devarenn@aar.alcatel-alsthom.fr (Agnes Devarenne) (04/22/91)
We are starting an important development in C++ and we are looking for basic libraries : sets, bags, collections, Btrees, stacks, lists... We have presently four candidates: - NIHCL - Tools.h from RogueWave - Classix.h from Classix - the gnu one Has anyone ever used these libraries? What do you know about them (efficiency, reusability, flexibility, bugs...)? Thanks in advance for any comment. *===============================================================* * * * Agnes Devarenne * * * * Alcatel-Alsthom Recherche * Route de Nozay, * * 91460 MARCOUSSIS FRANCE * * * * Tel. : (33) (1) 64.49.17.71, * * mail : devarenn@aar.alcatel-alsthom.fr *===============================================================* * Sometimes a cigar is just a cigar * * (S. Freud) * *===============================================================* -- *===============================================================* * * * Agnes Devarenne * * *
laborell@mimosa.unice.fr (Louis Laborelli) (04/24/91)
Bonjour, In article <1991Apr22.122217.486@aar.alcatel-alsthom.fr>, devarenn@aar.alcatel-alsthom.fr (Agnes Devarenne) writes: |> We are starting an important development in C++ and we are looking |> for basic libraries : sets, bags, collections, Btrees, stacks, lists... |> |> We have presently four candidates: |> - NIHCL |> - Tools.h from RogueWave |> - Classix.h from Classix |> - the gnu one |> |> Has anyone ever used these libraries? What do you know about them |> (efficiency, reusability, flexibility, bugs...)? |> |> Thanks in advance for any comment. Je ne connais pas Tools et Classix .Ce sont des produits commerciaux ? ( je crois me souvenir d'une librairie de methodes numeriques simples nomme rogue . ( Matrices , fft...)) J'ai surtout utilise depuis deux ans libg++ , pour les structures de donnees . Ca marche a peu pres bien , (les bugs sont en faible nombre et repertoriees dans gnu.libg++.bugs) C'est toutefois assez dur a utiliser . (il faut generer de nombreuses methodes , modifier a la main certains fichiers ...). Je peux me tromper mais mais je n'ai pas reussi a avoir plusieurs instanciations generiques d'un ensemble avec le meme code . De ce fait , ca devient vite gigantesque . J'attend avec impatience les templates.... efficacite : bonne , sauf pour certaines operations , et ce n'est que peu documente . Le probleme de la documentation est important , car on ne sait pas trop si on a droit de tout faire ( par exemple entrelacer des operations de parcours et d'ajout ) NIH n'etait pas documente il ya deux ans..donc... Il faut aussi considerer ET++ qui possede plusieurs structures de donnees. Je travaille avec libg++ et InterViews. -- Louis Laborelli Universite de Nice Sophia Antipolis / phone: (33) 92 94 26 89 I3S LISAN - CNRS Batiment 4 \ telex: GRP 970006F 250 rue Albert Einstein / fax: (33) 92 94 28 98 Sophia Antipolis \ e-mail:laborell@zig.inria.fr 06560 Valbonne FRANCE / or laborell@mimosa.unice.fr
rmartin@clear.com (Bob Martin) (04/25/91)
In article <1991Apr22.122217.486@aar.alcatel-alsthom.fr> devarenn@aar.alcatel-alsthom.fr () writes: >We are starting an important development in C++ and we are looking >for basic libraries : sets, bags, collections, Btrees, stacks, lists... > >We have presently four candidates: >- NIHCL >- Tools.h from RogueWave >- Classix.h from Classix >- the gnu one > >Has anyone ever used these libraries? What do you know about them >(efficiency, reusability, flexibility, bugs...)? I have been a user of Classix for a few months now. The library consists of simple basics, and so far everything I have tried works. Furthermore the library is concise and doesn't add a megabyte to my executables like NIHCL did. The classes are easy to use and reasonably complete.. I have had the following problems: The string class could be better. NIHCL's is much better. Classix string class has no converter from String to char*. You have to derive your own class if you want it. Also substrings are not supported. Although they claim to have a "parameterizing" utility which will convert their generic container classes into specific container classes, they have delayed shipment of this utility. It has been several months since I placed my order and I still have not received it. The DATUM concept is a bit week. First of all the type name DATUM conflicts with another 3rd party package the I am using, so I was forced to change DATUM to CDatum. Secondly I have declared CDatum to be void*. THis allows