Performance: improve end-of-line handling
-----------------------------------------
Key: XERCESC-1369
URL: http://issues.apache.org/jira/browse/XERCESC-1369
Project: Xerces-C++
Type: Improvement
Components: Miscellaneous
Versions: 2.6.0
Reporter: Christian Will
Priority: Minor
Attachments: XMLReader.cpp.patch, XMLReader.hpp.patch
We can improve the end-of-line handling by two steps.
1. We move the function XMLReader:handleEOL(...) from the header into the cpp
file, because the function is to big for inlining.
2. We create bit masks to avoid most of the handleEOL calls.
Here are two examples :
a)
We use the content information that our current character is a whitespace. The
bit mask selects all cases where we have to call handleEOL.
if (isWhitespace(curCh))
{
//
// 'curCh' is a whitespace(x20|x9|xD|xA), so we only can have
// end-of-line combinations with a leading chCR(xD) or chLF(xA)
//
// 100000 x20
// 001001 x9
// 001010 chLF
// 001101 chCR
// -----------
// 000110 == (chCR|chLF) & ~(0x9|0x20)
//
// if the result of thelogical-& operation is
// true : 'curCh' must be xA or xD
// false : 'curCh' must be x20 or x9
//
if ( ( curCh & (chCR|chLF) & ~(0x9|0x20) ) == 0 )
{
fCurCol++;
} else
{
handleEOL(curCh, false);
}
b)
We have no content information so we have to test for all four possible start
characters.
The bit masks selects only 128 cases (from before 63483) where we have to call
handleEOL.
//
// we can have end-of-line combinations with a leading
// chCR(xD), chLF(xA), chNEL(x85), or chLineSeparator(x2028)
//
// 0000000000001101 chCR
// 0000000000001010 chLF
// 0000000010000101 chNEL
// 0010000000101000 chLineSeparator
// -----------------------
// 1101111101010000 == ~(chCR|chLF|chNEL|chLineSeparator)
//
// if the result of the logical-& operation is
// true : 'curCh' can not be chCR, chLF, chNEL or chLineSeparator
// false : 'curCh' can be chCR, chLF, chNEL or chLineSeparator
//
if ( chGotten & (XMLCh) ~(chCR|chLF|chNEL|chLineSeparator) )
{
fCurCol++;
} else
{
handleEOL(chGotten, false);
}
I created bit masks for all (5) cases where we call handleEOL.
I attached patch files against the latest cvs version.
Regards,
Christian Will
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://issues.apache.org/jira/secure/Administrators.jspa
-
If you want more information on JIRA, or have a bug to report see:
http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
