git-svn-id: https://svn.apache.org/repos/asf/xerces/c/trunk@554580 13f79535-47bb-0310-9956-ffa450edef68

git-svn-id: https://svn.apache.org/repos/asf/xerces/c/trunk@471747 13f79535-47bb-0310-9956-ffa450edef68

git-svn-id: https://svn.apache.org/repos/asf/xerces/c/trunk@191036 13f79535-47bb-0310-9956-ffa450edef68

git-svn-id: https://svn.apache.org/repos/asf/xerces/c/trunk@176317 13f79535-47bb-0310-9956-ffa450edef68

git-svn-id: https://svn.apache.org/repos/asf/xerces/c/trunk@176279 13f79535-47bb-0310-9956-ffa450edef68

git-svn-id: https://svn.apache.org/repos/asf/xerces/c/trunk@176169 13f79535-47bb-0310-9956-ffa450edef68

git-svn-id: https://svn.apache.org/repos/asf/xerces/c/trunk@176026 13f79535-47bb-0310-9956-ffa450edef68

git-svn-id: https://svn.apache.org/repos/asf/xerces/c/trunk@175228 13f79535-47bb-0310-9956-ffa450edef68

git-svn-id: https://svn.apache.org/repos/asf/xerces/c/trunk@174989 13f79535-47bb-0310-9956-ffa450edef68

git-svn-id: https://svn.apache.org/repos/asf/xerces/c/trunk@174984 13f79535-47bb-0310-9956-ffa450edef68

git-svn-id: https://svn.apache.org/repos/asf/xerces/c/trunk@174311 13f79535-47bb-0310-9956-ffa450edef68

git-svn-id: https://svn.apache.org/repos/asf/xerces/c/trunk@173439 13f79535-47bb-0310-9956-ffa450edef68

git-svn-id: https://svn.apache.org/repos/asf/xerces/c/trunk@172633 13f79535-47bb-0310-9956-ffa450edef68

Details: (see xerces-c-dev mailing Jan 14)
XMLRecognizer.cpp: the internal encoding string XMLUni::fgXMLChEncodingString
was going through this function numerous times.  As a result, the top hot-spot
for the parse was _wcsicmp().  The real problem is that the Microsofts wide string
functions are unbelievably slow.  For things like encodings, it might be
better to use a special comparison function that only considers a-z and
A-Z as characters with case.  This works since the character set for
encodings is limit to printable ASCII characters.

 XMLScanner2.cpp: This also has some case-sensitive vs. insensitive compares.
They are also much faster.  The other tweak is to only make a copy of an attribute
string if it needs to be split.  And then, the strategy is to try to use a
stack-based buffer, rather than a dynamically-allocated one.

SAX2XMLReaderImpl.cpp: Again, more case-sensitive vs. insensitive comparisons.

KVStringPair.cpp & hpp: By storing the size of the allocation, the storage can
likely be re-used many times, cutting down on dynamic memory allocations.

XMLString.hpp: a more efficient implementation of stringLen().

DTDValidator.cpp: another case of using a stack-based buffer when possible

These patches made a big difference in parse time in some of our test
files, especially the ones are very attribute-heavy.


git-svn-id: https://svn.apache.org/repos/asf/xerces/c/trunk@172497 13f79535-47bb-0310-9956-ffa450edef68

1.1.0 code to be finished. I can't list them all here, but a list is
available elsewhere.


git-svn-id: https://svn.apache.org/repos/asf/xerces/c/trunk@171927 13f79535-47bb-0310-9956-ffa450edef68

git-svn-id: https://svn.apache.org/repos/asf/xerces/c/trunk@171712 13f79535-47bb-0310-9956-ffa450edef68

git-svn-id: https://svn.apache.org/repos/asf/xerces/c/trunk@171387 13f79535-47bb-0310-9956-ffa450edef68