[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kde-core-devel
Subject:    Re: RFC KURL rewite (1st STAGE)
From:       Dawit Alemayehu <adawit () earthlink ! net>
Date:       1999-10-09 16:42:54
[Download RAW message or body]

On Sat, 9 Oct 1999, weis wrote :
> Hi,

> while I consider it very useful to fix the shortcomings of
> the current KURL, I would strongly suggest NOT to use
> QRegExp! It is not very efficient!
>
> The current KURL and QUrl use state machines since they
> are much more efficient than this QRegExp.
>
> In KDE and especially konqueror URLs are parsed in masses, so
> parsing URLs is a time critical thing.

Thanks for the feedback.  My next step would have been to do some
benchmark since I basically had the same concerns.  I did see the use of the
state machines in KURL.

Perhaps I should have done that first before posting. A question though. 
Should all use of QRegExp be avoided ?  I ask this because while converting the
basic checks to a simple while loop is very trivial, the validation is not. As
an example, look at the test for on the hostname which can only consist of
alphanumeric & "-" values and they have to conform to a certain pattern. 

While we are on the performance topic, I saw a prevallant use of Qchar in the
original code.  Was this done for performance purposes as well ? Is invoking 

QString mid ( uint index, uint len=0xffffffff ) const 

more expensive than creating new Qstring object out of

QString ( const QChar * unicode, uint length )  ?


[OFF TOPIC :]  Would there be a possibility of having a Quri/Qurl class future
versions of QT ?

Regards,
Dawit A.

[prev in list] [next in list] [prev in thread] [next in thread]