|
//-----------------------------------------------------------------------
//
// Microsoft Windows Client Platform
// Copyright (C) Microsoft Corporation
//
// File: FullTextState.cs
//
// Contents: Formatting state of full text
//
// Created: 2-25-2004 Worachai Chaoweeraprasit (wchao)
// History: 1-29-2005 (wchao) refactor drawing state out, keep full text
// only for formatting state
//
//------------------------------------------------------------------------
using System;
using System.Security;
using System.Globalization;
using System.Diagnostics;
using System.Runtime.InteropServices;
using System.Windows;
using System.Windows.Media;
using System.Windows.Media.TextFormatting;
using MS.Internal.Shaping;
using MS.Internal.Generic;
namespace MS.Internal.TextFormatting
{
/// <summary>
/// Formatting state of full text
/// </summary>
internal sealed class FullTextState
{
private TextStore _store; // formatting store for main text
private TextStore _markerStore; // store specifically for marker
private StatusFlags _statusFlags; // status flags
private int _cpMeasured; // number of CP successfully measured and fit within column width
private int _lscpHyphenationLookAhead; // LSCP after the last character examined by the hyphenation code.
private bool _isSideways;
[Flags]
private enum StatusFlags
{
None = 0,
VerticalAdjust = 0x00000001, // vertical adjustment on some runs required
ForceWrap = 0x00000002, // force wrap
KeepState = 0x00000040, // state should be kept in the line
}
/// <summary>
/// A word smaller than seven characters does not require hyphenation. This is based on
/// observation in real Western books and publications. Advanced research on readability
/// seems to support this finding. It suggests that the 'fovea' which is the center
/// point of our vision, can only see three to four letters before and after the center
/// of a word.
/// </summary>
/// <remarks>
/// "It has been known for over 100 years that when we read, our eyes don't move smoothly
/// across the page, but rather make discrete jumps from word to word. We fixate on a word
/// for a period of time, roughly 200-250ms, then make a ballistic movement to another word.
/// These movements are called saccades and usually take 20-35ms. Most saccades are forward
/// movements from 7 to 9 letters.
/// ...
/// During a single fixation, there is a limit to the amount of information that can be
/// recognized. The fovea, which is the clear center point of our vision, can only see
/// three to four letters to the left and right of fixation at normal reading distances."
///
/// ["The Science of Word Recognition" by Kevin Larson; July 2004]
/// </remarks>
private const int MinCchWordToHyphenate = 7;
/// <summary>
/// Create a fulltext state object from formatting info for subsequent fulltext formatting
/// e.g. fulltext line, mix/max computation, potential breakpoints for optimal paragraph formatting.
/// </summary>
internal static FullTextState Create(
FormatSettings settings,
int cpFirst,
int finiteFormatWidth
)
{
// prepare text stores
TextStore store = new TextStore(
settings,
cpFirst,
0,
settings.GetFormatWidth(finiteFormatWidth)
);
ParaProp pap = settings.Pap;
TextStore markerStore = null;
if( pap.FirstLineInParagraph
&& pap.TextMarkerProperties != null
&& pap.TextMarkerProperties.TextSource != null)
{
// create text store specifically for marker
markerStore = new TextStore(
// create specialized settings for marker store e.g. with marker text source
new FormatSettings(
settings.Formatter,
pap.TextMarkerProperties.TextSource,
new TextRunCacheImp(), // no cross-call run cache available for marker store
pap, // marker by default use the same paragraph properties
null, // not consider previousLineBreak
true, // isSingleLineFormatting
settings.TextFormattingMode,
settings.IsSideways
),
0, // marker store always started with cp == 0
TextStore.LscpFirstMarker, // first lscp value for marker text
Constants.IdealInfiniteWidth // formatWidth
);
}
// construct a new fulltext state object
return new FullTextState(store, markerStore, settings.IsSideways);
}
/// <summary>
/// Construct full text state for formatting
/// </summary>
private FullTextState(
TextStore store,
TextStore markerStore,
bool isSideways
)
{
_isSideways = isSideways;
_store = store;
_markerStore = markerStore;
}
/// <summary>
/// Number of client CP for which we know the nominal widths fit in the
/// available margin (i.e., because we've measured them).
/// </summary>
internal int CpMeasured
{
get
{
return _cpMeasured;
}
set
{
_cpMeasured = value;
}
}
/// <summary>
/// LSCP immediately after the last LSCP examined by hyphenation code while being run.
/// This value is used to calculate a more precise DependentLength value for the line
/// ended by automatic hyphenation.
