/*
* Provides a fast-path for Encoder.canEncode, which drastically improves performance on Android post JellyBean.
* After KitKat, the implementation of canEncode degrades to the point of being useless. For non ASCII or UTF,
* performance may be bad. We can add more encoders for common character sets that are impacted by performance
* issues on Android if required.
*
* Benchmarks: *
* OLD toHtml() impl v New (fastpath) in millis
* Wiki: 1895, 16
* CNN: 6378, 55
* Alterslash: 3013, 28
* Jsoup: 167, 2
*/
private static bool CanEncode(Entities.CoreCharset charset, char c, System.Text.Encoding fallback)
{
// todo add more charset tests if impacted by Android's bad perf in canEncode
switch (charset)
{
case Entities.CoreCharset.ascii: {
return(c < 0x80);
}
case Entities.CoreCharset.utf: {
// real is:!(Character.isLowSurrogate(c) || Character.isHighSurrogate(c)); - but already check above
return(true);
}
default: {
return(fallback.CanEncode(c));
}
}
}
// this method is ugly, and does a lot. but other breakups cause rescanning and stringbuilder generations
internal static void Escape(StringBuilder accum, String str, OutputSettings outputSettings, bool inAttribute
, bool normaliseWhite, bool stripLeadingWhite)
{
bool lastWasWhite = false;
bool reachedNonWhite = false;
Entities.EscapeMode escapeMode = outputSettings.EscapeMode();
System.Text.Encoding encoder = outputSettings.Charset();
Entities.CoreCharset coreCharset = GetCoreCharsetByName(outputSettings.Charset().Name());
IDictionary <char, String> map = escapeMode.GetMap();
int length = str.Length;
int codePoint;
for (int offset = 0; offset < length; offset += iText.IO.Util.TextUtil.CharCount(codePoint))
{
codePoint = str.CodePointAt(offset);
if (normaliseWhite)
{
if (iText.StyledXmlParser.Jsoup.Helper.StringUtil.IsWhitespace(codePoint))
{
if ((stripLeadingWhite && !reachedNonWhite) || lastWasWhite)
{
continue;
}
accum.Append(' ');
lastWasWhite = true;
continue;
}
else
{
lastWasWhite = false;
reachedNonWhite = true;
}
}
// surrogate pairs, split implementation for efficiency on single char common case (saves creating strings, char[]):
if (codePoint < iText.IO.Util.TextUtil.CHARACTER_MIN_SUPPLEMENTARY_CODE_POINT)
{
char c = (char)codePoint;
// html specific and required escapes:
switch (c)
{
case '&': {
accum.Append("&");
break;
}
case (char)0xA0: {
if (escapeMode != Entities.EscapeMode.xhtml)
{
accum.Append(" ");
}
else
{
accum.Append(" ");
}
break;
}
case '<': {
// escape when in character data or when in a xml attribue val; not needed in html attr val
if (!inAttribute || escapeMode == Entities.EscapeMode.xhtml)
{
accum.Append("<");
}
else
{
accum.Append(c);
}
break;
}
case '>': {
if (!inAttribute)
{
accum.Append(">");
}
else
{
accum.Append(c);
}
break;
}
case '"': {
if (inAttribute)
{
accum.Append(""");
}
else
{
accum.Append(c);
}
break;
}
default: {
if (CanEncode(coreCharset, c, encoder))
{
accum.Append(c);
}
else
{
if (map.ContainsKey(c))
{
accum.Append('&').Append(map.Get(c)).Append(';');
}
else
{
accum.Append("&#x").Append(JavaUtil.IntegerToHexString(codePoint)).Append(';');
}
}
break;
}
}
}
else
{
String c = new String(iText.IO.Util.TextUtil.ToChars(codePoint));
if (encoder.CanEncode(c))
{
// uses fallback encoder for simplicity
accum.Append(c);
}
else
{
accum.Append("&#x").Append(JavaUtil.IntegerToHexString(codePoint)).Append(';');
}
}
}
}
