[yunqa.de] Re: DIRegEx - translating utf8 match info to utf16
- From: Delphi Inspiration <delphi@xxxxxxxx>
- To: yunqa@xxxxxxxxxxxxx
- Date: Fri, 16 Nov 2007 17:31:51 +0100
Hello Jim Bretti,
>I noticed that I can write my source widestrings to memory streams, and use
>TDiRegExSearchStream_Utf16LE instead of TDIPerlRegEx. The SearchNext method
>for the stream regex seems to return what I need ...
Yes, this does certainly work.
>is this the best way to handle?
Please notice that TDIRegExSearchStream_Utf16LE uses the DFA matching algorithm
instead of the Perl one. While mostly similar, DFA differs from Perl in some
important aspects: the greedy or ungreedy nature of repetition quantifiers is
not relevant, no substring capturing, no support for backreference, plus a few
others. To see the full list of differences, search the DIRegEx help for
"pcrematching".
I also believe that there is a performance trade-off between
TDIRegExSearchStream_Utf16LE and BufCountUtf8Chars approach I suggested in my
other answer. The DFA algorithm is somewhat slower than Perl, and
TDIRegExSearchStream_Utf16LE has some overhead for stream handling and on the
fly character conversion. On the other hand, only needs to perform a single
scan through the subject string.
I therefore expect that TDIRegExSearchStream_Utf16LE will be faster if
* the subject string is rather long.
* positions for multiple matches need to be determined.
For single matches in short strings I expect the BufCountUtf8Chars to perform
faster.
Ralf
PS: My appologies for not answering sooner, I have been sick for a few days.
_______________________________________________
Delphi Inspiration mailing list
yunqa@xxxxxxxxxxxxx
//www.freelists.org/list/yunqa
Other related posts:
