open TELEMAC-MASCARET The mathematically superior suite of solvers

Put yourself in the shoes of an outside user of Telemac who has successfully completed a simulation and wants to see the time series of a flow quantity at one location, in double precision.

Here I report amply on how troublesome and frustrating it is to visualize results with TELEMAC with a naive use of the information shown in the website.
Hope this user's dismay experience is helpful to improve the present state of
things -- assuming this is perceived as worth improving.

The page www.opentelemac.org/index.php/installation is the 'Introduction to the openTELEMAC-MASCARET system installation'. The building block 'Results visualisation and processing' (seemingly relevant) mentions four items, without giving any link to them. The links are the essence of the internet, after all.
These tools are

Assertion #1 Wouldn't it be friendly if a link is provided towards a fetch point or to more information that explains how possible/impossible is to retrieve these utilities?

Let's see what happens when one goes in search of these items (starting from exactly here).

1) FUDAA-Prepro

Google helps. There is a Sourceforge repository at sourceforge.net/projects/fudaa/
I can download the file prepro-ui-1.3.0RC11-install.zip. Side remark, I work with a Linux system preferably. After installation I run the script 'supervisor.sh' that immediately throws the error "bash: ./supervisor.sh: /bin/bash^M: bad interpreter: No such file or directory". The file lisez-moi.txt only mentions the software version, pretty uselessly.

Stackoverflows helps -- stackoverflow.com/questions/14219092/bas...r-directory#14219160. I save `supervisor.sh` with my text editor setting the line ending as Unix/Linux type instead of Windows.

Then, one discovers that the bash syntax is incorrect in that file. The script says ""$PWD fudaa-prepro-1.3.0.jar"" but it should be "$(pwd)/fudaa-prepro-1.3.0.jar". Clearly this script has never been tested.

Assertion #2-4 Wouldn't it be friendly that the shell script is provided using Unix/Linux formats and without blunders? Wouldn't be nice that files are tested before being shared? Or have a more informative lisez-moi.txt, perhaps called README?

Then FUDAA-Prepro starts. I look for data in the folders where my TELEMAC run has been successfully completed... FUDAA picks up nothing except directories, that I try to load. It asks for a file of type POST only and only of that type with the error message: Impossible de charger ce répertoire, Soit il n'est pas de type .POST soit in ne contient pas de fichier setup. Clear if you are conversant with Romance languages: you cannot load this folder, either it's not a .POST type or it does not contain the setup file.

More basically: What is a POST type? What is the `setup file`? Where is this all explained? Where are the manuals for FUDAA-Prepro? As a newcomer I can still be willing to make up for my knowledge gaps...

Let's talk about manuals, then. The manuals for Telemac are linked to in the home page www.opentelemac.org/index.php/manuals but a keyword search on POST in the Telemac2D User and Technical Manuals leads to no clue on what FUDAA demands.

By the way, if I go back to the Sourceforge page for FUDAA there's actually a folder `fudaa_docs` sourceforge.net/projects/fudaa/files/fudaa_docs/ with three subfolders with uninformative names... last modified between 2004 and 2006, bad moon rising... One of them is documentation on Java, irrelevant, the rest is to discover. In sum, how deep should one dig before getting to informative naming... and sure there's the rest of the internet to search in... and perhaps tens of forum threads to sieve from... I give up.

Further assertions. Wouldn't be nice that these manuals oF FUDAA are made available together with those of TELEMAC in the first pages of the website? The usability of FUDAA would gain a lot from this.


As it is now, it feels like visualization and post-processing are not regarded a part of flow modelling. What a poor treatment for a building block...

Blue-Kenue

Google helps. Accessible and pretty well documented at www.nrc-cnrc.gc.ca/eng/solutions/advisory/blue_kenue_index.html. But version 3.4.4 does not handle double-precision output files from TELEMAC. And I would love to have those double-precision results. Plus, it's Windows-only, so you have to work around this too.

Davit

Google does not help. The more I try to restrict the search, the more it tries to convince me that I was looking for something else. Bad moon rising...

