1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/content/base/public/nsLineBreaker.h Wed Dec 31 06:09:35 2014 +0100
1.3 @@ -0,0 +1,224 @@
1.4 +/* -*- Mode: C++; tab-width: 2; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
1.5 +/* This Source Code Form is subject to the terms of the Mozilla Public
1.6 + * License, v. 2.0. If a copy of the MPL was not distributed with this
1.7 + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
1.8 +
1.9 +#ifndef NSLINEBREAKER_H_
1.10 +#define NSLINEBREAKER_H_
1.11 +
1.12 +#include "nsString.h"
1.13 +#include "nsTArray.h"
1.14 +#include "nsILineBreaker.h"
1.15 +
1.16 +class nsIAtom;
1.17 +class nsHyphenator;
1.18 +
1.19 +/**
1.20 + * A receiver of line break data.
1.21 + */
1.22 +class nsILineBreakSink {
1.23 +public:
1.24 + /**
1.25 + * Sets the break data for a substring of the associated text chunk.
1.26 + * One or more of these calls will be performed; the union of all substrings
1.27 + * will cover the entire text chunk. Substrings may overlap (i.e., we may
1.28 + * set the break-before state of a character more than once).
1.29 + * @param aBreakBefore the break-before states for the characters in the substring.
1.30 + * These are enum values from gfxTextRun::CompressedGlyph:
1.31 + * FLAG_BREAK_TYPE_NONE - no linebreak is allowed here
1.32 + * FLAG_BREAK_TYPE_NORMAL - a normal (whitespace) linebreak
1.33 + * FLAG_BREAK_TYPE_HYPHEN - a hyphenation point
1.34 + */
1.35 + virtual void SetBreaks(uint32_t aStart, uint32_t aLength, uint8_t* aBreakBefore) = 0;
1.36 +
1.37 + /**
1.38 + * Indicates which characters should be capitalized. Only called if
1.39 + * BREAK_NEED_CAPITALIZATION was requested.
1.40 + */
1.41 + virtual void SetCapitalization(uint32_t aStart, uint32_t aLength, bool* aCapitalize) = 0;
1.42 +};
1.43 +
1.44 +/**
1.45 + * A line-breaking state machine. You feed text into it via AppendText calls
1.46 + * and it computes the possible line breaks. Because break decisions can
1.47 + * require a lot of context, the breaks for a piece of text are sometimes not
1.48 + * known until later text has been seen (or all text ends). So breaks are
1.49 + * returned via a call to SetBreaks on the nsILineBreakSink object passed
1.50 + * with each text chunk, which might happen during the corresponding AppendText
1.51 + * call, or might happen during a later AppendText call or even a Reset()
1.52 + * call.
1.53 + *
1.54 + * The linebreak results MUST NOT depend on how the text is broken up
1.55 + * into AppendText calls.
1.56 + *
1.57 + * The current strategy is that we break the overall text into
1.58 + * whitespace-delimited "words". Then those words are passed to the nsILineBreaker
1.59 + * service for deeper analysis if they contain a "complex" character as described
1.60 + * below.
1.61 + *
1.62 + * This class also handles detection of which characters should be capitalized
1.63 + * for text-transform:capitalize. This is a good place to handle that because
1.64 + * we have all the context we need.
1.65 + */
1.66 +class nsLineBreaker {
1.67 +public:
1.68 + nsLineBreaker();
1.69 + ~nsLineBreaker();
1.70 +
1.71 + static inline bool IsSpace(char16_t u) { return NS_IsSpace(u); }
1.72 +
1.73 + static inline bool IsComplexASCIIChar(char16_t u)
1.74 + {
1.75 + return !((0x0030 <= u && u <= 0x0039) ||
1.76 + (0x0041 <= u && u <= 0x005A) ||
1.77 + (0x0061 <= u && u <= 0x007A) ||
1.78 + (0x000a == u));
1.79 + }
1.80 +
1.81 + static inline bool IsComplexChar(char16_t u)
1.82 + {
1.83 + return IsComplexASCIIChar(u) ||
1.84 + NS_NeedsPlatformNativeHandling(u) ||
1.85 + (0x1100 <= u && u <= 0x11ff) || // Hangul Jamo
1.86 + (0x2000 <= u && u <= 0x21ff) || // Punctuations and Symbols
1.87 + (0x2e80 <= u && u <= 0xd7ff) || // several CJK blocks
1.88 + (0xf900 <= u && u <= 0xfaff) || // CJK Compatibility Idographs
1.89 + (0xff00 <= u && u <= 0xffef); // Halfwidth and Fullwidth Forms
1.90 + }
1.91 +
1.92 + // Break opportunities exist at the end of each run of breakable whitespace
1.93 + // (see IsSpace above). Break opportunities can also exist between pairs of
1.94 + // non-whitespace characters, as determined by nsILineBreaker. We pass a whitespace-
1.95 + // delimited word to nsILineBreaker if it contains at least one character
1.96 + // matching IsComplexChar.
1.97 + // We provide flags to control on a per-chunk basis where breaks are allowed.
1.98 + // At any character boundary, exactly one text chunk governs whether a
1.99 + // break is allowed at that boundary.
1.100 + //
1.101 + // We operate on text after whitespace processing has been applied, so
1.102 + // other characters (e.g. tabs and newlines) may have been converted to
1.103 + // spaces.
1.104 +
1.105 + /**
1.106 + * Flags passed with each chunk of text.
1.107 + */
1.108 + enum {
1.109 + /*
1.110 + * Do not introduce a break opportunity at the start of this chunk of text.
1.111 + */
1.112 + BREAK_SUPPRESS_INITIAL = 0x01,
1.113 + /**
1.114 + * Do not introduce a break opportunity in the interior of this chunk of text.
