/* -*- Mode: C++; tab-width: 20; indent-tabs-mode: nil; c-basic-offset: 4 -*- * ***** BEGIN LICENSE BLOCK ***** * Version: MPL 1.1/GPL 2.0/LGPL 2.1 * * The contents of this file are subject to the Mozilla Public License Version * 1.1 (the "License"); you may not use this file except in compliance with * the License. You may obtain a copy of the License at * http://www.mozilla.org/MPL/ * * Software distributed under the License is distributed on an "AS IS" basis, * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License * for the specific language governing rights and limitations under the * License. * * The Original Code is Mozilla Foundation code. * * The Initial Developer of the Original Code is Mozilla Foundation. * Portions created by the Initial Developer are Copyright (C) 2005 * the Initial Developer. All Rights Reserved. * * Contributor(s): * Stuart Parmenter * Masayuki Nakano * * Alternatively, the contents of this file may be used under the terms of * either the GNU General Public License Version 2 or later (the "GPL"), or * the GNU Lesser General Public License Version 2.1 or later (the "LGPL"), * in which case the provisions of the GPL or the LGPL are applicable instead * of those above. If you wish to allow use of your version of this file only * under the terms of either the GPL or the LGPL, and not to allow others to * use your version of this file under the terms of the MPL, indicate your * decision by deleting the provisions above and replace them with the notice * and other provisions required by the GPL or the LGPL. If you do not delete * the provisions above, a recipient may use your version of this file under * the terms of any one of the MPL, the GPL or the LGPL. * * ***** END LICENSE BLOCK ***** */ #ifndef GFX_FONT_H #define GFX_FONT_H #include "prtypes.h" #include "gfxTypes.h" #include "nsString.h" #include "gfxPoint.h" #include "nsTArray.h" #include "nsTHashtable.h" #include "nsHashKeys.h" #include "gfxSkipChars.h" #include "gfxRect.h" #include "nsExpirationTracker.h" #include "nsMathUtils.h" #include "nsBidiUtils.h" #ifdef DEBUG #include #endif class gfxContext; class gfxTextRun; class nsIAtom; class gfxFont; class gfxFontGroup; #define FONT_STYLE_NORMAL 0 #define FONT_STYLE_ITALIC 1 #define FONT_STYLE_OBLIQUE 2 #define FONT_WEIGHT_NORMAL 400 #define FONT_WEIGHT_BOLD 700 #define FONT_MAX_SIZE 2000.0 struct THEBES_API gfxFontStyle { gfxFontStyle(); gfxFontStyle(PRUint8 aStyle, PRUint16 aWeight, gfxFloat aSize, const nsACString& aLangGroup, float aSizeAdjust, PRPackedBool aSystemFont, PRPackedBool aFamilyNameQuirks); gfxFontStyle(const gfxFontStyle& aStyle); // The style of font (normal, italic, oblique) PRUint8 style : 7; // Say that this font is a system font and therefore does not // require certain fixup that we do for fonts from untrusted // sources. PRPackedBool systemFont : 1; // True if the character set quirks (for treatment of "Symbol", // "Wingdings", etc.) should be applied. PRPackedBool familyNameQuirks : 1; // The weight of the font. 100, 200, ... 900 are the weights, and // single integer offsets request the next bolder/lighter font // available. For example, for a font available in weights 200, // 400, 700, and 900, a weight of 898 should lead to the weight 400 // font being used, since it is two weights lighter than 900. PRUint16 weight; // The logical size of the font, in pixels gfxFloat size; // the language group nsCString langGroup; // The aspect-value (ie., the ratio actualsize:actualxheight) that any // actual physical font created from this font structure must have when // rendering or measuring a string. A value of 0 means no adjustment // needs to be done. float sizeAdjust; // Return the final adjusted font size for the given aspect ratio. // Not meant to be called when sizeAdjust = 0. gfxFloat GetAdjustedSize(gfxFloat aspect) const { NS_ASSERTION(sizeAdjust != 0.0, "Not meant to be called when sizeAdjust = 0"); gfxFloat adjustedSize = PR_MAX(NS_round(size*(sizeAdjust/aspect)), 1.0); return PR_MIN(adjustedSize, FONT_MAX_SIZE); } PLDHashNumber Hash() const { return ((style + (systemFont << 7) + (familyNameQuirks << 8) + (weight << 9)) + PRUint32(size*1000) + PRUint32(sizeAdjust*1000)) ^ HashString(langGroup); } void ComputeWeightAndOffset(PRInt8 *outBaseWeight, PRInt8 *outOffset) const; PRBool Equals(const gfxFontStyle& other) const { return (size == other.size) && (style == other.style) && (systemFont == other.systemFont) && (familyNameQuirks == other.familyNameQuirks) && (weight == other.weight) && (langGroup.Equals(other.langGroup)) && (sizeAdjust == other.sizeAdjust); } }; /** * Font cache design: * * The mFonts hashtable contains most fonts, indexed by (name, style). * It does not add a reference to the fonts it contains. * When a font's refcount decreases to zero, instead of deleting it we * add it to our expiration tracker. * The expiration tracker tracks fonts with zero refcount. After a certain * period of time, such fonts expire and are deleted. * * We're using 3 generations with a ten-second generation interval, so * zero-refcount fonts will be deleted 20-30 seconds after their refcount * goes to zero, if timer events fire in a timely manner. */ class THEBES_API gfxFontCache : public nsExpirationTracker { public: enum { TIMEOUT_SECONDS = 10 }; gfxFontCache() : nsExpirationTracker(TIMEOUT_SECONDS*1000) { mFonts.Init(); } ~gfxFontCache() { // Expire everything that has a zero refcount, so we don't leak them. AgeAllGenerations(); // All fonts should be gone. NS_WARN_IF_FALSE(mFonts.Count() == 0, "Fonts still alive while shutting down gfxFontCache"); // Note that we have to delete everything through the expiration // tracker, since there might be fonts not in the hashtable but in // the tracker. } /* * Get the global gfxFontCache. You must call Init() before * calling this method --- the result will not be null. */ static gfxFontCache* GetCache() { return gGlobalCache; } static nsresult Init(); // It's OK to call this even if Init() has not been called. static void Shutdown(); // Look up a font in the cache. Returns an addrefed pointer, or null // if there's nothing matching in the cache already_AddRefed Lookup(const nsAString &aName, const gfxFontStyle *aFontGroup); // We created a new font (presumably because Lookup returned null); // put it in the cache. The font's refcount should be nonzero. It is // allowable to add a new font even if there is one already in the // cache with the same key; we'll forget about the old one. void AddNew(gfxFont *aFont); // The font's refcount has gone to zero; give ownership of it to // the cache. We delete it if it's not acquired again after a certain // amount of time. void NotifyReleased(gfxFont *aFont); // This gets called when the timeout has expired on a zero-refcount // font; we just delete it. virtual void NotifyExpired(gfxFont *aFont); protected: void DestroyFont(gfxFont *aFont); static gfxFontCache *gGlobalCache; struct Key { const nsAString& mString; const gfxFontStyle* mStyle; Key(const nsAString& aString, const gfxFontStyle* aStyle) : mString(aString), mStyle(aStyle) {} }; class HashEntry : public PLDHashEntryHdr { public: typedef const Key& KeyType; typedef const Key* KeyTypePointer; // When constructing a new entry in the hashtable, we'll leave this // blank. The caller of Put() will fill this in. HashEntry(KeyTypePointer aStr) : mFont(nsnull) { } HashEntry(const HashEntry& toCopy) : mFont(toCopy.mFont) { } ~HashEntry() { } PRBool KeyEquals(const KeyTypePointer aKey) const; static KeyTypePointer KeyToPointer(KeyType aKey) { return &aKey; } static PLDHashNumber HashKey(const KeyTypePointer aKey) { return HashString(aKey->mString) ^ aKey->mStyle->Hash(); } enum { ALLOW_MEMMOVE = PR_TRUE }; gfxFont* mFont; }; nsTHashtable mFonts; }; /** * This stores glyph bounds information for a particular gfxFont, at * a particular appunits-per-dev-pixel ratio (because the compressed glyph * width array is stored in appunits). * * We store a hashtable from glyph IDs to float bounding rects. For the * common case where the glyph has no horizontal left bearing, and no * y overflow above the font ascent or below the font descent, and tight * bounding boxes are not required, we avoid storing the glyph ID in the hashtable * and instead consult an array of 16-bit glyph XMost values (in appunits). * This array always has an entry for the font's space glyph --- the width is * assumed to be zero. */ class THEBES_API gfxGlyphExtents { public: gfxGlyphExtents(PRUint32 aAppUnitsPerDevUnit) : mAppUnitsPerDevUnit(aAppUnitsPerDevUnit) { MOZ_COUNT_CTOR(gfxGlyphExtents); mTightGlyphExtents.Init(); } ~gfxGlyphExtents(); enum { INVALID_WIDTH = 0xFFFF }; // returns INVALID_WIDTH => not a contained glyph // Otherwise the glyph has no before-bearing or vertical bearings, // and the result is its width measured from the baseline origin, in // appunits. PRUint16 GetContainedGlyphWidthAppUnits(PRUint32 aGlyphID) const { return mContainedGlyphWidths.Get(aGlyphID); } PRBool IsGlyphKnown(PRUint32 aGlyphID) const { return mContainedGlyphWidths.Get(aGlyphID) != INVALID_WIDTH || mTightGlyphExtents.GetEntry(aGlyphID) != nsnull; } PRBool IsGlyphKnownWithTightExtents(PRUint32 aGlyphID) const { return mTightGlyphExtents.GetEntry(aGlyphID) != nsnull; } // Get glyph extents; a rectangle relative to the left baseline origin // Returns true on success. Can fail on OOM or when aContext is null // and extents were not (successfully) prefetched. PRBool GetTightGlyphExtentsAppUnits(gfxFont *aFont, gfxContext *aContext, PRUint32 aGlyphID, gfxRect *aExtents); void SetContainedGlyphWidthAppUnits(PRUint32 aGlyphID, PRUint16 aWidth) { mContainedGlyphWidths.Set(aGlyphID, aWidth); } void SetTightGlyphExtents(PRUint32 aGlyphID, const gfxRect& aExtentsAppUnits); PRUint32 GetAppUnitsPerDevUnit() { return mAppUnitsPerDevUnit; } private: class HashEntry : public nsUint32HashKey { public: // When constructing a new entry in the hashtable, we'll leave this // blank. The caller of Put() will fill this in. HashEntry(KeyTypePointer aPtr) : nsUint32HashKey(aPtr) {} HashEntry(const HashEntry& toCopy) : nsUint32HashKey(toCopy) { x = toCopy.x; y = toCopy.y; width = toCopy.width; height = toCopy.height; } float x, y, width, height; }; typedef unsigned long PtrBits; enum { BLOCK_SIZE_BITS = 7, BLOCK_SIZE = 1 << BLOCK_SIZE_BITS }; // 128-glyph blocks class GlyphWidths { public: void Set(PRUint32 aIndex, PRUint16 aValue); PRUint16 Get(PRUint32 aIndex) const { PRUint32 block = aIndex >> BLOCK_SIZE_BITS; if (block >= mBlocks.Length()) return INVALID_WIDTH; PtrBits bits = mBlocks[block]; if (!bits) return INVALID_WIDTH; PRUint32 indexInBlock = aIndex & (BLOCK_SIZE - 1); if (bits & 0x1) { if (GetGlyphOffset(bits) != indexInBlock) return INVALID_WIDTH; return GetWidth(bits); } PRUint16 *widths = reinterpret_cast(bits); return widths[indexInBlock]; } #ifdef DEBUG PRUint32 ComputeSize(); #endif ~GlyphWidths(); private: static PRUint32 GetGlyphOffset(PtrBits aBits) { NS_ASSERTION(aBits & 0x1, "This is really a pointer..."); return (aBits >> 1) & ((1 << BLOCK_SIZE_BITS) - 1); } static PRUint32 GetWidth(PtrBits aBits) { NS_ASSERTION(aBits & 0x1, "This is really a pointer..."); return aBits >> (1 + BLOCK_SIZE_BITS); } static PtrBits MakeSingle(PRUint32 aGlyphOffset, PRUint16 aWidth) { return (aWidth << (1 + BLOCK_SIZE_BITS)) + (aGlyphOffset << 1) + 1; } nsTArray mBlocks; }; GlyphWidths mContainedGlyphWidths; nsTHashtable mTightGlyphExtents; PRUint32 mAppUnitsPerDevUnit; }; /* a SPECIFIC single font family */ class THEBES_API gfxFont { public: nsrefcnt AddRef(void) { NS_PRECONDITION(PRInt32(mRefCnt) >= 0, "illegal refcnt"); if (mExpirationState.IsTracked()) { gfxFontCache::GetCache()->RemoveObject(this); } ++mRefCnt; NS_LOG_ADDREF(this, mRefCnt, "gfxFont", sizeof(*this)); return mRefCnt; } nsrefcnt Release(void) { NS_PRECONDITION(0 != mRefCnt, "dup release"); --mRefCnt; NS_LOG_RELEASE(this, mRefCnt, "gfxFont"); if (mRefCnt == 0) { // Don't delete just yet; return the object to the cache for // possibly recycling within some time limit gfxFontCache::GetCache()->NotifyReleased(this); return 0; } return mRefCnt; } PRInt32 GetRefCount() { return mRefCnt; } protected: nsAutoRefCnt mRefCnt; public: gfxFont(const nsAString &aName, const gfxFontStyle *aFontGroup); virtual ~gfxFont(); const nsString& GetName() const { return mName; } const gfxFontStyle *GetStyle() const { return &mStyle; } virtual nsString GetUniqueName() = 0; // Font metrics struct Metrics { gfxFloat xHeight; gfxFloat superscriptOffset; gfxFloat subscriptOffset; gfxFloat strikeoutSize; gfxFloat strikeoutOffset; gfxFloat underlineSize; gfxFloat underlineOffset; gfxFloat height; gfxFloat internalLeading; gfxFloat externalLeading; gfxFloat emHeight; gfxFloat emAscent; gfxFloat emDescent; gfxFloat maxHeight; gfxFloat maxAscent; gfxFloat maxDescent; gfxFloat maxAdvance; gfxFloat aveCharWidth; gfxFloat spaceWidth; gfxFloat zeroOrAveCharWidth; // width of '0', or if there is // no '0' glyph in this font, // equal to .aveCharWidth }; virtual const gfxFont::Metrics& GetMetrics() = 0; /** * We let layout specify spacing on either side of any * character. We need to specify both before and after * spacing so that substring measurement can do the right things. * These values are in appunits. They're always an integral number of * appunits, but we specify them in floats in case very large spacing * values are required. */ struct Spacing { gfxFloat mBefore; gfxFloat mAfter; }; /** * Metrics for a particular string */ struct RunMetrics { RunMetrics() { mAdvanceWidth = mAscent = mDescent = 0.0; mBoundingBox = gfxRect(0,0,0,0); } void CombineWith(const RunMetrics& aOtherOnRight) { mAscent = PR_MAX(mAscent, aOtherOnRight.mAscent); mDescent = PR_MAX(mDescent, aOtherOnRight.mDescent); mBoundingBox = mBoundingBox.Union(aOtherOnRight.mBoundingBox + gfxPoint(mAdvanceWidth, 0)); mAdvanceWidth += aOtherOnRight.mAdvanceWidth; } // can be negative (partly due to negative spacing). // Advance widths should be additive: the advance width of the // (offset1, length1) plus the advance width of (offset1 + length1, // length2) should be the advance width of (offset1, length1 + length2) gfxFloat mAdvanceWidth; // For zero-width substrings, these must be zero! gfxFloat mAscent; // always non-negative gfxFloat mDescent; // always non-negative // Bounding box that is guaranteed to include everything drawn. // If aTightBoundingBox was set to true when these metrics were // generated, this will tightly wrap the glyphs, otherwise it is // "loose" and may be larger than the true bounding box. // Coordinates are relative to the baseline left origin, so typically // mBoundingBox.y == -mAscent gfxRect mBoundingBox; }; /** * Draw a series of glyphs to aContext. The direction of aTextRun must * be honoured. * @param aStart the first character to draw * @param aEnd draw characters up to here * @param aBaselineOrigin the baseline origin; the left end of the baseline * for LTR textruns, the right end of the baseline for RTL textruns. On return, * this should be updated to the other end of the baseline. In application * units, really! * @param aSpacing spacing to insert before and after characters (for RTL * glyphs, before-spacing is inserted to the right of characters). There * are aEnd - aStart elements in this array, unless it's null to indicate * that there is no spacing. * @param aDrawToPath when true, add the glyph outlines to the current path * instead of drawing the glyphs * * Callers guarantee: * -- aStart and aEnd are aligned to cluster and ligature boundaries * -- all glyphs use this font * * The default implementation builds a cairo glyph array and * calls cairo_show_glyphs or cairo_glyph_path. */ virtual void Draw(gfxTextRun *aTextRun, PRUint32 aStart, PRUint32 aEnd, gfxContext *aContext, PRBool aDrawToPath, gfxPoint *aBaselineOrigin, Spacing *aSpacing); /** * Measure a run of characters. See gfxTextRun::Metrics. * @param aTight if false, then return the union of the glyph extents * with the font-box for the characters (the rectangle with x=0,width= * the advance width for the character run,y=-(font ascent), and height= * font ascent + font descent). Otherwise, we must return as tight as possible * an approximation to the area actually painted by glyphs. * @param aContextForTightBoundingBox when aTight is true, this must * be non-null. * @param aSpacing spacing to insert before and after glyphs. The bounding box * need not include the spacing itself, but the spacing affects the glyph * positions. null if there is no spacing. * * Callers guarantee: * -- aStart and aEnd are aligned to cluster and ligature boundaries * -- all glyphs use this font * * The default implementation just uses font metrics and aTextRun's * advances, and assumes no characters fall outside the font box. In * general this is insufficient, because that assumption is not always true. */ virtual RunMetrics Measure(gfxTextRun *aTextRun, PRUint32 aStart, PRUint32 aEnd, PRBool aTightBoundingBox, gfxContext *aContextForTightBoundingBox, Spacing *aSpacing); /** * Line breaks have been changed at the beginning and/or end of a substring * of the text. Reshaping may be required; glyph updating is permitted. * @return true if anything was changed, false otherwise */ PRBool NotifyLineBreaksChanged(gfxTextRun *aTextRun, PRUint32 aStart, PRUint32 aLength) { return PR_FALSE; } // Expiration tracking nsExpirationState *GetExpirationState() { return &mExpirationState; } // Get the glyphID of a space virtual PRUint32 GetSpaceGlyph() = 0; gfxGlyphExtents *GetOrCreateGlyphExtents(PRUint32 aAppUnitsPerDevUnit); // You need to call SetupCairoFont on the aCR just before calling this virtual void SetupGlyphExtents(gfxContext *aContext, PRUint32 aGlyphID, PRBool aNeedTight, gfxGlyphExtents *aExtents); // This is called by the default Draw() implementation above. virtual PRBool SetupCairoFont(gfxContext *aContext) = 0; PRBool IsSyntheticBold() { return mSyntheticBoldOffset != 0; } PRUint32 GetSyntheticBoldOffset() { return mSyntheticBoldOffset; } protected: // The family name of the font nsString mName; nsExpirationState mExpirationState; gfxFontStyle mStyle; nsAutoTArray mGlyphExtentsArray; // synthetic bolding for environments where this is not supported by the platform PRUint32 mSyntheticBoldOffset; // number of devunit pixels to offset double-strike, 0 ==> no bolding // some fonts have bad metrics, this method sanitize them. // if this font has bad underline offset, aIsBadUnderlineFont should be true. void SanitizeMetrics(gfxFont::Metrics *aMetrics, PRBool aIsBadUnderlineFont); }; class THEBES_API gfxTextRunFactory { THEBES_INLINE_DECL_REFCOUNTING(gfxTextRunFactory) public: // Flags in the mask 0xFFFF0000 are reserved for textrun clients // Flags in the mask 0x0000F000 are reserved for per-platform fonts // Flags in the mask 0x00000FFF are set by the textrun creator. enum { CACHE_TEXT_FLAGS = 0xF0000000, USER_TEXT_FLAGS = 0x0FFF0000, PLATFORM_TEXT_FLAGS = 0x0000F000, TEXTRUN_TEXT_FLAGS = 0x00000FFF, SETTABLE_FLAGS = CACHE_TEXT_FLAGS | USER_TEXT_FLAGS, /** * When set, the text string pointer used to create the text run * is guaranteed to be available during the lifetime of the text run. */ TEXT_IS_PERSISTENT = 0x0001, /** * When set, the text is known to be all-ASCII (< 128). */ TEXT_IS_ASCII = 0x0002, /** * When set, the text is RTL. */ TEXT_IS_RTL = 0x0004, /** * When set, spacing is enabled and the textrun needs to call GetSpacing * on the spacing provider. */ TEXT_ENABLE_SPACING = 0x0008, /** * When set, GetSpacing can return negative spacing. */ TEXT_ENABLE_NEGATIVE_SPACING = 0x0010, /** * When set, GetHyphenationBreaks may return true for some character * positions, otherwise it will always return false for all characters. */ TEXT_ENABLE_HYPHEN_BREAKS = 0x0040, /** * When set, the text has no characters above 255 and it is stored * in the textrun in 8-bit format. */ TEXT_IS_8BIT = 0x0080, /** * When set, the text may have UTF16 surrogate pairs, otherwise it * doesn't. */ TEXT_HAS_SURROGATES = 0x0100, /** * When set, the RunMetrics::mBoundingBox field will be initialized * properly based on glyph extents, in particular, glyph extents that * overflow the standard font-box (the box defined by the ascent, descent * and advance width of the glyph). When not set, it may just be the * standard font-box even if glyphs overflow. */ TEXT_NEED_BOUNDING_BOX = 0x0200, /** * When set, optional ligatures are disabled. Ligatures that are * required for legible text should still be enabled. */ TEXT_DISABLE_OPTIONAL_LIGATURES = 0x0400, /** * When set, the textrun should favour speed of construction over * quality. This may involve disabling ligatures and/or