MATLAB-based tools

Only if I move away from the home page and discover the download area of the TELEMAC website (www.opentelemac.org/index.php/download) can I then find those.
A-ha, so there is a page called `supporting tools` www.opentelemac.org/index.php/supporting-tools with one item from the visualisation and post-processing building block! The tools are at
nl.mathworks.com/matlabcentral/fileexchange/25021-telemac-tools -- So eventually there is a glimpse of hope... you only have to despair to enjoy it near the starting point.

Incidentally, here it is pity that MATLAB is not quite as open-source as TELEMAC, which is a regretful inconsistency, but not an uncommon one. For good measure, out of the four items listed, no one is accessible, is open-source and deals with double-precision numbers at the same and one time.

Overall

Overall, the newcomer spends hours trusting clues, searching for user-friendly, working tools. One can now perhaps understand how irritating it can be to poke the forum post and get back anything in the line of 'read the manual first' or go trial-and-error blindly.

Again, I hope this user's dismay experience is helpful to improve the present state of things -- again, assuming this is perceived as worth improving.


P.S. There's also the Paraview option, but the TELEMAC web page www.opentelemac.org/index.php/installation does not mention it, strangely.

Hello,

Thank you -- this is helpful indeed and keep us going while on shoe strings.

You should also know - you are not going to like it - that there are two other software packages (at least) coming along to the mix ... these are not mentioned here yet because they are in a testing mode:

- a post-processing QGIS plugin (works everywhere), developed by ARTELIA on the basis of the existing python scripts.
- a very comprehensive GUI called SALOME, developed by EDF and Co., available on Linux.

Therefore, I would like to re-emphasise that we are doing our very best to provide the best software environment we can ... and also that it was, for us as a consortium, better to open up the source code of the TELEMAC system in 2010 than holding on it while concentrating on graphics ...

Good things come to those who wait.

Sébastien.

In my view the root problem is that selafin is a binary format. If there was a way to output ASCII files one could write his/her own scripts. I might well have missed this possibility... if so, please anyone point me towards the way to get that.

Sébastien, great answer, thank you --- does that Linux GUI really work? (Sorry for the shameless question...) --- can I get hold of it? (Even more shameless...)

I appreciate anyone's good work, but I cannot deny that precision helps and pays off when it comes to software development and numerical modelling. Personally, I dislike false economies.

I feel that you realised how bad is to work with a numerical model of which inspecting the results is quagmire or a minefield. It's like an engine without wheels. Overall, I am glad that you saw purpose in my mail, happy work with that.

Hello,

Just a quick personal feedback. I indeed can remember some dark times when I first started with TELEMAC about 6 years ago 100% self-taught. Since then I introduced other colleagues to the software and tried to lower their learning curve’s slope.

But if we assume that new comers will mainly work with TELEMAC-2D on Windows (read Blue Kenue), it is actually rather easy to start using the program oneself by following the tutorials available on the homepage (www.opentelemac.org/index.php/training-and-tutorials) which provides a detailed workflow step by step (and even click by click). I have example of colleagues succeeding running a first 2D model this way (almost) on their own. I now have about 5-10 colleagues working with TELEMAC and I would say that the main part of them gets sorted out with Blue Kenue only.

After, when increasing model complexity it is indeed hard not to increase the workflow complexity. But that’s something to be expected and one should always remember that it is indeed great to be able to work with such open source tools (which often offer more functionalities than commercial alternatives) for free.

If I can understand the advice about “read the manual first”, I can also think that this should be associated with “… and then search on the forum” where there is a LOT of useful information.

And while we speak about how to improve the whole package, my personal wish would be to have a clear documentation about the numerous and powerful python scripts. There are certainly many tools (both pre and post-processing) hidden in those scripts that only the expert python user can have benefit of at the moment. A documentation could dramatically increase their use I think…

Best regards
PL

Hello,
Watermotion.eu wrote:
That's true, it is not easy to read/write selafin files correctly (little/big endian, single/double precision). But there are some python scripts distributed with TELEMAC which allow the user in 2d (sorry I'm only work in 2d) to get the data. I think you can make such scripts also for 3D.

www.opentelemac.org/index.php/kunena/oth...lotlib-pytel-scripts

www.opentelemac.org/index.php/kunena/scr...-on-mesh-with-python

I prefer to extract the data (mainly arrays) from the selafin files matplotlib for plotting an data analysis in 2d, but it should be easy to convert the data into another format. So maybe this is a possibility for you?