/// </summary>
internal int LscpHyphenationLookAhead
{
get
{
return _lscpHyphenationLookAhead;
}
}
internal TextFormattingMode TextFormattingMode
{
get
{
return Formatter.TextFormattingMode;
}
}
internal bool IsSideways
{
get
{
return _isSideways;
}
}
/// <summary>
/// Set tab stops
/// </summary>
/// <SecurityNote>
/// Critical - as this calls TextFormatterContext.SetTabs which is Critical in four
/// different places.
/// Safe - as the buffers are all created in the same function and the count
/// passed in the same as used for creating the buffer.
/// </SecurityNote>
[SecurityCritical, SecurityTreatAsSafe]
internal void SetTabs(TextFormatterContext context)
{
unsafe
{
ParaProp pap = _store.Pap;
FormatSettings settings = _store.Settings;
// set up appropriate tab stops
int incrementalTab = TextFormatterImp.RealToIdeal(pap.DefaultIncrementalTab);
int lsTbdCount = pap.Tabs != null ? pap.Tabs.Count : 0;
LsTbd[] lsTbds;
if (_markerStore != null)
{
if (pap.Tabs != null && pap.Tabs.Count > 0)
{
lsTbdCount = pap.Tabs.Count + 1;
lsTbds = new LsTbd[lsTbdCount];
lsTbds[0].ur = settings.TextIndent; // marker requires a tab stop at text start position
fixed (LsTbd* plsTbds = &lsTbds[1])
{
CreateLsTbds(pap, plsTbds, lsTbdCount - 1);
context.SetTabs(incrementalTab, plsTbds - 1, lsTbdCount);
}
}
else
{
LsTbd markerRequiredLsTbd = new LsTbd();
markerRequiredLsTbd.ur = settings.TextIndent; // marker requires a tab stop at text start position
context.SetTabs(incrementalTab, &markerRequiredLsTbd, 1);
}
}
else
{
if (pap.Tabs != null && pap.Tabs.Count > 0)
{
lsTbds = new LsTbd[lsTbdCount];
fixed (LsTbd* plsTbds = &lsTbds[0])
{
CreateLsTbds(pap, plsTbds, lsTbdCount);
context.SetTabs(incrementalTab, plsTbds, lsTbdCount);
}
}
else
{
// work with only incremental tab
context.SetTabs(incrementalTab, null, 0);
}
}
}
}
/// <summary>
/// Fill a fixed buffer of LsTbd with
/// </summary>
/// <SecurityNote>
/// Critical - This method writes into unmanaged array.
/// </SecurityNote>
[SecurityCritical]
private unsafe void CreateLsTbds(
ParaProp pap,
LsTbd* plsTbds,
int lsTbdCount
)
{
for (int i = 0; i < lsTbdCount; i++)
{
TextTabProperties tab = (TextTabProperties)pap.Tabs[i];
plsTbds[i].lskt = Convert.LsKTabFromTabAlignment(tab.Alignment);
plsTbds[i].ur = TextFormatterImp.RealToIdeal(tab.Location);
if (tab.TabLeader != 0)
{
// Note: LS does not currently support surrogate character as tab leader and aligning character
plsTbds[i].wchTabLeader = (char)tab.TabLeader;
// tab leader requires state at display time for tab leader width fetching
_statusFlags |= StatusFlags.KeepState;
}
plsTbds[i].wchCharTab = (char)tab.AligningCharacter;
}
}
/// <summary>
/// Get distance from the start of main text to the end of marker
/// </summary>
/// <remarks>
/// Positive distance is filtered out. Marker overlapping the main text is not supported.
/// </remarks>
internal int GetMainTextToMarkerIdealDistance()
{
if (_markerStore != null)
{
return Math.Min(0, TextFormatterImp.RealToIdeal(_markerStore.Pap.TextMarkerProperties.Offset) - _store.Settings.TextIndent);
}
return 0;
}
/// <summary>
/// Map LSCP to host CP, and return the last LSRun
/// before the specified limit.
/// </summary>
internal LSRun CountText(
int lscpLim,
int cpFirst,
out int count
)
{
LSRun lastRun = null;
count = 0;
int lsccp = lscpLim - _store.CpFirst;
Debug.Assert(lsccp > 0, "Zero-length text line!");
foreach (Span span in _store.PlsrunVector)
{
if (lsccp <= 0)
break;
Plsrun plsrun = (Plsrun)span.element;
// There should be no marker runs in _plsrunVector.