kerning or * other effects. */ TEXT_OPTIMIZE_SPEED = 0x0800 }; /** * This record contains all the parameters needed to initialize a textrun. */ struct Parameters { // A reference context suggesting where the textrun will be rendered gfxContext *mContext; // Pointer to arbitrary user data (which should outlive the textrun) void *mUserData; // A description of which characters have been stripped from the original // DOM string to produce the characters in the textrun. May be null // if that information is not relevant. gfxSkipChars *mSkipChars; // A list of where linebreaks are currently placed in the textrun. May // be null if mInitialBreakCount is zero. PRUint32 *mInitialBreaks; PRUint32 mInitialBreakCount; // The ratio to use to convert device pixels to application layout units PRUint32 mAppUnitsPerDevUnit; }; virtual ~gfxTextRunFactory() {} }; /** * gfxTextRun is an abstraction for drawing and measuring substrings of a run * of text. It stores runs of positioned glyph data, each run having a single * gfxFont. The glyphs are associated with a string of source text, and the * gfxTextRun APIs take parameters that are offsets into that source text. * * gfxTextRuns are not refcounted. They should be deleted when no longer required. * * gfxTextRuns are mostly immutable. The only things that can change are * inter-cluster spacing and line break placement. Spacing is always obtained * lazily by methods that need it, it is not cached. Line breaks are stored * persistently (insofar as they affect the shaping of glyphs; gfxTextRun does * not actually do anything to explicitly account for line breaks). Initially * there are no line breaks. The textrun can record line breaks before or after * any given cluster. (Line breaks specified inside clusters are ignored.) * * It is important that zero-length substrings are handled correctly. This will * be on the test! * * gfxTextRun stores a list of zero or more glyphs for each character. For each * glyph we store the glyph ID, the advance, and possibly an xoffset and yoffset. * The idea is that a string is rendered by a loop that draws each glyph * at its designated offset from the current point, then advances the current * point by the glyph's advance in the direction of the textrun (LTR or RTL). * Each glyph advance is always rounded to the nearest appunit; this ensures * consistent results when dividing the text in a textrun into multiple text * frames (frame boundaries are always aligned to appunits). We optimize * for the case where a character has a single glyph and zero xoffset and yoffset, * and the glyph ID and advance are in a reasonable range so we can pack all * necessary data into 32 bits. * * gfxTextRun methods that measure or draw substrings will associate all the * glyphs in a cluster with the first character of the cluster; if that character * is in the substring, the glyphs will be measured or drawn, otherwise they * won't. */ class THEBES_API gfxTextRun { public: // Override operator delete because we used custom allocation void operator delete(void* aPtr); virtual ~gfxTextRun(); typedef gfxFont::RunMetrics Metrics; // Public textrun API for general use PRBool IsClusterStart(PRUint32 aPos) { NS_ASSERTION(0 <= aPos && aPos < mCharacterCount, "aPos out of range"); return mCharacterGlyphs[aPos].IsClusterStart(); } PRBool IsLigatureGroupStart(PRUint32 aPos) { NS_ASSERTION(0 <= aPos && aPos < mCharacterCount, "aPos out of range"); return mCharacterGlyphs[aPos].IsLigatureGroupStart(); } PRBool CanBreakLineBefore(PRUint32 aPos) { NS_ASSERTION(0 <= aPos && aPos < mCharacterCount, "aPos out of range"); return mCharacterGlyphs[aPos].CanBreakBefore(); } PRUint32 GetLength() { return mCharacterCount; } // All PRUint32 aStart, PRUint32 aLength ranges below are restricted to // grapheme cluster boundaries! All offsets are in terms of the string // passed into MakeTextRun. // All coordinates are in layout/app units /** * Set the potential linebreaks for a substring of the textrun. These are * the "allow break before" points. Initially, there are no potential * linebreaks. * * This can change glyphs and/or geometry! Some textruns' shapes * depend on potential line breaks (e.g., title-case-converting textruns). * This function is virtual so that those textruns can reshape themselves. * * @return true if this changed the linebreaks, false if the new line * breaks are the same as the old */ virtual PRBool SetPotentialLineBreaks(PRUint32 aStart, PRUint32 aLength, PRPackedBool *aBreakBefore, gfxContext *aRefContext); /** * Layout provides PropertyProvider objects. These allow detection of * potential line break points and computation of spacing. We pass the data * this way to allow lazy data acquisition; for example BreakAndMeasureText * will want to only ask for properties of text it's actually looking at. * * NOTE that requested spacing may not actually be applied, if the textrun * is unable to apply it in some context. Exception: spacing around a * whitespace character MUST always be applied. */ class PropertyProvider { public: // Detect hyphenation break opportunities in the given range; breaks // not at cluster boundaries will be ignored. virtual void GetHyphenationBreaks(PRUint32 aStart, PRUint32 aLength, PRPackedBool *aBreakBefore) = 0; // Returns the extra width that will be consumed by a hyphen. This should // be constant for a given textrun. virtual gfxFloat GetHyphenWidth() = 0; typedef gfxFont::Spacing Spacing; /** * Get the spacing around the indicated characters. Spacing must be zero * inside clusters. In other words, if character i is not * CLUSTER_START, then character i-1 must have zero after-spacing and * character i must have zero before-spacing. */ virtual void GetSpacing(PRUint32 aStart, PRUint32 aLength, Spacing *aSpacing) = 0; }; /** * Draws a substring. Uses only GetSpacing from aBreakProvider. * The provided point is the baseline origin on the left of the string * for LTR, on the right of the string for RTL. * @param aDirtyRect if non-null, drawing outside of the rectangle can be * (but does not need to be) dropped. Note that if this is null, we cannot * draw partial ligatures and we will assert if partial ligatures * are detected. * @param aAdvanceWidth if non-null, the advance width of the substring * is returned here. * * Drawing should respect advance widths in the sense that for LTR runs, * Draw(ctx, pt, offset1, length1, dirty, &provider, &advance) followed by * Draw(ctx, gfxPoint(pt.x + advance, pt.y), offset1 + length1, length2, * dirty, &provider, nsnull) should have the same effect as * Draw(ctx, pt, offset1, length1+length2, dirty, &provider, nsnull). * For RTL runs the rule is: * Draw(ctx, pt, offset1 + length1, length2, dirty, &provider, &advance) followed by * Draw(ctx, gfxPoint(pt.x + advance, pt.y), offset1, length1, * dirty, &provider, nsnull) should have the same effect as * Draw(ctx, pt, offset1, length1+length2, dirty, &provider, nsnull). * * Glyphs should be drawn in logical content order, which can be significant * if they overlap (perhaps due to negative spacing). */ void Draw(gfxContext *aContext, gfxPoint aPt, PRUint32 aStart, PRUint32 aLength, const gfxRect *aDirtyRect, PropertyProvider *aProvider, gfxFloat *aAdvanceWidth); /** * Renders a substring to a path. Uses only GetSpacing from aBreakProvider. * The provided point is the baseline origin on the left of the string * for LTR, on the right of the string for RTL. * @param aAdvanceWidth if non-null, the advance width of the substring * is returned here. * * Drawing should respect advance widths in the way that Draw above does. * * Glyphs should be drawn in logical content order. * * UNLIKE Draw above, this cannot be used to render substrings that start or * end inside a ligature. */ void DrawToPath(gfxContext *aContext, gfxPoint aPt, PRUint32 aStart, PRUint32 aLength, PropertyProvider *aBreakProvider, gfxFloat *aAdvanceWidth); /** * Computes the ReflowMetrics for a substring. * Uses GetSpacing from aBreakProvider. * @param aTightBoundingBox if true, we make the bounding box tight */ Metrics MeasureText(PRUint32 aStart, PRUint32 aLength, PRBool aTightBoundingBox, gfxContext *aRefContextForTightBoundingBox, PropertyProvider *aProvider); /** * Computes just the advance width for a substring. * Uses GetSpacing from aBreakProvider. */ gfxFloat GetAdvanceWidth(PRUint32 aStart, PRUint32 aLength, PropertyProvider *aProvider); /** * Clear all stored line breaks for the given range (both before and after), * and then set the line-break state before aStart to aBreakBefore and * after the last cluster to aBreakAfter. * * We require that before and after line breaks be consistent. For clusters * i and i+1, we require that if there is a break after cluster i, a break * will be specified before cluster i+1. This may be temporarily violated * (e.g. after reflowing line L and before reflowing line L+1); to handle * these temporary violations, we say that there is a break betwen i and i+1 * if a break is specified after i OR a break is specified before i+1. * * This can change textrun geometry! The existence of a linebreak can affect * the advance width of the cluster before the break (when kerning) or the * geometry of one cluster before the break or any number of clusters * after the break. (The one-cluster-before-the-break limit is somewhat * arbitrary; if some scripts require breaking it, then we need to * alter nsTextFrame::TrimTrailingWhitespace, perhaps drastically becase * it could affect the layout of frames before it...) * * We return true if glyphs or geometry changed, false otherwise. This * function is virtual so that gfxTextRun subclasses can reshape * properly. * * @param aAdvanceWidthDelta if non-null, returns the change in advance * width of the given range. */ virtual PRBool SetLineBreaks(PRUint32 aStart, PRUint32 aLength, PRBool aLineBreakBefore, PRBool aLineBreakAfter, gfxFloat *aAdvanceWidthDelta, gfxContext *aRefContext); /** * Finds the longest substring that will fit into the given width. * Uses GetHyphenationBreaks and GetSpacing from aBreakProvider. * Guarantees the following: * -- 0 <= result <= aMaxLength * -- result is the maximal value of N such that either * N < aMaxLength && line break at N && GetAdvanceWidth(aStart, N) <= aWidth * OR N < aMaxLength && hyphen break at N && GetAdvanceWidth(aStart, N) + GetHyphenWidth() <= aWidth * OR N == aMaxLength && GetAdvanceWidth(aStart, N) <= aWidth * where GetAdvanceWidth assumes the effect of * SetLineBreaks(aStart, N, aLineBreakBefore, N < aMaxLength, aProvider) * -- if no such N exists, then result is the smallest N such that * N < aMaxLength && line break at N * OR N < aMaxLength && hyphen break at N * OR N == aMaxLength * * The call has the effect of * SetLineBreaks(aStart, result, aLineBreakBefore, result < aMaxLength, aProvider) * and the returned metrics and the invariants above reflect this. * * @param aMaxLength this can be PR_UINT32_MAX, in which case the length used * is up to the end of the string * @param aLineBreakBefore set to true if and only if there is an actual * line break at the start of this string. * @param aSuppressInitialBreak if true, then we assume there is no possible * linebreak before aStart. If false, then we will check the internal * line break opportunity state before deciding whether to return 0 as the * character to break before. * @param aTrimWhitespace if non-null, then we allow a trailing run of * spaces to be trimmed; the width of the space(s) will not be included in * the measured string width for comparison with the limit aWidth, and * trimmed spaces will not be included in returned metrics. The width * of the trimmed spaces will be returned in aTrimWhitespace. * Trimmed spaces are still counted in the "characters fit" result. * @param aMetrics if non-null, we fill this in for the returned substring. * If a hyphenation break was used, the hyphen is NOT included in the returned metrics. * @param aTightBoundingBox if true, we make the bounding box in aMetrics tight * @param aRefContextForTightBoundingBox a reference context to get the * tight bounding box, if aTightBoundingBox is true * @param aUsedHyphenation if non-null, records if we selected a hyphenation break * @param aLastBreak if non-null and result is aMaxLength, we set this to * the maximal N such that * N < aMaxLength && line break at N && GetAdvanceWidth(aStart, N) <= aWidth * OR N < aMaxLength && hyphen break at N && GetAdvanceWidth(aStart, N) + GetHyphenWidth() <= aWidth * or PR_UINT32_MAX if no such N exists, where GetAdvanceWidth assumes * the effect of * SetLineBreaks(aStart, N, aLineBreakBefore, N < aMaxLength, aProvider) * * @param aCanWordWrap true if we can break between any two grapheme * clusters. This is set by word-wrap: break-word * * @param aBreakPriority in/out the priority of the break opportunity * saved in the line. If we are prioritizing break opportunities, we will * not set a break with a lower priority. @see gfxBreakPriority. * * Note that negative advance widths are possible especially if negative * spacing is provided. */ PRUint32 BreakAndMeasureText(PRUint32 aStart, PRUint32 aMaxLength, PRBool aLineBreakBefore, gfxFloat aWidth, PropertyProvider *aProvider, PRBool aSuppressInitialBreak, gfxFloat *aTrimWhitespace, Metrics *aMetrics, PRBool aTightBoundingBox, gfxContext *aRefContextForTightBoundingBox, PRBool *aUsedHyphenation, PRUint32 *aLastBreak, PRBool aCanWordWrap, gfxBreakPriority *aBreakPriority); /** * Update the reference context. * XXX this is a hack. New text frame does not call this. Use only * temporarily for old text frame. */ void SetContext(gfxContext *aContext) {} // Utility getters PRBool IsRightToLeft() const { return (mFlags & gfxTextRunFactory::TEXT_IS_RTL) != 0; } gfxFloat GetDirection() const { return (mFlags & gfxTextRunFactory::TEXT_IS_RTL) ? -1.0 : 1.0; } void *GetUserData() const { return mUserData; } void SetUserData(void *aUserData) { mUserData = aUserData; } PRUint32 GetFlags() const { return mFlags; } void SetFlagBits(PRUint32 aFlags) { NS_ASSERTION(!