@DAVID
this is like Janet a product of smile consult, a company.
A further way might be Tecplot with a plugin, but most people convert selfain to the Tecplot format.


In my opinion it is not easy to make a good decision which data format should be preferred/supported by a numerical code like TELEMAC. VTK might be widely supported in the Open Source community, but it has also different types and some drawbacks (e.g reading XML-VTK is horrible). Maybe some developers prefer formats that support parallel write (MPI) into one file to speed up things. Some codes start to use NETCDF or only HDF5, but there is no standard for unstructured grids (some is on the way).
I think selafin + python tools of TELEMAC are a good way, maybe there should be some improvements when writing selafin files, but its ok.


Best regards

Leo

Lufia wrote:
I am thankful for this tip and I am playing around with this to see how far it can help. At the same time this already brings new arguments as to why the look-in-the-forum approach is far from optimal.

The posts indicated are years old and the terminology is outdated. There is nothing like named 'pytel' in the distribution any longer. So, instead of one solution, the newcomer finds one more problem: I did not know yet what I was about to find (just trusting), but now I don't even know its name (trusting a bit less).

Nowadays the folder is called `python27`. You get to know this starting one another forum search, in search of a soul that has been kind enough to mention this point. Perhaps you find it, perhaps not. Forum searching are like quests for the holy grail: you only pursue it as long as you believe that something to get exists. (I did find a tip, or one of many, but I leave you the joy to experience that discovery moment yourselves, if you ever want).

OK, I understand that if you are looking for Python scripts for Telemac, wherever else can they be than in the folder <telemac version folder>/scripts/python27?

But like John Lennon and others, pytel lives on. And where? In the website that has never been updated. Try yourself www.opentelemac.org/index.php/search?q=pytel and see where you end up. So anyone who uses Telemac to understand Telemac is too far doomed to face confusing information. One more question arises: should I perhaps double check my understanding again?

But probably, by looking in the forum, you will also find someone who --- in all good faith and with the best intentions --- said 'true, we should have updated the website, we are working on that'. I understand this too.

But the main message is: how does this compare with a link on one of the front web pages that orderly lists all possibilities for post-processing simulation results, with a short description of what they do, where they are and which limitation the user incurs? To which extent ease of access to correct information is optional for open-source systems?

I go back and check out those scripts. After all I need that time series in double precision mentioned at the beginning of this thread. Thanks to all contributors, by the way.

My personal point of view.

Yes, using TELEMAC is sometimes hard, the documentation is often outdated,... we can make a long list of things that must be fixed or should be improved. So it can be frustrating to work with TELEMAC and it is no excuse that some other open source /commercial codes are not better. It is possible to get some commercial support, not sure if this can help some of the beginners.

On the other side, there are Open Source projects (e.g. scipy, matplotlib,...) with a extreme good documentation, easy to use and with many tutorials.

The reason for this situation is perhaps the history and how the development is driven. The question we should ask is: How can we improve TELEMAC?

There are some users that help and write install instructions, some users don't like or have no time to do so. Looking at the list of things that should be improve at TELEMAC everyone can help to do so. It is not easy to contribute to the main code but it is easy to help at other parts.

Hello,

soon we will provide the first informations about TUC 2017! Maybe the training workshop in the next Telemac User Conference 2017 could be a good opportunity to introduce the capabilities of the Python scripts?! Furthermore QGIS and SALOME could be hot topics. The details of the training workshop will be announced in a later step.

@Sébastien: could we launch a poll about the topics for the training workshop?

Best regards,
Clemens

OK. So you want to extract time series data at a point from a double precision 3d *.slf file, run my extract_pt.py script (see this post).

pprodano wrote:
Kind of you. I could write my own in the meantime but will have a look for benchmarking asap.