Where is Code_Aster (website) going?

corra

Hello there,

I have been using Code_Aster since 2009. I love the software, and I especially love the documentation coming with it, which allows everyone to understand the theory behind it. For some time, I have used the Windows version; then I went back to Linux systems and, I must say, all the hassle coming from the installation.

I have noticed that the downloading page has not been updated since 2020 (version 15.2), while, scrolling the forum, I got in touch with this gitlab repository:

https ://gitlab.com/codeaster-opensource-documentation

where it seems that many changes have occurred since 2020 (Code Aster seems mainly moved to containers; more close relationship with Salome Meca; Python API and new architecture), which are, at once, a bit confusing. It seems strange that there is no mention of all this in the official website. So my basic questions are: is the website still maintained? Has Code Aster been forked by a new team? Where should a generic user look for news, documentation and official versions?

Thank you very much for any kind answer.

Corrado

paul18

Hi,

Quite recently I came back to Code Aster after more than 15 years; as expected and logically, some changes have occurred in the meantime, even if the Code Aster skeleton remains more or less the same through last decades.

The main difference I’ve been noticing is the community behaviour and more precisely the lack of reaction: years ago, any question in the forum found an answer quite fastly (in the day from memory), both internal and external Code Aster people felt involved … that’s not clearly the case today: some posts has few answers or not at all.

My feeling: Code Aster developments started in the 1990’s i.e. more than 30 years ago => another time, another developers (another strategies? for interna use only?).

It seems that learning Code Aster has become more hard today; I can imagine any new user will wonder if it's worth getting started with a new solver with limited support?… it’s a pity because as Corrado said, it’s probably one of the most complete FEM solver in terms of functionalities, docs, test cases and so on.

Paul

AsterO'dactyle

Hello,

ode_aster is even more developed and active than ever.

On the other hand, we don't have any time to devote to animating the free community, including the website. This was already the case 15 years ago, and the time was taken from people's personal time (evenings and weekends).

Now, it's mainly devoted to EDF industrial issues.

I'd also like to make two comments:

- 10 or 15 years ago, one or two questions a week was manageable. Now it's several dozen, and the questions are particularly difficult.

- code distribution and tools keep pace with modern developments: containers, use of Git. There will also be major changes in documentation.

Finally, I notice that the community can't organize itself. It's open-source code, and apart from a few notable exceptions (like our Italian friends), it's mostly a collection of isolated people who come in once to ask a couple of questions and then disappear. Nevertheless, as many people answer the questions with accuracy and enthusiasm, I think it's not very nice not to notice.

After many years of hard work by our historical fathers, users are getting organized on the industrial side, with co-development agreements, particularly in the nuclear sector, or via highly professional solutions such as SimScale. I invite you to follow the news we regularly post on LinkedIn (often at weekends too...).

corra

Dear AsterO'dactyle,

Thank you for your answer. My question was not meant to criticize anything, it was just a genuine question on the "modern developments", as you said. Where are they posted and released? Is the website to be considered outdated? Where should a generic user look for news, documentation and official versions?

About the forum I completely agree with you, it is made by enthusiasts and experts in their free time. So the next question is: how should a user (an academic one, in my case) look for more structured support? Is it possible in Code_Aster, for instance in form of formal agreement or paid support?

I am sorry for the long list of questions, but I would really like to understand.

Kind regards,

Corrado

laurent

I learned Aster when you first had to compile it with Intel compiler and MKL around 2003/2004.

The whole learning was really hard but i had the time.

Very useful learning experience and an opportunity to compare with others softwares (Abaqus)

I lost touch around 2016.

Coming back to it recently to assess a thermal problem (1st Aster, then i will see if i need to move to Syrthes):

a. Coudn't do a Ubuntu install with the Singularity stuff and didn't want to lose too much time if others solutions would work (hence: Windows & Salome-for-Windows 2021). Salome-Meca direct install also wouldn't work on Ubuntu.

b. Had issue to just download Salome stand-alone: website link to working to get the file after filling up all infos...

So, yes, starting/restarting is difficult but maybe easily fixable?

Yet, if/when Salome-Meca is working (Windows today sadly for me, but hopefull it will change), i got my simulation set-up and working in much less time that i would have thought it would take.

The integration is really an improvement.

And i can see that User Meetings started again, with sharing of the presentations...

Hopefully it will follow a good trajectory/path and this discussion might help getting attention on some of the issues

ps: now reading AsterO'dactyle's comments: i am one of those individuals, using the strength of an industrial code @no cost, which has served me well. I try to help whenever i can, but as mentioned, some questions reflect how complex simulation can be.

Personnaly, Aster & Saturne helped me pushing topics i could never push without demonstrating by myself (with those tools) a few points conceptually. Then things got started. So, thank to the community for that. But no nice presentation to share! I could relate some experience if that helps

ps2: @Corra have you checked the Pronet code-aster.org/V2/spip.php?rubrique62 page regarding support?

AsterO'dactyle

Hello,

Don't worry.

These are legitimate and very frequent questions.

We have to deploy our software on many different types of media. The best tool at the moment is containers. A container contains salome_meca, but there's nothing to stop you using only code_aster, especially in the case of problems such as graphics cards.

Internally, code is qualified on platforms such as Debian10. Soon Debian11.

You can always compile code_aster externally. The only difference is that we no longer make "aster-full" packages, which had become too time-consuming. There are many prerequisites for code-aster. The salome-meca container contains them all, with the guarantee that they'll work whatever happens.

In addition, the sources are updated every week (in fact, every day) on the Gitlab repository, so nothing has changed in that respect.

About the support. The answer is: no.

EDF is an electricity producer, not a software vendor.

Some companies have tried to set up this kind of service, but the problem is that the French market is too small and it's not profitable for them.

For a long time now, we've known that the only solution is for the open-source community to organize itself so that this kind of service can emerge.

It's possible, but it's difficult. Don't forget that a mechanical computation code is an extremely complex object on a restricted market. Look at the size and power of the software editors in this market (ANSYS, Dassault Systèmes, etC.). This is no coincidence...

pbedenbaugh

AsterO'dactyle,

Do you have a group of core users in the USA ? Perhaps a suitablly positioned user would be able to apply for funding from the USA National Science Foundation for an open source software community development grant, the deadline is pretty soon, however.

https://new.nsf.gov/funding/opportunities/pathways-enable-open-source-ecosystems-pose
https://new.nsf.gov/funding/opportunities/pathways-enable-open-source-ecosystems-pose/nsf23-556/solicitation

The eligibility is pretty broad, although if a commercial institution without prior government grants or contracts would need to start right away to fulfill registration requirements in time. The recipient has to be USA based.

They offer up to $300,000 for a planning of community grant, or up to $1.5 million over two years for a sustainable community implementation and expansion grant.

krande

Hey, for what its worth, there is ongoing work in the open source community for improving the distribution and ease of installing Code Aster.

For example the weekly releases of Code Aster are distributed on conda-forge. See https://github.com/conda-forge/code-aster-feedstock. Currently only linux is supported for python 3.8-3.11 (3.12 support is ready, but not yet released). And Windows (MSVC) support is on its way (see https://github.com/conda-forge/code-aster-feedstock/issues/65 for the progress tracker).

Assuming you have a conda distribution installed (like https://github.com/conda-forge/miniforge)
to install Code Aster for a given python version simply open the miniforge prompt and type:

mamba create -n myca-env code-aster python=3.11

That will install code-aster and all the necessary dependencies in a new python environment inside your conda installation called "myca-env".

Please note that the conda-forge distribution of code-aster is still very much under development and is by no means perfect. We are still working on resolving compatibility with the existing conda-forge ecosystem and we are actively working on improving it.

Best Regards
Kristoffer

jeanpierreaubry

hello

thanks for this effort but i am disappointed as

the following command
as_run $full_path/$exportfile.export
results in the following error
<F>_FILE_NOT_FOUND fichier non trouvé : $HOME/miniforge3/testing/config.txt

which is more or less the same with the official code_aster version since 2023 version

and astk fails with the same error as in the official version

any hints how to solve the issue

jean pierre aubry

krande

jeanpierreaubry

Good catch! I was able to reproduce this locally on my machine also.