(aFlags & ~gfxTextRunFactory::SETTABLE_FLAGS), "Only user flags should be mutable"); mFlags |= aFlags; } void ClearFlagBits(PRUint32 aFlags) { NS_ASSERTION(!(aFlags & ~gfxTextRunFactory::SETTABLE_FLAGS), "Only user flags should be mutable"); mFlags &= ~aFlags; } const gfxSkipChars& GetSkipChars() const { return mSkipChars; } PRUint32 GetAppUnitsPerDevUnit() const { return mAppUnitsPerDevUnit; } gfxFontGroup *GetFontGroup() const { return mFontGroup; } const PRUint8 *GetText8Bit() const { return (mFlags & gfxTextRunFactory::TEXT_IS_8BIT) ? mText.mSingle : nsnull; } const PRUnichar *GetTextUnicode() const { return (mFlags & gfxTextRunFactory::TEXT_IS_8BIT) ? nsnull : mText.mDouble; } const void *GetTextAt(PRUint32 aIndex) { return (mFlags & gfxTextRunFactory::TEXT_IS_8BIT) ? static_cast(mText.mSingle + aIndex) : static_cast(mText.mDouble + aIndex); } const PRUnichar GetChar(PRUint32 i) const { return (mFlags & gfxTextRunFactory::TEXT_IS_8BIT) ? mText.mSingle[i] : mText.mDouble[i]; } PRUint32 GetHashCode() const { return mHashCode; } void SetHashCode(PRUint32 aHash) { mHashCode = aHash; } // Call this, don't call "new gfxTextRun" directly. This does custom // allocation and initialization static gfxTextRun *Create(const gfxTextRunFactory::Parameters *aParams, const void *aText, PRUint32 aLength, gfxFontGroup *aFontGroup, PRUint32 aFlags); // Clone this textrun, according to the given parameters. This textrun's // glyph data is copied, so the text and length must be the same as this // textrun's. If there's a problem, return null. Actual linebreaks will // be set as per aParams; there will be no potential linebreaks. // If aText is not persistent (aFlags & TEXT_IS_PERSISTENT), the // textrun will copy it. virtual gfxTextRun *Clone(const gfxTextRunFactory::Parameters *aParams, const void *aText, PRUint32 aLength, gfxFontGroup *aFontGroup, PRUint32 aFlags); /** * This class records the information associated with a character in the * input string. It's optimized for the case where there is one glyph * representing that character alone. * * A character can have zero or more associated glyphs. Each glyph * has an advance width and an x and y offset. * A character may be the start of a cluster. * A character may be the start of a ligature group. * A character can be "missing", indicating that the system is unable * to render the character. * * All characters in a ligature group conceptually share all the glyphs * associated with the characters in a group. */ class CompressedGlyph { public: CompressedGlyph() { mValue = 0; } enum { // Indicates that a cluster and ligature group starts at this // character; this character has a single glyph with a reasonable // advance and zero offsets. A "reasonable" advance // is one that fits in the available bits (currently 14) (specified // in appunits). FLAG_IS_SIMPLE_GLYPH = 0x80000000U, // Indicates that a linebreak is allowed before this character FLAG_CAN_BREAK_BEFORE = 0x40000000U, // The advance is stored in appunits ADVANCE_MASK = 0x3FFF0000U, ADVANCE_SHIFT = 16, GLYPH_MASK = 0x0000FFFFU, // Non-simple glyphs may or may not have glyph data in the // corresponding mDetailedGlyphs entry. They have the following // flag bits: // When NOT set, indicates that this character corresponds to a // missing glyph and should be skipped (or possibly, render the character // Unicode value in some special way). If there are glyphs, // the mGlyphID is actually the UTF16 character code. The bit is // inverted so we can memset the array to zero to indicate all missing. FLAG_NOT_MISSING = 0x01, FLAG_NOT_CLUSTER_START = 0x02, FLAG_NOT_LIGATURE_GROUP_START = 0x04, FLAG_LOW_SURROGATE = 0x08, GLYPH_COUNT_MASK = 0x00FFFF00U, GLYPH_COUNT_SHIFT = 8 }; // "Simple glyphs" have a simple glyph ID, simple advance and their // x and y offsets are zero. Also the glyph extents do not overflow // the font-box defined by the font ascent, descent and glyph advance width. // These case is optimized to avoid storing DetailedGlyphs. // Returns true if the glyph ID aGlyph fits into the compressed representation static PRBool IsSimpleGlyphID(PRUint32 aGlyph) { return (aGlyph & GLYPH_MASK) == aGlyph; } // Returns true if the advance aAdvance fits into the compressed representation. // aAdvance is in appunits. static PRBool IsSimpleAdvance(PRUint32 aAdvance) { return (aAdvance & (ADVANCE_MASK >> ADVANCE_SHIFT)) == aAdvance; } PRBool IsSimpleGlyph() const { return (mValue & FLAG_IS_SIMPLE_GLYPH) != 0; } PRUint32 GetSimpleAdvance() const { return (mValue & ADVANCE_MASK) >> ADVANCE_SHIFT; } PRUint32 GetSimpleGlyph() const { return mValue & GLYPH_MASK; } PRBool IsMissing() const { return (mValue & (FLAG_NOT_MISSING|FLAG_IS_SIMPLE_GLYPH)) == 0; } PRBool IsLowSurrogate() const { return (mValue & (FLAG_LOW_SURROGATE|FLAG_IS_SIMPLE_GLYPH)) == FLAG_LOW_SURROGATE; } PRBool IsClusterStart() const { return (mValue & FLAG_IS_SIMPLE_GLYPH) || !(mValue & FLAG_NOT_CLUSTER_START); } PRBool IsLigatureGroupStart() const { return (mValue & FLAG_IS_SIMPLE_GLYPH) || !(mValue & FLAG_NOT_LIGATURE_GROUP_START); } PRBool CanBreakBefore() const { return (mValue & FLAG_CAN_BREAK_BEFORE) != 0; } // Returns FLAG_CAN_BREAK_BEFORE if the setting changed, 0 otherwise PRUint32 SetCanBreakBefore(PRBool aCanBreakBefore) { NS_ASSERTION(aCanBreakBefore == PR_FALSE || aCanBreakBefore == PR_TRUE, "Bogus break-before value!"); PRUint32 breakMask = aCanBreakBefore*FLAG_CAN_BREAK_BEFORE; PRUint32 toggle = breakMask ^ (mValue & FLAG_CAN_BREAK_BEFORE); mValue ^= toggle; return toggle; } CompressedGlyph& SetSimpleGlyph(PRUint32 aAdvanceAppUnits, PRUint32 aGlyph) { NS_ASSERTION(IsSimpleAdvance(aAdvanceAppUnits), "Advance overflow"); NS_ASSERTION(IsSimpleGlyphID(aGlyph), "Glyph overflow"); mValue = (mValue & FLAG_CAN_BREAK_BEFORE) | FLAG_IS_SIMPLE_GLYPH | (aAdvanceAppUnits << ADVANCE_SHIFT) | aGlyph; return *this; } CompressedGlyph& SetComplex(PRBool aClusterStart, PRBool aLigatureStart, PRUint32 aGlyphCount) { mValue = (mValue & FLAG_CAN_BREAK_BEFORE) | FLAG_NOT_MISSING | (aClusterStart ? 0 : FLAG_NOT_CLUSTER_START) | (aLigatureStart ? 