBidi object was created for.
michael@0:    */
michael@0:   nsresult GetLength(int32_t* aLength);
michael@0: 
michael@0:   /**
michael@0:    * Get the level for one character.
michael@0:    *
michael@0:    * @param aCharIndex the index of a character.
michael@0:    *
michael@0:    * @param aLevel receives the level for the character at aCharIndex.
michael@0:    *
michael@0:    * @see nsBidiLevel
michael@0:    */
michael@0:   nsresult GetLevelAt(int32_t aCharIndex,  nsBidiLevel* aLevel);
michael@0: 
michael@0:   /**
michael@0:    * Get an array of levels for each character.<p>
michael@0:    *
michael@0:    * Note that this function may allocate memory under some
michael@0:    * circumstances, unlike <code>GetLevelAt</code>.
michael@0:    *
michael@0:    * @param aLevels receives a pointer to the levels array for the text,
michael@0:    *       or <code>nullptr</code> if an error occurs.
michael@0:    *
michael@0:    * @see nsBidiLevel
michael@0:    */
michael@0:   nsresult GetLevels(nsBidiLevel** aLevels);
michael@0: #endif // FULL_BIDI_ENGINE
michael@0:   /**
michael@0:    * Get the bidirectional type for one character.
michael@0:    *
michael@0:    * @param aCharIndex the index of a character.
michael@0:    *
michael@0:    * @param aType receives the bidirectional type of the character at aCharIndex.
michael@0:    */
michael@0:   nsresult GetCharTypeAt(int32_t aCharIndex,  nsCharType* aType);
michael@0: 
michael@0:   /**
michael@0:    * Get a logical run.
michael@0:    * This function returns information about a run and is used
michael@0:    * to retrieve runs in logical order.<p>
michael@0:    * This is especially useful for line-breaking on a paragraph.
michael@0:    *
michael@0:    * @param aLogicalStart is the first character of the run.
michael@0:    *
michael@0:    * @param aLogicalLimit will receive the limit of the run.
michael@0:    *      The l-value that you point to here may be the
michael@0:    *      same expression (variable) as the one for
michael@0:    *      <code>aLogicalStart</code>.
michael@0:    *      This pointer can be <code>nullptr</code> if this
michael@0:    *      value is not necessary.
michael@0:    *
michael@0:    * @param aLevel will receive the level of the run.
michael@0:    *      This pointer can be <code>nullptr</code> if this
michael@0:    *      value is not necessary.
michael@0:    */
michael@0:   nsresult GetLogicalRun(int32_t aLogicalStart, int32_t* aLogicalLimit, nsBidiLevel* aLevel);
michael@0: 
michael@0:   /**
michael@0:    * Get the number of runs.
michael@0:    * This function may invoke the actual reordering on the
michael@0:    * <code>nsBidi</code> object, after <code>SetPara</code>
michael@0:    * may have resolved only the levels of the text. Therefore,
michael@0:    * <code>CountRuns</code> may have to allocate memory,
michael@0:    * and may fail doing so.
michael@0:    *
michael@0:    * @param aRunCount will receive the number of runs.
michael@0:    */
michael@0:   nsresult CountRuns(int32_t* aRunCount);
michael@0: 
michael@0:   /**
michael@0:    * Get one run's logical start, length, and directionality,
michael@0:    * which can be 0 for LTR or 1 for RTL.
michael@0:    * In an RTL run, the character at the logical start is
michael@0:    * visually on the right of the displayed run.
michael@0:    * The length is the number of characters in the run.<p>
michael@0:    * <code>CountRuns</code> should be called
michael@0:    * before the runs are retrieved.
michael@0:    *
michael@0:    * @param aRunIndex is the number of the run in visual order, in the
michael@0:    *      range <code>[0..CountRuns-1]</code>.
michael@0:    *
michael@0:    * @param aLogicalStart is the first logical character index in the text.
michael@0:    *      The pointer may be <code>nullptr</code> if this index is not needed.
michael@0:    *
michael@0:    * @param aLength is the number of characters (at least one) in the run.
michael@0:    *      The pointer may be <code>nullptr</code> if this is not needed.
michael@0:    *
michael@0:    * @param aDirection will receive the directionality of the run,
michael@0:    *       <code>NSBIDI_LTR==0</code> or <code>NSBIDI_RTL==1</code>,
michael@0:    *       never <code>NSBIDI_MIXED</code>.