me general access to the container classes but since void* has no meaningful comparison operators the "sorted" classes which depend on object comparisons aren't worth much. Finally they used the type `Node' as a subtype in the tree classes. This conflicts with some of my own code which also declared a type `Node', so I had to change their source to use a less common name. I am using their library, and I think I will be a bit happier when I can parameterize the container classes properly. Right now its slim-pickin's out there and we have to use what's available. THE INDUSTRY NEEDS TO DEVELOP SOME LIBRARY STANDARDS like an IC data book. -- Path: clrcom!balr!tellab5!laidbak!ism.isc.com!ispd-newsserver!rpi!zaphod.mps.ohio-state.edu!mips!apple!tahoe!jimi!arrakis!niobium From: niobium@nevada.edu (Christopher W. Carlson) Newsgroups: comp.sys.amiga.datacomm Subject: Re: Fixing NiftyTerm to use SHARED mode on serial.device
balden@wimsey.bc.ca (Bruce Balden) (04/28/91)
Here follows an interlinear translation of Louis Laborelli's useful comments on libg++ and other related stuff, which, however, American and Canadian readers may not fully understand in the French language. In article <77@taloa.unice.fr> laborell@mimosa.unice.fr (Louis Laborelli) writes: >Bonjour, In article <1991Apr22.122217.486@aar.alcatel-alsthom.fr>, devarenn@aar.alcatel-alsthom.fr (Agnes Devarenne) writes: >|> We are starting an important development in C++ and we are looking >|> for basic libraries : sets, bags, collections, Btrees, stacks, lists... >|> We have presently four candidates: >|> - NIHCL >|> - Tools.h from RogueWave >|> - Classix.h from Classix >|> - the gnu one >|> Has anyone ever used these libraries? What do you know about them >|> (efficiency, reusability, flexibility, bugs...)? >Je ne connais pas Tools et Classix .Ce sont des produits commerciaux ? I'm not familiar with Tools or Classix (are these commercial products?) >( je crois me souvenir d'une librairie de methodes numeriques simples >nomme rogue . ( Matrices , fft...)) I seem to remember a library of simple numerical methods called rogue (matrices, fft, etc.) >J'ai surtout utilise depuis deux ans libg++ , pour les structures de donnees . The library I've used most over the past two years is libg++, for data structures. >Ca marche a peu pres bien , (les bugs sont en faible nombre et repertoriees dans gnu.libg++.bugs) It works more or less well; its bugs are few in number and reported in gnu.libg++.bugs. >C'est toutefois assez dur a utiliser . (il faut generer de nombreuses methodes , modifier a la main certains fichiers ...). It is nevertheless rather difficult to use (you have to generate many methods and modify certain files by hand) >Je peux me tromper mais mais je n'ai pas reussi a avoir plusieurs instanciations generiques d'un ensemble avec le meme code . De ce fait , ca devient vite gigantesque . I could be making a mistake, but I've not succeeded in making multiple instantiations of the same generic type share code. Because of this fact, the code quickly mushrooms in size. >J'attend avec impatience les templates.... I impatiently await templates ... >efficacite : bonne , sauf pour certaines operations , et ce n'est que peu documente . Efficiency: good, except for certain operations, and its only a little documented. >Le probleme de la documentation est important , car on ne sait pas trop si on a droit de tout faire ( par exemple entrelacer des operations de parcours et d'ajout ) The lack of documentation is serious, because one never sure how methods are going to interact (for example, interleaving the operations of traversal and adding [to a tree]). >NIH n'etait pas documente il ya deux ans..donc... NIH was not decoumented two years ago, hence... >Il faut aussi considerer ET++ qui possede plusieurs structures de donnees. You should consider ET++ which has many data structure [classes]. >Je travaille avec libg++ et InterViews. I have worked with libg++ and Interviews. > Louis Laborelli > Universite de Nice Sophia Antipolis / phone: (33) 92 94 26 89 > I3S LISAN - CNRS Batiment 4 \ telex: GRP 970006F > 250 rue Albert Einstein / fax: (33) 92 94 28 98 > Sophia Antipolis \ e-mail:laborell@zig.inria.fr > 06560 Valbonne FRANCE / or laborell@mimosa.unice.fr -- ******************************************************************************* * Bruce E. Balden Computer Signal Corporation Canada * * Thaumaturgist 225B Evergreen Drive * * balden@xenophon.wimsey.bc.ca Port Moody, B.C. V3H 1S1 CANADA *