Debug.Assert(!TextStore.IsMarker(plsrun));
// Is it a normal, non-static, LSRun?
if (plsrun >= Plsrun.FormatAnchor)
{
// Get the run and remember the last run.
lastRun = _store.GetRun(plsrun);
// Accumulate the length.
int cpRun = lastRun.Length;
if (cpRun > 0)
{
if (lsccp < span.length && cpRun == span.length)
{
count += lsccp;
break;
}
count += cpRun;
}
}
lsccp -= span.length;
}
// make char count relative to cpFirst as the cpFirst of this metrics may not
// be the same as the cpFirst of the store in optimal paragraph formatting.
count = count - cpFirst + _store.CpFirst;
return lastRun;
}
/// <summary>
/// Convert the specified external CP to LSCP corresponding to a possible break position for optimal break.
/// </summary>
/// <remarks>
/// There is a generic issue that one CP could map to multiple LSCPs when it comes to
/// lsrun that occupies no actual CP. Such lsrun is generated by line layout during
/// formatting to accomplsih specific purpose i.e. run representing open and close
/// reverse object or a fake-linebreak lsrun.
///
/// According to SergeyGe and Antons, LS will never breaks immediately after open reverse
/// or immediately before close reverse. It also never break before a linebreak character.
///
/// Therefore, it is safe to make an assumption here that one CP will indeed map to one
/// LSCP given being the LSCP before the open reverse and the LSCP after the close reverse.
/// Never the vice-versa.
///
/// This is the reason why the loop inside this method may overread the PLSRUN span by
/// one span at the end. The loop is designed to skip over all the lsruns with zero CP
/// which is not the open reverse.
///
/// This logic is exactly the same as the one used by FullTextLine.PrefetchLSRuns. Any
/// attempt to change it needs to be thoroughly reviewed. The same logic is to be applied
/// accordingly on PrefetchLSRuns.
///
/// [Wchao, 5-24-2005]
///
/// </remarks>
internal int GetBreakpointInternalCp(int cp)
{
int ccp = cp - _store.CpFirst;
int lscp = _store.CpFirst;
int ccpCurrent = 0;
SpanVector plsrunVector = _store.PlsrunVector;
LSRun lsrun;
int i = 0;
int lastSpanLength = 0;
int lastRunLength = 0;
do
{
Span span = plsrunVector[i];
Plsrun plsrun = (Plsrun)span.element;
lsrun = _store.GetRun(plsrun);
if (ccp == ccpCurrent && lsrun.Type == Plsrun.Reverse)
{
break;
}
lastSpanLength = span.length;
lastRunLength = (plsrun >= Plsrun.FormatAnchor ? lsrun.Length : 0);
lscp += lastSpanLength;
ccpCurrent += lastRunLength;
} while ( ++i < plsrunVector.Count
&& lsrun.Type != Plsrun.ParaBreak
&& ccp >= ccpCurrent
);
// Since we may overread the span vector by one span,
// we need to subtract the accumulated lscp by the number of cp we may have overread.
if (ccpCurrent == ccp || lastSpanLength == lastRunLength)
return lscp - ccpCurrent + ccp;
Invariant.Assert(ccpCurrent - ccp == lastRunLength);
return lscp - lastSpanLength;
}
/// <summary>
/// Find the hyphen break following or preceding the specified current LSCP
/// </summary>
/// <remarks>
/// This method never checks whether the specified current LSCP is already right
/// at a hyphen break. It either finds the next or the previous break regardless.
///
/// A negative lscchLim param value indicates the caller finds the hyphen immediately
/// before the specified character index.
/// </remarks>
/// <param name="lscpCurrent">the current LSCP</param>
/// <param name="lscchLim">the number of LSCP to search for break</param>
/// <param name="isCurrentAtWordStart">flag indicates whether lscpCurrent is the beginning of the word to hyphenate</param>
/// <param name="lscpHyphen">LSCP of the hyphen</param>
/// <param name="lshyph">Hyphen properties</param>
internal bool FindNextHyphenBreak(
int lscpCurrent,
int lscchLim,
bool isCurrentAtWordStart,
ref int lscpHyphen,
ref LsHyph lshyph
)
{
lshyph = new LsHyph(); // no additional hyphen properties for now
if (_store.Pap.Hyphenator != null)
{
int lscpChunk;
int lscchChunk;
LexicalChunk chunk = GetChunk(
_store.Pap.Hyphenator,
lscpCurrent,
lscchLim,
isCurrentAtWordStart,
out lscpChunk,
out lscchChunk
);
_lscpHyphenationLookAhead = lscpChunk + lscchChunk;
if (!chunk.IsNoBreak)
{
int ichCurrent = chunk.LSCPToCharacterIndex(lscpCurrent - lscpChunk);
int ichLim = chunk.LSCPToCharacterIndex(lscpCurrent + lscchLim - lscpChunk);
if (lscchLim >= 0)
{
int ichNext = chunk.Breaks.GetNextBreak(ichCurrent);
if (ichNext >= 0 && ichNext > ichCurrent && ichNext <= ichLim)
{
// -1 because ichNext is the character index where break occurs in front of it,
// while LSCP is the position where break occurs after it.