michael@0:    *
michael@0:    * @see CountRuns<p>
michael@0:    *
michael@0:    * Example:
michael@0:    * @code
michael@0:    *  int32_t i, count, logicalStart, visualIndex=0, length;
michael@0:    *  nsBidiDirection dir;
michael@0:    *  pBidi->CountRuns(&count);
michael@0:    *  for(i=0; i<count; ++i) {
michael@0:    *    pBidi->GetVisualRun(i, &logicalStart, &length, &dir);
michael@0:    *    if(NSBIDI_LTR==dir) {
michael@0:    *      do { // LTR
michael@0:    *        show_char(text[logicalStart++], visualIndex++);
michael@0:    *      } while(--length>0);
michael@0:    *    } else {
michael@0:    *      logicalStart+=length;  // logicalLimit
michael@0:    *      do { // RTL
michael@0:    *        show_char(text[--logicalStart], visualIndex++);
michael@0:    *      } while(--length>0);
michael@0:    *    }
michael@0:    *  }
michael@0:    * @endcode
michael@0:    *
michael@0:    * Note that in right-to-left runs, code like this places
michael@0:    * modifier letters before base characters and second surrogates
michael@0:    * before first ones.
michael@0:    */
michael@0:   nsresult GetVisualRun(int32_t aRunIndex, int32_t* aLogicalStart, int32_t* aLength, nsBidiDirection* aDirection);
michael@0: 
michael@0: #ifdef FULL_BIDI_ENGINE
michael@0:   /**
michael@0:    * Get the visual position from a logical text position.
michael@0:    * If such a mapping is used many times on the same
michael@0:    * <code>nsBidi</code> object, then calling
michael@0:    * <code>GetLogicalMap</code> is more efficient.<p>
michael@0:    *
michael@0:    * Note that in right-to-left runs, this mapping places
michael@0:    * modifier letters before base characters and second surrogates
michael@0:    * before first ones.
michael@0:    *
michael@0:    * @param aLogicalIndex is the index of a character in the text.
michael@0:    *
michael@0:    * @param aVisualIndex will receive the visual position of this character.
michael@0:    *
michael@0:    * @see GetLogicalMap
michael@0:    * @see GetLogicalIndex
michael@0:    */
michael@0:   nsresult GetVisualIndex(int32_t aLogicalIndex, int32_t* aVisualIndex);
michael@0: 
michael@0:   /**
michael@0:    * Get the logical text position from a visual position.
michael@0:    * If such a mapping is used many times on the same
michael@0:    * <code>nsBidi</code> object, then calling
michael@0:    * <code>GetVisualMap</code> is more efficient.<p>
michael@0:    *
michael@0:    * This is the inverse function to <code>GetVisualIndex</code>.
michael@0:    *
michael@0:    * @param aVisualIndex is the visual position of a character.
michael@0:    *
michael@0:    * @param aLogicalIndex will receive the index of this character in the text.
michael@0:    *
michael@0:    * @see GetVisualMap
michael@0:    * @see GetVisualIndex
michael@0:    */
michael@0:   nsresult GetLogicalIndex(int32_t aVisualIndex, int32_t* aLogicalIndex);
michael@0: 
michael@0:   /**
michael@0:    * Get a logical-to-visual index map (array) for the characters in the nsBidi
michael@0:    * (paragraph or line) object.
michael@0:    *
michael@0:    * @param aIndexMap is a pointer to an array of <code>GetLength</code>
michael@0:    *      indexes which will reflect the reordering of the characters.
michael@0:    *      The array does not need to be initialized.<p>
michael@0:    *      The index map will result in <code>aIndexMap[aLogicalIndex]==aVisualIndex</code>.<p>
michael@0:    *
michael@0:    * @see GetVisualMap
michael@0:    * @see GetVisualIndex
michael@0:    */
michael@0:   nsresult GetLogicalMap(int32_t *aIndexMap);
michael@0: 
michael@0:   /**
michael@0:    * Get a visual-to-logical index map (array) for the characters in the nsBidi
michael@0:    * (paragraph or line) object.
michael@0:    *
michael@0:    * @param aIndexMap is a pointer to an array of <code>GetLength</code>
michael@0:    *      indexes which will reflect the reordering of the characters.
michael@0:    *      The array does not need to be initialized.<p>
michael@0:    *      The index map will result in <code>aIndexMap[aVisualIndex]==aLogicalIndex</code>.<p>
michael@0:    *
michael@0:    * @see GetLogicalMap
michael@0:    * @see GetLogicalIndex
michael@0:    */
michael@0:   nsresult GetVisualMap(int32_t *aIndexMap);
michael@0: 
michael@0:   /**
michael@0:    * This is a convenience function that does not use a nsBidi object.
