www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.announce - Indigo template containers 0.6

reply "Uwe Salomon" <post uwesalomon.de> writes:
Indigo is a project to provide general containers for D that look & feel  
like the Qt 4.0 container classes, and additionally harmonize very well  
with D specialities. This is version 0.6, containing Vector, List,  
LinkedList, Map, and String. This release completes the String class,  
though i could not implement every feature of QString (especially RegExps  
and complex localization functionality). As this is all i wanted to  
accomplish, this project is somehow finished now (though i will implement  
bugfixes and requested features, of course).

String contains a some borrowed code from Mango again: the UTF16 to UTF8  
decoder, mostly copied and a bit improved (about 30% faster). Thanks a lot  
for that, too.

You reach the docs under:
   http://www.uwesalomon.de/code/indigo/index.html
And the source under:
   http://www.uwesalomon.de/code/indigo/indigo.tar.gz

In the docs there are also some results of tests i made that show the very  
high performance of the containers.

This is not my first programming project, but my first open source  
project, thus i will be very grateful for advices from you all.

Thanks & ciao
uwe
May 15 2005
next sibling parent reply Victor Nakoryakov <nail-mail mail.ru> writes:
Uwe Salomon wrote:

 Indigo is a project to provide general containers for D that look & 
 feel  like the Qt 4.0 container classes, and additionally harmonize very 
 well  with D specialities. This is version 0.6, containing Vector, 
 List,  LinkedList, Map, and String. This release completes the String 
 class,  though i could not implement every feature of QString 
 (especially RegExps  and complex localization functionality). As this is 
 all i wanted to  accomplish, this project is somehow finished now 
 (though i will implement  bugfixes and requested features, of course).
 
 String contains a some borrowed code from Mango again: the UTF16 to 
 UTF8  decoder, mostly copied and a bit improved (about 30% faster). 
 Thanks a lot  for that, too.
 
 You reach the docs under:
   http://www.uwesalomon.de/code/indigo/index.html
 And the source under:
   http://www.uwesalomon.de/code/indigo/indigo.tar.gz
 
 In the docs there are also some results of tests i made that show the 
 very  high performance of the containers.
 
 This is not my first programming project, but my first open source  
 project, thus i will be very grateful for advices from you all.
 
 Thanks & ciao
 uwe


Great library. Too few algorithms yet :), but already implemented 
classes are very useful anyway. I'ld be very happy if Indigo will become 
part of standart D library in someway. Keep good work ;)
May 16 2005
parent reply "Uwe Salomon" <post uwesalomon.de> writes:
 Great library.


Thanks :)


 Too few algorithms yet :)


Yes, i totally forgot about them. :)  What algorithms would you want to  
see implemented at first? Binary search comes to my mind. Sorting is  
already implemented in the containers themselves, is it needed as an  
algorithm anyways?

Ciao
uwe
May 16 2005
parent reply Victor Nakoryakov <nail-mail mail.ru> writes:
Uwe Salomon wrote:

 Yes, i totally forgot about them. :)  What algorithms would you want to  
 see implemented at first? Binary search comes to my mind. Sorting is  
 already implemented in the containers themselves, is it needed as an  
 algorithm anyways?


I find most useful algorithms those that perform "boolean" operations. 
E.g. return elements that are in both containers, that in first but not 
in second, that just only in one not in both, etc, etc. This was my IMHO.

Also I find container "Set" very useful in many cases, so it would be 
very nice to see it in further library versions.
May 16 2005
parent "Uwe Salomon" <post uwesalomon.de> writes:
 I find most useful algorithms those that perform "boolean" operations.  
 E.g. return elements that are in both containers, that in first but not  
 in second, that just only in one not in both, etc, etc. This was my IMHO.

 Also I find container "Set" very useful in many cases, so it would be  
 very nice to see it in further library versions.


Ok, i will work on that for 0.8. Note that boolean algorithms are very  
slow in containers like List or Vector, thus i will implement them for  
Set, perhaps for Hash/Map.

Thanks for the feedback
uwe
May 21 2005
prev sibling next sibling parent "G.Vidal" <gyvidal wanadoo.fr> writes:
By the way, Natural Doc is really cool, thanks for the idea.
May 17 2005
prev sibling parent reply "Kris" <fu bar.com> writes:
"Uwe Salomon" <post uwesalomon.de> wrote in message
news:op.sqtug0y46yjbe6 sandmann.maerchenwald.net...

 String contains a some borrowed code from Mango again: the UTF16 to UTF8
 decoder, mostly copied and a bit improved (about 30% faster). Thanks a lot
 for that, too.


You're welcome!

But you did fail to note that those functions of yours insist on allocating
memory, making them 300% slower in a brief test against the current
mango.io.Utf   ;-)

These routines are quite efficient though, especially where one can avoid
heap-allocation (e.g. conversion of fragments).
May 17 2005
parent "Uwe Salomon" <post uwesalomon.de> writes:
 String contains a some borrowed code from Mango again: the UTF16 to UTF8
 decoder, mostly copied and a bit improved (about 30% faster). Thanks a  
 lot for that, too.


 You're welcome!

 But you did fail to note that those functions of yours insist on  
 allocating
 memory, making them 300% slower in a brief test against the current
 mango.io.Utf   ;-)


No, they don't insist. Use setUtf8() repeatedly on the same string.  
Providing the buffer would make no sense for this algorithm (its purpose  
is to fill the string's internal buffer). But it is easy to write it  
stand-alone, with an optional second buffer argument.


 These routines are quite efficient though, especially where one can avoid
 heap-allocation (e.g. conversion of fragments).


Perhaps you want to look at the conversion strategy: It is a bit different  
 from yours, saving a variable and some code size. That makes up the 30%  
(Tested on a random german email i took from my girlfriend). Should work  
with your algorithm, too...

Ciao
uwe
May 17 2005