I guess it is because I (and I guess Luca also) uses primarily run_aster instead of as_run. You should be able to simply substitute as_run with run_aster (I tried that just now, and I worked fine).

Fyi: we use as_run in our tests on the compiled package in conda-forge:

  commands:
    - as_run --info
    - homard || true  # [unix]
    - as_run --test forma01a
    - as_run --test sslp114a
    - as_run --test mumps05a
    - as_run --test mumps01a
    - as_run --test umat001a
    - as_run --test zzzz121a

I have created an issue ticket on our conda-forge feedstock -> https://github.com/conda-forge/code-aster-feedstock/issues/77 to keep track of it. Based on the error message, I believe it should be easy to fix.

Unfortunately I don't have any experience with astk (and the error message was a bit more cryptic). If you have any example commands and example files for running astk you could share here, I could take a quick look?

Best Regards
Kristoffer

jeanpierreaubry

thanks for that flash answer

copying the skeletal forma01a files to my usual working dir tree and running run_aster with the export file does the job

running a more realistic problem with the same style of skeletal .export does the job

running with a more realistic .export containing a lot more infos fails

i think there must be some incompatible stuff in the .export file

so much for the compatibility with older problem files!!!!!!!!

regarding astk
in the grand old days just issuing astk command would open the astk.tcl that was a great help in setting up all the parameters of a problem
it was creating a true life .export file around 50 or more lines
a little bit more than the less than ten lines used in the examples that looks like a joke if you want to tackle a true life problem

on the first launch it would also create an astkrc dir in $HOME that contained a bunch of information about the environment

and this was compatible from version to version from the beginning until 2020 and with a little tweaking until 2002
i can prepare an example and send you the file via email

k_zurawski

I hesitated a while to share my opinion. Despite the fact that I invested at least a thousand+ hours
in learning the tool I'm really considering moving to some other. The last Salome_Meca I have running stable is 2020 under Ubuntu 20.04 (which will be outdated in a month). I tried Debians + containers, Ubuntus + containers, Win10 WSL2 + containers... Result? I'm getting SIGSEGV, GlibC version problems, random shutdowns of S_M, reloading data in paravis results losing model in viewport and a lot more... Site is outdated, it's inaccessible today when I found time to dive into documentation (Parse error: syntax error, unexpected identifier "outil", expecting "," or ";" in /home/webcodeaos/www/plugins-dist/porte_plume/porte_plume_fonctions.php on line 34), I don't know where to look for the comparison between v15.x v16.x v.17. What will be used, what new things do I need to learn? I don't know and I don't know where to get the information from. I spent a lot of time on Singularity, which was supposed to be a cure for all the evil. And after downloading the newest version of S_M, well, it is not...

When AsterO'dactyle wrote about animating the free community it struck me. I understand you are focused on solving your industrial problems, but since you made a decision to go opensource you could at least find the budget, spend more time and provide one stable version every few years. We don't need to be animated. But I think even the most devoted believers are tired of solving the same problems all over again. Look on the topics in the forum after you issue another version...

I'm a mechanical engineer, not a programmer. I'd like to focus on solving mechanical problems, not IT stuff. I don't know anything about git, gitlab, github. It's not my field of expertise. I'd gladly do whatever you need me to do to have a stable environment. If you'd say install system x on machine with this specification, install Singularity version x.x, do this and that... and it will work. And if you need some more information, go here and there. That would be enough. I'd even buy a new or old machine with required specification. I think for you guys is one day work and for us discovering solutions all over again takes ages. I understand that you don't have budget or resources to do update the site. But be honest, updating a website is rather easy job, maybe a day or two once a year. But if you still think this is too much, there are few of us you could trust to help you out guys. Maybe code-aster.org should also go opensource with some guidance from your side?

Yes, I'm tired and I'm complaining. I'm not a malcontent, there are reasons for what I wrote.

