[Springframework-developer] HtmlUtils refactoring

[Martin K. <Mar...@St...>]

Hi there,
 
   Since I critised the Spring code base within a forum post made
in the architectural section (topic: Spring Code Remarks),
I felt like refactoring the prime evil class HtmlUtils. Please 
don't take the prime evil thing as an offence. But the 
HtmlUtil implementation features some of the biggest abuses 
in terms of if statements and code duplication, I have ever seen.
So this implementation is surely blessed by the dark side! ;-)

I have spend around three hours to rework the current 
implementation. I could eleminate 200 SLOC lines of code
and refactored the character entity references setup to
be more readable (boy that was a hot editing session).

Also the unescape and escape methods got some
attantion. (found a way to improve the 
performance as well).

The current trouble maker is the 
isSpecialCharacter method. I would like to scrap this 
method and let the CharacterEntityReference class take 
care of it. You know this class has the appropriate 
knowledge and it can be easily computed at startup 
using an int-array.

There are two opptions:

First use 3000 bits(! not bytes!) and go for a 
mapping of the chars 0....1000 and 
8000...10000. This would provide O(1) complexity.
Another way would be the use of a sorted char array 
with O(log n) but around 1000 bytes.

I would like to implement the first one. It is nearly
a no brainer but it might be a bit difficult for the 
reader. But I would like to get away from 
104 McCabe decision points, anyway.

Another point is the unescape method. I am
not sure if EMPTY_REFERENCE and
MALFORMED_REFERENCE are really
needed. Is there anything in the spec about
'&;' being allowed?

I would like to drop that too.

Another thing is that I think I found a 
semantical bug in the orginal implementation. 
Check out this test-case:

&  
In the current implementation the first & and 
the ';' is threatened as a reference. Since
this reference can not be resolved, it will be
droped. I would think it should be 
evaluated to & ' '.

I couldn't test the refactoring (it is still 
incomplete anyways) since I did it
within a test runtime environment. I use
the Spring codebase of 1.4 as the default
use-case scenario for my final exams work 
(which is about to be completed in two weeks I hope).


Cheers,

Martin (Kersten)