0 : FLAG_NOT_LIGATURE_GROUP_START) | (aGlyphCount << GLYPH_COUNT_SHIFT); return *this; } /** * Missing glyphs are treated as cluster and ligature group starts. */ CompressedGlyph& SetMissing(PRUint32 aGlyphCount) { mValue = (mValue & FLAG_CAN_BREAK_BEFORE) | (aGlyphCount << GLYPH_COUNT_SHIFT); return *this; } /** * Low surrogates don't have any glyphs and are not the start of * a cluster or ligature group. */ CompressedGlyph& SetLowSurrogate() { mValue = (mValue & FLAG_CAN_BREAK_BEFORE) | FLAG_NOT_MISSING | FLAG_LOW_SURROGATE; return *this; } PRUint32 GetGlyphCount() const { NS_ASSERTION(!IsSimpleGlyph(), "Expected non-simple-glyph"); return (mValue & GLYPH_COUNT_MASK) >> GLYPH_COUNT_SHIFT; } private: PRUint32 mValue; }; /** * When the glyphs for a character don't fit into a CompressedGlyph record * in SimpleGlyph format, we use an array of DetailedGlyphs instead. */ struct DetailedGlyph { /** The glyphID, or the Unicode character * if this is a missing glyph */ PRUint32 mGlyphID; /** The advance, x-offset and y-offset of the glyph, in appunits * mAdvance is in the text direction (RTL or LTR) * mXOffset is always from left to right * mYOffset is always from bottom to top */ PRInt32 mAdvance; float mXOffset, mYOffset; }; // The text is divided into GlyphRuns as necessary struct GlyphRun { nsRefPtr mFont; // never null PRUint32 mCharacterOffset; // into original UTF16 string }; class THEBES_API GlyphRunIterator { public: GlyphRunIterator(gfxTextRun *aTextRun, PRUint32 aStart, PRUint32 aLength) : mTextRun(aTextRun), mStartOffset(aStart), mEndOffset(aStart + aLength) { mNextIndex = mTextRun->FindFirstGlyphRunContaining(aStart); } PRBool NextRun(); GlyphRun *GetGlyphRun() { return mGlyphRun; } PRUint32 GetStringStart() { return mStringStart; } PRUint32 GetStringEnd() { return mStringEnd; } private: gfxTextRun *mTextRun; GlyphRun *mGlyphRun; PRUint32 mStringStart; PRUint32 mStringEnd; PRUint32 mNextIndex; PRUint32 mStartOffset; PRUint32 mEndOffset; }; class GlyphRunOffsetComparator { public: PRBool Equals(const GlyphRun& a, const GlyphRun& b) const { return a.mCharacterOffset == b.mCharacterOffset; } PRBool LessThan(const GlyphRun& a, const GlyphRun& b) const { return a.mCharacterOffset < b.mCharacterOffset; } }; friend class GlyphRunIterator; friend class FontSelector; // API for setting up the textrun glyphs. Should only be called by // things that construct textruns. /** * Record every character that is the second half of a surrogate pair. * This should be called after creating a Unicode textrun. */ void RecordSurrogates(const PRUnichar *aString); /** * We've found a run of text that should use a particular font. Call this * only during initialization when font substitution has been computed. * Call it before setting up the glyphs for the characters in this run; * SetMissingGlyph requires that the correct glyphrun be installed. * * If aForceNewRun, a new glyph run will be added, even if the * previously added run uses the same font. If glyph runs are * added out of strictly increasing aStartCharIndex order (via * force), then SortGlyphRuns must be called after all glyph runs * are added before any further operations are performed with this * TextRun. */ nsresult AddGlyphRun(gfxFont *aFont, PRUint32 aStartCharIndex, PRBool aForceNewRun = PR_FALSE); void ResetGlyphRuns() { mGlyphRuns.Clear(); } void SortGlyphRuns(); // Call the following glyph-setters during initialization or during reshaping // only. It is OK to overwrite existing data for a character. /** * Set the glyph data for a character. aGlyphs may be null if aGlyph is a * simple glyph or has no associated glyphs. If non-null the data is copied, * the caller retains ownership. */ void SetSimpleGlyph(PRUint32 aCharIndex, CompressedGlyph aGlyph) { NS_ASSERTION(aGlyph.IsSimpleGlyph(), "Should be a simple glyph here"); if (mCharacterGlyphs) { mCharacterGlyphs[aCharIndex] = aGlyph; } if (mDetailedGlyphs) { mDetailedGlyphs[aCharIndex] = nsnull; } } void SetGlyphs(PRUint32 aCharIndex, CompressedGlyph aGlyph, const DetailedGlyph *aGlyphs); void SetMissingGlyph(PRUint32 aCharIndex, PRUint32 aUnicodeChar); void SetSpaceGlyph(gfxFont *aFont, gfxContext *aContext, PRUint32 aCharIndex); /** * Prefetch all the glyph extents needed to ensure that Measure calls * on this textrun with aTightBoundingBox false will succeed. Note * that some glyph extents might not be fetched due to OOM or other * errors. */ void FetchGlyphExtents(gfxContext *aRefContext); // API for access to the raw glyph data, needed by gfxFont::Draw // and gfxFont::GetBoundingBox const CompressedGlyph *GetCharacterGlyphs() { return mCharacterGlyphs; } const DetailedGlyph *GetDetailedGlyphs(PRUint32 aCharIndex) { return mDetailedGlyphs ? mDetailedGlyphs[aCharIndex].get() : nsnull; } PRBool HasDetailedGlyphs() { return mDetailedGlyphs.get() != nsnull; } PRUint32 CountMissingGlyphs(); const GlyphRun *GetGlyphRuns(PRUint32 *aNumGlyphRuns) { *aNumGlyphRuns = mGlyphRuns.Length(); return mGlyphRuns.Elements(); } // Returns the index of the GlyphRun containing the given offset. // Returns mGlyphRuns.Length() when aOffset is mCharacterCount. PRUint32 FindFirstGlyphRunContaining(PRUint32 aOffset); // Copy glyph data for a range of characters from aSource to this // textrun. If aStealData is true then we actually steal the glyph data, // setting the data in aSource to "missing". aDest should be in the last // glyphrun. virtual void CopyGlyphDataFrom(gfxTextRun *aSource, PRUint32 aStart, PRUint32 aLength, PRUint32 aDest, PRBool aStealData); nsExpirationState *GetExpirationState() { return &mExpirationState; } struct LigatureData { // textrun offsets of the start and end of the containing ligature PRUint32 mLigatureStart; PRUint32 mLigatureEnd; // appunits advance to the start of the ligature part within the ligature; // never includes any spacing gfxFloat mPartAdvance; // appunits width of the ligature part; includes before-spacing // when the part is at the start of the ligature, and after-spacing // when the part is as the end of the ligature gfxFloat mPartWidth; PRPackedBool mClipBeforePart; PRPackedBool mClipAfterPart; }; #ifdef DEBUG // number of entries referencing this textrun in the gfxTextRunWordCache PRUint32 mCachedWords; void Dump(FILE* aOutput); #endif // post-process glyph advances to deal with synthetic bolding void AdjustAdvancesForSyntheticBold(PRUint32 aStart, PRUint32 aLength); protected: // Allocates extra space for the CompressedGlyph array and the text // (if needed) void *operator new(size_t aSize, PRUint32 aLength, PRUint32 aFlags); /** * Initializes the textrun to blank. * @param aObjectSize the size of the object; this lets us fine * where our CompressedGlyph array and string have been allocated */ gfxTextRun(const gfxTextRunFactory::Parameters *aParams, const void *aText, PRUint32 aLength, gfxFontGroup *aFontGroup, PRUint32 aFlags, PRUint32 aObjectSize); private: // **** general helpers **** // Allocate aCount DetailedGlyphs for the given index DetailedGlyph *AllocateDetailedGlyphs(PRUint32 aCharIndex, PRUint32 aCount); // Spacing for characters outside the range aSpacingStart/aSpacingEnd // is assumed to be zero; such characters are not passed to aProvider. // This is useful to protect aProvider from being passed character indices // it is not currently able to handle. PRBool GetAdjustedSpacingArray(PRUint32 aStart, PRUint32 aEnd, PropertyProvider *aProvider, PRUint32 aSpacingStart, PRUint32 aSpacingEnd, nsTArray *aSpacing); // **** ligature helpers **** // (Platforms do the actual ligaturization, but we need to do a bunch of stuff // to handle requests that begin or end inside a ligature) // if aProvider is null then mBeforeSpacing and mAfterSpacing are set to zero LigatureData ComputeLigatureData(PRUint32 aPartStart, PRUint32 aPartEnd, PropertyProvider *aProvider); gfxFloat ComputePartialLigatureWidth(PRUint32 aPartStart, PRUint32 aPartEnd, PropertyProvider *aProvider); void DrawPartialLigature(gfxFont *aFont, gfxContext *aCtx, PRUint32 aStart, PRUint32 aEnd, const gfxRect *aDirtyRect, gfxPoint *aPt, PropertyProvider *aProvider); // Advance aStart to the start of the nearest ligature; back up aEnd // to the nearest ligature end; may result in *aStart == *aEnd void ShrinkToLigatureBoundaries(PRUint32 *aStart, PRUint32 *aEnd); // result in appunits gfxFloat GetPartialLigatureWidth(PRUint32 aStart, PRUint32 aEnd, PropertyProvider *aProvider); void AccumulatePartialLigatureMetrics(gfxFont *aFont, PRUint32 aStart, PRUint32 aEnd, PRBool aTight, gfxContext *aRefContext, PropertyProvider *aProvider, Metrics *aMetrics); // **** measurement helper **** void AccumulateMetricsForRun(gfxFont *aFont, PRUint32 aStart, PRUint32 aEnd, PRBool aTight, gfxContext *aRefContext, PropertyProvider *aProvider, PRUint32 aSpacingStart, PRUint32 aSpacingEnd, Metrics *aMetrics); // **** drawing helper **** void DrawGlyphs(gfxFont *aFont, gfxContext *aContext, PRBool aDrawToPath, gfxPoint *aPt, PRUint32 aStart, PRUint32 aEnd, PropertyProvider *aProvider, PRUint32 aSpacingStart, PRUint32 aSpacingEnd); // All our glyph data is in logical order, not visual. // mCharacterGlyphs is allocated fused with this object. We need a pointer // to it because gfxTextRun subclasses exist with extra fields, so we don't // know where it starts without a virtual method call or an explicit pointer. CompressedGlyph* mCharacterGlyphs; nsAutoArrayPtr > mDetailedGlyphs; // only non-null if needed // XXX this should be changed to a GlyphRun plus a maybe-null GlyphRun*, // for smaller size especially in the super-common one-glyphrun case nsAutoTArray mGlyphRuns; // When TEXT_IS_8BIT is set, we use mSingle, otherwise we use mDouble. // When TEXT_IS_PERSISTENT is set, we don't own the text, otherwise we // own the text. When we own the text, it's allocated fused with this // object, so it need not be deleted. // This text is not null-terminated. union { const PRUint8 *mSingle; const PRUnichar *mDouble; } mText; void *mUserData; gfxFontGroup *mFontGroup; // addrefed gfxSkipChars mSkipChars; nsExpirationState mExpirationState; PRUint32 mAppUnitsPerDevUnit; PRUint32 mFlags; PRUint32 mCharacterCount; PRUint32 mHashCode; }; class THEBES_API gfxFontGroup : public gfxTextRunFactory { public: gfxFontGroup(const nsAString& aFamilies, const gfxFontStyle *aStyle); virtual ~gfxFontGroup() { mFonts.Clear(); } virtual gfxFont *GetFontAt(PRInt32 i) { return static_cast(mFonts[i]); } virtual PRUint32 FontListLength() const { return mFonts.Length(); } PRBool Equals(const gfxFontGroup& other) const { return mFamilies.Equals(other.mFamilies) && mStyle.Equals(other.mStyle); } const gfxFontStyle *GetStyle() const { return &mStyle; } virtual gfxFontGroup *Copy(const gfxFontStyle *aStyle) = 0; /** * The listed characters should not be passed in to MakeTextRun and should * be treated as invisible and zero-width. */ static PRBool IsInvalidChar(PRUnichar ch) { if (ch >= 32) { return ch == 0x0085/*NEL*/ || ((ch & 0xFF00) == 0x2000 /* Unicode control character */ && (ch == 0x200B/*ZWSP*/ || ch == 0x2028/*LSEP*/ || ch == 0x2029/*PSEP*/ || IS_BIDI_CONTROL_CHAR(ch))); } // We could just blacklist all control characters, but it seems better // to only blacklist the ones we know cause problems for native font // engines. return ch == 0x0B || ch == '\t' || ch == '\r' || ch == '\n' || ch == '\f' || (ch >= 0x1c && ch <= 0x1f); } /** * Make a textrun for an empty string. This is fast; if you call it, * don't bother caching the result. */ gfxTextRun *MakeEmptyTextRun(const Parameters *aParams, PRUint32 aFlags); /** * Make a textrun for a single ASCII space. This is fast; if you call it, * don't bother caching the result. */ gfxTextRun *MakeSpaceTextRun(const Parameters *aParams, PRUint32 aFlags); /** * Make a textrun for a given string. * If aText is not persistent (aFlags & TEXT_IS_PERSISTENT), the * textrun will copy it. * This calls FetchGlyphExtents on the textrun. */ virtual gfxTextRun *MakeTextRun(const PRUnichar *aString, PRUint32 aLength, const Parameters *aParams, PRUint32 aFlags) = 0; /** * Make a textrun for a given string. * If aText is not persistent (aFlags & TEXT_IS_PERSISTENT), the * textrun will copy it. * This calls FetchGlyphExtents on the textrun. */ virtual gfxTextRun *MakeTextRun(const PRUint8 *aString, PRUint32 aLength, const Parameters *aParams, PRUint32 aFlags) = 0; /* helper function for splitting font families on commas and * calling a function for each family to fill the mFonts array */ typedef PRBool (*FontCreationCallback) (const nsAString& aName, const nsACString& aGenericName, void *closure); static PRBool ForEachFont(const nsAString& aFamilies, const nsACString& aLangGroup, FontCreationCallback fc, void *closure); PRBool ForEachFont(FontCreationCallback fc, void *closure); const nsString& GetFamilies() { return mFamilies; } // This returns the preferred underline for this font group. // Some CJK fonts have wrong underline offset in its metrics. // If this group has such "bad" font, each platform's gfxFontGroup initialized mUnderlineOffset. // The value should be lower value of first font's metrics and the bad font's metrics. // Otherwise, this returns from first font's metrics. enum { UNDERLINE_OFFSET_NOT_SET = PR_INT16_MAX }; gfxFloat GetUnderlineOffset() { if (mUnderlineOffset == UNDERLINE_OFFSET_NOT_SET) mUnderlineOffset = GetFontAt(0)->GetMetrics().underlineOffset; return mUnderlineOffset; } protected: nsString mFamilies; gfxFontStyle mStyle; nsTArray< nsRefPtr > mFonts; gfxFloat mUnderlineOffset; // Init this font group's font metrics. If there no bad fonts, you don't need to call this. // But if there are one or more bad fonts which have bad underline offset, // you should call this with the *first* bad font. void InitMetricsForBadFont(gfxFont* aBadFont); /* If aResolveGeneric is true, then CSS/Gecko generic family names are * replaced with preferred fonts. * * If aResolveFontName is true then fc() is called only for existing fonts * and with actual font names. If false then fc() is called with each * family name in aFamilies (after resolving CSS/Gecko generic family names * if aResolveGeneric). */ static PRBool ForEachFontInternal(const nsAString& aFamilies, const nsACString& aLangGroup, PRBool aResolveGeneric, PRBool aResolveFontName, FontCreationCallback fc, void *closure); static PRBool FontResolverProc(const nsAString& aName, void *aClosure); }; #endif