AnatoleCG

Hello everyone, i was thinking about it since i see some people complain about the difficulty around code_aster learning and how to use it. As a beginner myself, i think it could be cool to create a fandom of code_aster where people can share some example of code or anything else and to regroup the tutorials we can see online. Fandom is an easy tool to create a database like that so i'm proposing that if you guys are interested in creating one. we could also share our methodology and all 🙂

k_zurawski

AnatoleCG I wasn't referring to difficulty around learning, I was referring to major version changes and no information what has changed, what is required, which Singularity etc... I'm not able to configure my system to have stable Salome_Meca installation which renders software unusable in real case scenarios. As an example - I won't risk spending few days and have problems visualizing outcomes.

There's a lot of useful information online, there are at least 4 books about Salome_Meca, there are companies that provide training, and you can find some videos on YT. But yes, this could be gathered in one place as someone mentioned before (Wiki pages).

k_zurawski

AnatoleCG several years ago I came across this:
http://caelinux.org/wiki/index.php/Doc:Code-Aster
I think after logging in you could contribute.

LiuShibo

The code_aster and Salome_Meca are both very good softwares. However, it is esay to use it but difficult to install it.

In Linux, is it possible to provide just one command to install code_aster or Salome_Meca?
In Window, is it possible to provide an .exe file for installation?

For other topup functions, such as one command line in Conda to install, shouldn't it be step 2, step 3, or even step n?

These improtant developments are crucial to increase user number, providing more opportunities for industry collaborations.

Further consideration, if no offical site, no standard, no easy installation and the learning curve very steep, with limited users, it is not good for this software to further develop.

Really hope Code_Aster or Salome_Meca have a bright future. And companies grow up from C_A or S_M, just like the Redhat from Linux.

k_zurawski

LiuShibo From what was said before the move from normal installation to Singularity container was to ease installation and maintenance. In theory your point 1 has been achieved. In theory, because all the "container" versions of Salome_Meca are unstable.

There's a Windows port of Salome_Meca (https_://code-aster-windows.com/download/), but is not provided by Code_Aster development team.

LiuShibo

k_zurawski Hi Zurawski, thanks for the comments.

Personally, I still think using the Singularity is somehow not that straigthforward. Especially, when I am not familiar with Singularity and have to install it (need to spend some time to install Singularity, and maybe even more time if the dependencies of Singularity are not fully found and installed). Thus, I said that: "it is esay to use it but difficult to install it. " It would be very great if Code_Aster or Salome_Meca can provide a xxxx.tar.gz for installation (no matter how big it is). Or some command like " pip install ....". (The key idea is easy to install it, with fewer steps and no bugs hides on the installation path).

For the Windows port of Salome_Meca, actually I think it is very great and I am acutally using it. However, one concern is that it is not the most updated one. The other is that it may disappear for later versions. It would be great if this can get offical support for versions every year or every 5 years if too time consuming. (like Ubuntu LTS version)

Opinions from a user who wish all the best for the C_A and S_M.

FYI

Best regards,

Shibo

k_zurawski

LiuShibo Singularity - I spent a lot of time trying to understand what it is, watching youtube, learning commands and some trial and errors...

Is the Windows version stable for you? In my case it's shutting down due to some vtk errors while working with paravis.

ib-guder

Hello. I'm in CA since 2009. I started with CAELinux and learned a lot from the tutorials. I have booked courses at code-aster. de and that also helped alot.

The situation currently with new colleagues who are learning to use Code Aster is that they are using a very well-known AI as support. It has an answer for everything. You can ask her to create complete *.comm files. You can have any error messages explained in detail simply by copying and pasting the error messages (in any language). I am now without further value as far as training new colleagues in Code Aster is concerned and can devote myself to new things.
Try it out - it still blows my mind!
Tom

AnatoleCG

ib-guder Hello, do you know what IA they are using cause i use chatgpt 4o and it is not answering this well. Or maybe i have to train him with all code_aster documentation maybe?

ib-guder

AnatoleCG We did not train chatGPT at all. I consider it knows the docs because they are available. Here is a quick example: h ttps://chatgpt.com/share/e/b7e362b7-3bea-40d2-9c19-37b2935ce525

mf

ib-guder
This does not work, by design, conversations are hidden to other users.