coolprop
coolprop copied to clipboard
ibell
Matlab compiling and vectorized matlab function
Here is some modifications to make Matlab compiling go smoother. (the original did not compile due to -ansi which was default on my installation.) In addition there is new function which take vector input and output. The aim was to be able to use your excellent work as an add on to open source Matlab Reservoir ToolBox www.sintef.no/Projectweb/MRST/
Your work was very useful to be able to test realistic fluids for simulation of reservoir with temperature. I have seen that you had started on an interface for mixtures. I would be very intrested to test that for our code also.
See also http://www.mathworks.com/matlabcentral/answers/269-using-c-style-comments-in-mex-files or http://www.mathworks.com/matlabcentral/answers/8870-c-comments-in-mex-files-under-linux - I think this sort of idea is a nicer solution than a separate mexopts.sh file.
Apart from that, nice work. I am looking to deprecate the Props function fairly soon, so only using the PropsSI variants. Could you not have added your code as a conditional branch in the PropsSI mex file? That would be a nicer solution I think.
Wrapping C/C++ libraries for MATLAB sucks, and it involves a lot of ugly hacking. Have you looked into python before? I guess with your existing library, its a bit less exciting...
Also, we are completely rewriting the guts of CoolProp, partly in order to support mixtures. Watch this space. It's taking a lot more time than originally planned, which is not really surprising.
Hei
Thank you for your answer. We are doing quite some C++ programing http://www.opm-project.org/ and have Wrapped c++ code to matlab before and know the trouble. In a bit longer time we will look at using some of your fluids in the c++ code. In this project the "advanced " fluids are based on dumux. www.dumux.orghttp://www.dumux.org
MRST is of historical reasons in matlab. Today we probably would have started in python (even it exist advantages of the matlab programing enviroment also) If we at some point have time( and money), we would like to translate some of the MRST code to python. I do not think it will take long time to get a OK simulator up and running.
I will look at putting the vectorized version as a branch in PropsSI.
Best regards Halvor M. Nilsen
@hnil Thanks for your reply. If you can implement the vectorized code as a conditional branch in PropsSI and implement the build process as I suggest, I'd be happy to merge that PR into CoolProp.
Do the people in your project know about CoolProp? If not, who should I contact to let them know about it? I figure if we are both open-source, it would be great if we could work together. We could use some manpower on our side too. Would welcome it really.
I might be able to help a bit with the python side of things. I have quite a good amount of experience with SWIG, and experience with Cython too. There's also the new xdress project. So there are pretty good tools for python, certainly much more mature than MATLAB. I don't deny that MATLAB has some nice features, but I find it extremely hard to justify the price they are trying to charge for a full license when you can't even use C++ in MATLAB, amongst many other small issues. Why pay thousands when python offers most of the same functionality for $0?
On Tue, Mar 25, 2014 at 10:44 AM, hnil [email protected] wrote:
Hei
Thank you for your answer. We are doing quite some C++ programing http://www.opm-project.org/ and have Wrapped c++ code to matlab before and know the trouble. In a bit longer time we will look at using some of your fluids in the c++ code. In this project the "advanced " fluids are based on dumux. www.dumux.orghttp://www.dumux.org
MRST is of historical reasons in matlab. Today we probably would have started in python (even it exist advantages of the matlab programing enviroment also) If we at some point have time( and money), we would like to translate some of the MRST code to python. I do not think it will take long time to get a OK simulator up and running.
I will look at putting the vectorized version as a branch in PropsSI.
Best regards Halvor M. Nilsen
From: Ian Bell [mailto:[email protected]] Sent: 25. mars 2014 09:44 To: ibell/coolprop Cc: Halvor Møll Nilsen Subject: Re: [coolprop] Matlab compiling and vectoriced matlab function (#194)
See also http://www.mathworks.com/matlabcentral/answers/269-using-c-style-comments-in-mex-filesor http://www.mathworks.com/matlabcentral/answers/8870-c-comments-in-mex-files-under-linux- I think this sort of idea is a nicer solution than a separate mexopts.sh file.
Apart from that, nice work. I am looking to deprecate the Props function fairly soon, so only using the PropsSI variants. Could you not have added your code as a conditional branch in the PropsSI mex file? That would be a nicer solution I think.
Wrapping C/C++ libraries for MATLAB sucks, and it involves a lot of ugly hacking. Have you looked into python before? I guess with your existing library, its a bit less exciting...
Also, we are completely rewriting the guts of CoolProp, partly in order to support mixtures. Watch this space. It's taking a lot more time than originally planned, which is not really surprising.
Reply to this email directly or view it on GitHub< https://github.com/ibell/coolprop/pull/194#issuecomment-38541165>.
Reply to this email directly or view it on GitHubhttps://github.com/ibell/coolprop/pull/194#issuecomment-38545821 .
Hei
Sorry for abit late answere.
I have talked to some of the developers of OPM (c++ code). I think that it would Be a good idea be able to call your code from this code since at least some parts of it is related to simulations with CO2.
I do not disagree with you about matlab/python. But due to historical reasons it is at the moment no issue to change. We currently have about 2000 downloads of the last release. And our research is contract based so if we do not have founding for changing we will not do it.
Else I am using python on my spare time. What is your prefered working enviroment for python. Do you use something with integrated debugger like spyder? (The integrated enviroment at least for some years ago was one of the advantages of matlab)
When I have time I will look at the vertorized version and the compiling options. I will consult one of may colleages which has struggled with the mex interface many times.
Best regards Halvor
@hnil You might want to hold off with the C++ integration for a bit since the API is getting a full overhaul at the C++ level. I can't say when it will be finished. But the high-level integration will remain the same - the Props and PropsSI and IPropsSI functions will be supported indefinitely.
For python, I have used Eclipse+pydev - quite nice and free, debugger, etc. A bit difficult to get set up, but once you have, works fine. For lighter weight stuff, I use SciTe most of the time, or IEP is a new entrant to this market.
As soon as you've got the vectorized code ready, let me know. I don't have the time to modify/merge your changes right now unfortunately.
Hei
I will let you know have done it. No trouble for me. (I will try to test it on windows also.)
Best wishes Halvor