1.115 + * Also, whitespace in this chunk is treated as non-breakable.
1.116 + */
1.117 + BREAK_SUPPRESS_INSIDE = 0x02,
1.118 + /**
1.119 + * The sink currently is already set up to have no breaks in it;
1.120 + * if no breaks are possible, nsLineBreaker does not need to call
1.121 + * SetBreaks on it. This is useful when handling large quantities of
1.122 + * preformatted text; the textruns will never have any breaks set on them,
1.123 + * and there is no need to ever actually scan the text for breaks, except
1.124 + * at the end of textruns in case context is needed for following breakable
1.125 + * text.
1.126 + */
1.127 + BREAK_SKIP_SETTING_NO_BREAKS = 0x04,
1.128 + /**
1.129 + * We need to be notified of characters that should be capitalized
1.130 + * (as in text-transform:capitalize) in this chunk of text.
1.131 + */
1.132 + BREAK_NEED_CAPITALIZATION = 0x08,
1.133 + /**
1.134 + * Auto-hyphenation is enabled, so we need to get a hyphenator
1.135 + * (if available) and use it to find breakpoints.
1.136 + */
1.137 + BREAK_USE_AUTO_HYPHENATION = 0x10
1.138 + };
1.139 +
1.140 + /**
1.141 + * Append "invisible whitespace". This acts like whitespace, but there is
1.142 + * no actual text associated with it. Only the BREAK_SUPPRESS_INSIDE flag
1.143 + * is relevant here.
1.144 + */
1.145 + nsresult AppendInvisibleWhitespace(uint32_t aFlags);
1.146 +
1.147 + /**
1.148 + * Feed Unicode text into the linebreaker for analysis. aLength must be
1.149 + * nonzero.
1.150 + * @param aSink can be null if the breaks are not actually needed (we may
1.151 + * still be setting up state for later breaks)
1.152 + */
1.153 + nsresult AppendText(nsIAtom* aHyphenationLanguage, const char16_t* aText, uint32_t aLength,
1.154 + uint32_t aFlags, nsILineBreakSink* aSink);
1.155 + /**
1.156 + * Feed 8-bit text into the linebreaker for analysis. aLength must be nonzero.
1.157 + * @param aSink can be null if the breaks are not actually needed (we may
1.158 + * still be setting up state for later breaks)
1.159 + */
1.160 + nsresult AppendText(nsIAtom* aHyphenationLanguage, const uint8_t* aText, uint32_t aLength,
1.161 + uint32_t aFlags, nsILineBreakSink* aSink);
1.162 + /**
1.163 + * Reset all state. This means the current run has ended; any outstanding
1.164 + * calls through nsILineBreakSink are made, and all outstanding references to
1.165 + * nsILineBreakSink objects are dropped.
1.166 + * After this call, this linebreaker can be reused.
1.167 + * This must be called at least once between any call to AppendText() and
1.168 + * destroying the object.
1.169 + * @param aTrailingBreak this is set to true when there is a break opportunity
1.170 + * at the end of the text. This will normally only be declared true when there
1.171 + * is breakable whitespace at the end.
1.172 + */
1.173 + nsresult Reset(bool* aTrailingBreak);
1.174 +
1.175 + /*
1.176 + * Set word-break mode for linebreaker. This is set by word-break property.
1.177 + * @param aMode is nsILineBreaker::kWordBreak_* value.
1.178 + */
1.179 + void SetWordBreak(uint8_t aMode) { mWordBreak = aMode; }
1.180 +
1.181 +private:
1.182 + // This is a list of text sources that make up the "current word" (i.e.,
1.183 + // run of text which does not contain any whitespace). All the mLengths
1.184 + // are are nonzero, these cannot overlap.
1.185 + struct TextItem {
1.186 + TextItem(nsILineBreakSink* aSink, uint32_t aSinkOffset, uint32_t aLength,
1.187 + uint32_t aFlags)
1.188 + : mSink(aSink), mSinkOffset(aSinkOffset), mLength(aLength), mFlags(aFlags) {}
1.189 +
1.190 + nsILineBreakSink* mSink;
1.191 + uint32_t mSinkOffset;
1.192 + uint32_t mLength;
1.193 + uint32_t mFlags;
1.194 + };
1.195 +
1.196 + // State for the nonwhitespace "word" that started in previous text and hasn't
1.197 + // finished yet.
1.198 +
1.199 + // When the current word ends, this computes the linebreak opportunities
1.200 + // *inside* the word (excluding either end) and sets them through the
1.201 + // appropriate sink(s). Then we clear the current word state.
1.202 + nsresult FlushCurrentWord();
1.203 +
1.204 + void UpdateCurrentWordLanguage(nsIAtom *aHyphenationLanguage);
1.205 +
1.206 + void FindHyphenationPoints(nsHyphenator *aHyphenator,
1.207 + const char16_t *aTextStart,
1.208 + const char16_t *aTextLimit,
1.209 + uint8_t *aBreakState);
1.210 +
1.211 + nsAutoTArray<char16_t,100> mCurrentWord;
1.212 + // All the items that contribute to mCurrentWord
1.213 + nsAutoTArray<TextItem,2> mTextItems;
1.214 + nsIAtom* mCurrentWordLanguage;
1.215 + bool mCurrentWordContainsMixedLang;
1.216 + bool mCurrentWordContainsComplexChar;
1.217 +
1.218 + // True if the previous character was breakable whitespace
1.219 + bool mAfterBreakableSpace;
1.220 + // True if a break must be allowed at the current position because
1.221 + // a run of breakable whitespace ends here
1.222 + bool mBreakHere;
1.223 + // line break mode by "word-break" style
1.224 + uint8_t mWordBreak;
1.225 +};
1.226 +
1.227 +#endif /*NSLINEBREAKER_H_*/