michael@0:    * It is intended to be used for when an application has determined the levels
michael@0:    * of objects (character sequences) and just needs to have them reordered (L2).
michael@0:    * This is equivalent to using <code>GetLogicalMap</code> on a
michael@0:    * <code>nsBidi</code> object.
michael@0:    *
michael@0:    * @param aLevels is an array with <code>aLength</code> levels that have been determined by
michael@0:    *      the application.
michael@0:    *
michael@0:    * @param aLength is the number of levels in the array, or, semantically,
michael@0:    *      the number of objects to be reordered.
michael@0:    *      It must be <code>aLength>0</code>.
michael@0:    *
michael@0:    * @param aIndexMap is a pointer to an array of <code>aLength</code>
michael@0:    *      indexes which will reflect the reordering of the characters.
michael@0:    *      The array does not need to be initialized.<p>
michael@0:    *      The index map will result in <code>aIndexMap[aLogicalIndex]==aVisualIndex</code>.
michael@0:    */
michael@0:   static nsresult ReorderLogical(const nsBidiLevel *aLevels, int32_t aLength, int32_t *aIndexMap);
michael@0: #endif // FULL_BIDI_ENGINE
michael@0:   /**
michael@0:    * This is a convenience function that does not use a nsBidi object.
michael@0:    * It is intended to be used for when an application has determined the levels
michael@0:    * of objects (character sequences) and just needs to have them reordered (L2).
michael@0:    * This is equivalent to using <code>GetVisualMap</code> on a
michael@0:    * <code>nsBidi</code> object.
michael@0:    *
michael@0:    * @param aLevels is an array with <code>aLength</code> levels that have been determined by
michael@0:    *      the application.
michael@0:    *
michael@0:    * @param aLength is the number of levels in the array, or, semantically,
michael@0:    *      the number of objects to be reordered.
michael@0:    *      It must be <code>aLength>0</code>.
michael@0:    *
michael@0:    * @param aIndexMap is a pointer to an array of <code>aLength</code>
michael@0:    *      indexes which will reflect the reordering of the characters.
michael@0:    *      The array does not need to be initialized.<p>
michael@0:    *      The index map will result in <code>aIndexMap[aVisualIndex]==aLogicalIndex</code>.
michael@0:    */
michael@0:   static nsresult ReorderVisual(const nsBidiLevel *aLevels, int32_t aLength, int32_t *aIndexMap);
michael@0: 
michael@0: #ifdef FULL_BIDI_ENGINE
michael@0:   /**
michael@0:    * Invert an index map.
michael@0:    * The one-to-one index mapping of the first map is inverted and written to
michael@0:    * the second one.
michael@0:    *
michael@0:    * @param aSrcMap is an array with <code>aLength</code> indexes
michael@0:    *      which define the original mapping.
michael@0:    *
michael@0:    * @param aDestMap is an array with <code>aLength</code> indexes
michael@0:    *      which will be filled with the inverse mapping.
michael@0:    *
michael@0:    * @param aLength is the length of each array.
michael@0:    */
michael@0:   nsresult InvertMap(const int32_t *aSrcMap, int32_t *aDestMap, int32_t aLength);
michael@0: #endif // FULL_BIDI_ENGINE
michael@0:   /**
michael@0:    * Reverse a Right-To-Left run of Unicode text.
michael@0:    *
michael@0:    * This function preserves the integrity of characters with multiple
michael@0:    * code units and (optionally) modifier letters.
michael@0:    * Characters can be replaced by mirror-image characters
michael@0:    * in the destination buffer. Note that "real" mirroring has
michael@0:    * to be done in a rendering engine by glyph selection
michael@0:    * and that for many "mirrored" characters there are no
michael@0:    * Unicode characters as mirror-image equivalents.
michael@0:    * There are also options to insert or remove Bidi control
michael@0:    * characters; see the description of the <code>aDestSize</code>
michael@0:    * and <code>aOptions</code> parameters and of the option bit flags.
michael@0:    *
michael@0:    * Since no Bidi controls are inserted here, this function will never
michael@0:    * write more than <code>aSrcLength</code> characters to <code>aDest</code>.
michael@0:    *
michael@0:    * @param aSrc A pointer to the RTL run text.
michael@0:    *
michael@0:    * @param aSrcLength The length of the RTL run.