lscpHyphen = chunk.CharacterIndexToLSCP(ichNext - 1) + lscpChunk;
return true;
}
}
else
{
int ichPrev = chunk.Breaks.GetPreviousBreak(ichCurrent);
if (ichPrev >= 0 && ichPrev <= ichCurrent && ichPrev > ichLim)
{
// -1 because ichPrev is the character index where break occurs in front of it,
// while LSCP is the position where break occurs after it.
lscpHyphen = chunk.CharacterIndexToLSCP(ichPrev - 1) + lscpChunk;
return true;
}
}
}
}
return false;
}
/// <summary>
/// Get the lexical chunk the specified current LSCP is within.
/// </summary>
private LexicalChunk GetChunk(
TextLexicalService lexicalService,
int lscpCurrent,
int lscchLim,
bool isCurrentAtWordStart,
out int lscpChunk,
out int lscchChunk
)
{
int lscpStart = lscpCurrent;
int lscpLim = lscpStart + lscchLim;
int cpFirst = _store.CpFirst;
if (lscpStart > lscpLim)
{
// Start is always before limit
lscpStart = lscpLim;
lscpLim = lscpCurrent;
}
LexicalChunk chunk = new LexicalChunk();
int cchWordMax;
CultureInfo textCulture;
SpanVector<int> textVector;
char[] rawText = _store.CollectRawWord(
lscpStart,
isCurrentAtWordStart,
_isSideways,
out lscpChunk,
out lscchChunk,
out textCulture,
out cchWordMax,
out textVector
);
if ( rawText != null
&& cchWordMax >= MinCchWordToHyphenate
&& lscpLim < lscpChunk + lscchChunk
&& textCulture != null
&& lexicalService != null
&& lexicalService.IsCultureSupported(textCulture)
)
{
// analyze the chunk and produce the lexical chunk to cache
TextLexicalBreaks breaks = lexicalService.AnalyzeText(
rawText,
rawText.Length,
textCulture
);
if (breaks != null)
{
chunk = new LexicalChunk(breaks, textVector);
}
}
return chunk;
}
/// <summary>
/// Get a text store containing the specified plsrun
/// </summary>
internal TextStore StoreFrom(Plsrun plsrun)
{
return TextStore.IsMarker(plsrun) ? _markerStore : _store;
}
/// <summary>
/// Get a text store containing the specified lscp
/// </summary>
internal TextStore StoreFrom(int lscp)
{
return lscp < 0 ? _markerStore : _store;
}
/// <summary>
/// Flag indicating whether vertical adjustment of some runs is required
/// </summary>
internal bool VerticalAdjust
{
get { return (_statusFlags & StatusFlags.VerticalAdjust) != 0; }
set
{
if (value)
_statusFlags |= StatusFlags.VerticalAdjust;
else
_statusFlags &= ~StatusFlags.VerticalAdjust;
}
}
/// <summary>
/// Flag indicating whether force wrap is required
/// </summary>
internal bool ForceWrap
{
get { return (_statusFlags & StatusFlags.ForceWrap) != 0; }
set
{
if (value)
_statusFlags |= StatusFlags.ForceWrap;
else
_statusFlags &= ~StatusFlags.ForceWrap;
}
}
/// <summary>
/// Flag indicating whether state should be kept in the line
/// </summary>
internal bool KeepState
{
get { return (_statusFlags & StatusFlags.KeepState) != 0; }
}
/// <summary>
/// Formatting store for main text
/// </summary>
internal TextStore TextStore
{
get { return _store; }
}
/// <summary>
/// Formatting store for marker text
/// </summary>
internal TextStore TextMarkerStore
{
get { return _markerStore; }
}
/// <summary>
/// Current formatter
/// </summary>
internal TextFormatterImp Formatter
{
get { return _store.Settings.Formatter; }
}
/// <summary>
/// formattng ideal width
/// </summary>
internal int FormatWidth
{
get { return _store.FormatWidth; }
}
}
}
|