michael@0:    *                 If the <code>NSBIDI_REMOVE_BIDI_CONTROLS</code> option
michael@0:    *                 is set, then the destination length may be less than
michael@0:    *                 <code>aSrcLength</code>.
michael@0:    *                 If this option is not set, then the destination length
michael@0:    *                 will be exactly <code>aSrcLength</code>.
michael@0:    *
michael@0:    * @param aDest A pointer to where the reordered text is to be copied.
michael@0:    *             <code>aSrc[aSrcLength]</code> and <code>aDest[aSrcLength]</code>
michael@0:    *             must not overlap.
michael@0:    *
michael@0:    * @param aOptions A bit set of options for the reordering that control
michael@0:    *                how the reordered text is written.
michael@0:    *
michael@0:    * @param aDestSize will receive the number of characters that were written to <code>aDest</code>.
michael@0:    */
michael@0:   nsresult WriteReverse(const char16_t *aSrc, int32_t aSrcLength, char16_t *aDest, uint16_t aOptions, int32_t *aDestSize);
michael@0: 
michael@0: protected:
michael@0:   friend class nsBidiPresUtils;
michael@0: 
michael@0:   /** length of the current text */
michael@0:   int32_t mLength;
michael@0: 
michael@0:   /** memory sizes in bytes */
michael@0:   size_t mDirPropsSize, mLevelsSize, mRunsSize;
michael@0: 
michael@0:   /** allocated memory */
michael@0:   DirProp* mDirPropsMemory;
michael@0:   nsBidiLevel* mLevelsMemory;
michael@0:   Run* mRunsMemory;
michael@0: 
michael@0:   /** indicators for whether memory may be allocated after construction */
michael@0:   bool mMayAllocateText, mMayAllocateRuns;
michael@0: 
michael@0:   const DirProp* mDirProps;
michael@0:   nsBidiLevel* mLevels;
michael@0: 
michael@0:   /** the paragraph level */
michael@0:   nsBidiLevel mParaLevel;
michael@0: 
michael@0:   /** flags is a bit set for which directional properties are in the text */
michael@0:   Flags mFlags;
michael@0: 
michael@0:   /** the overall paragraph or line directionality - see nsBidiDirection */
michael@0:   nsBidiDirection mDirection;
michael@0: 
michael@0:   /** characters after trailingWSStart are WS and are */
michael@0:   /* implicitly at the paraLevel (rule (L1)) - levels may not reflect that */
michael@0:   int32_t mTrailingWSStart;
michael@0: 
michael@0:   /** fields for line reordering */
michael@0:   int32_t mRunCount;     /* ==-1: runs not set up yet */
michael@0:   Run* mRuns;
michael@0: 
michael@0:   /** for non-mixed text, we only need a tiny array of runs (no malloc()) */
michael@0:   Run mSimpleRuns[1];
michael@0: 
michael@0: private:
michael@0: 
michael@0:   void Init();
michael@0: 
michael@0:   bool GetMemory(void **aMemory, size_t* aSize, bool aMayAllocate, size_t aSizeNeeded);
michael@0: 
michael@0:   void Free();
michael@0: 
michael@0:   void GetDirProps(const char16_t *aText);
michael@0: 
michael@0:   nsBidiDirection ResolveExplicitLevels();
michael@0: 
michael@0:   nsresult CheckExplicitLevels(nsBidiDirection *aDirection);
michael@0: 
michael@0:   nsBidiDirection DirectionFromFlags(Flags aFlags);
michael@0: 
michael@0:   void ResolveImplicitLevels(int32_t aStart, int32_t aLimit, DirProp aSOR, DirProp aEOR);
michael@0: 
michael@0:   void AdjustWSLevels();
michael@0: 
michael@0:   void SetTrailingWSStart();
michael@0: 
michael@0:   bool GetRuns();
michael@0: 
michael@0:   void GetSingleRun(nsBidiLevel aLevel);
michael@0: 
michael@0:   void ReorderLine(nsBidiLevel aMinLevel, nsBidiLevel aMaxLevel);
michael@0: 
michael@0:   static bool PrepareReorder(const nsBidiLevel *aLevels, int32_t aLength, int32_t *aIndexMap, nsBidiLevel *aMinLevel, nsBidiLevel *aMaxLevel);
michael@0: 
michael@0:   int32_t doWriteReverse(const char16_t *src, int32_t srcLength,
michael@0:                          char16_t *dest, uint16_t options);
michael@0: 
michael@0: };
michael@0: 
michael@0: #endif // _nsBidi_h_