Add MATLAB Bindings #494

BenjaminRodenberg · 2019-08-29T09:07:31Z

Closes #102.

This branch currently contains all the changes by @Dominanz that were necessary for his thesis. One particularly useful contribution for preCICE are the MATLAB bindings that have been developed by @Dominanz. We want to merge @Dominanz implementation of the MATLAB bindings into https://github.com/precice/precice within this PR.

I would suggest to release the MATLAB bindings as soon as possible in order to be able to provide them to our users. However, I would suggest to explicitly declare them as experimental since I expect that we will need to make some experiences until we arrive at the final version of the Interface.

Requirements

We require the following components for the MATLAB bindings:

all source code that is necessary for installation of the MATLAB bindings
documentation, such that a user can install the MATLAB bindings
a solverdummy using the MATLAB bindings

Todo

Before we can merge this PR, we have to do the following things:

get rid of everything that is not directly connected to the MATLAB bindings (see requirements above)
add documentation, if it is missing (-> documentation is in good shape, more will follow in a later PR)
write tests for the MATLAB bindings (might also be postponed after merging, since we might need to test the MATLAB bindings using a system test on https://github.com/precice/systemtests/tree/master/Test_bindings)
update the Changelog
update the wiki as soon as PR is merged

…elop

Note: Bindings not finished, currently only those required for solverdummy are implemented.

…elop

Added bindings for three more functions. Fixed a minimal typo in solverdummy.

…elop

The MATLAB bindings now support both character arrays and strings for inputs where the C++ API requires strings.

Changed the write functions in SolverInterface (and the functions called by them) to take a const double* (resp. const double&) instead of a double* (resp. double) for the value argument. The purpose of these functions is to copy the contents of the buffer provided as value argument to the other solver. Hence, the access to the write buffer is read-only, consequently it can be const.

…elop This includes the const-correctness which is valuable for continuing the MATLAB bindings

…elop Adds the remaining changes for const correctness.

…elop

The compiling script now uses pkgconfig to determine necessary flags and matlabs `mfilename` to determine the path, making it portable. The dokumenation was extended. Many API functions were added, though several are still missing.

…elop

For writeBlockScalarData, the user may now specify to transpose the output if desired, which is helpful if the code is geared to work with column vectors (manual transposing causes unnecessary memory overhead). Fixed bugs in Constants class and mesh quad functions.

…elop

fsimonis

Do we need the legacy matlab bindings?

Once the new matlab interface is released, we end up with a total of 3 matlab bindings.

I cannot test nor review the .m files.

src/precice/bindings/matlab/+precice/@Constants/private/preciceGateway.cpp

tools/solverdummies/precice-config.xml

src/precice/impl/SolverInterfaceImpl.cpp

fsimonis

This code is solid but needs to be polished a bit.

Also, these bindings are not compiled into preCICE and therefore should not be in the src directory.
It is time that we tackle this issue and move such bindings into something like extra/bindings/.
Alternatively we could put them into a separate repository.

tools/solverdummies/matlab-legacy/solverdummy.cpp

fsimonis · 2019-09-04T11:40:24Z

tools/solverdummies/matlab-legacy/solverdummy.cpp

+  int meshID = interface.getMeshID(meshName);
+  int dimensions = interface.getDimensions();
+  mexPrintf(meshName.c_str());
+  std::vector<std::vector<double>> vertices(N, std::vector<double>(dimensions, 0)); 


We should populate the dimensions somehow.
std::generate and some random distribution maybe?

I'm not sure I understand what you mean by this.

fsimonis · 2019-09-04T11:43:21Z

src/precice/bindings/matlab/+precice/@SolverInterface/private/preciceGateway.cpp

+        bool valid=false;
+        // define the functionID
+        const TypedArray<uint8_t> functionIDArray = inputs[0];
+        int functionID = functionIDArray[0];


Can't we merge these two calls?

See my comment above.

src/precice/bindings/matlab/+precice/@SolverInterface/private/preciceGateway.cpp

BenjaminRodenberg · 2019-09-09T09:09:45Z

From my perspective unrealistic to get this done until the release. Let's postpone it to the next one.

- Switched to an enum class for the function IDs - Undid unrelated changes - Some minor fixes and changes

Dominanz · 2019-09-12T20:16:38Z

Do we need the legacy matlab bindings?

They are not meant to be bindings. The solverdummy is rather a minimal example showing how bindings would look like in the old C MEX API. If users are forced to use a pre-R2018a release of MATLAB, they might be interested in this. However, we could also just exclude them from the PR.

- Switched to an enum class for the function IDs - Undid unrelated changes - Some minor fixes and changes

BenjaminRodenberg

Can we entirely remove tools/solverdummies/matlab-legacy from the PR? Merging something that is already called legacy does not look right. edit: I renamed them and decided to keep them for the reasons @Dominanz is giving in #494 (comment)

src/precice/impl/SolverInterfaceImpl.cpp

src/precice/SolverInterface.hpp

src/precice/SolverInterface.cpp

…, the will be introduced via precice#530

BenjaminRodenberg · 2019-09-26T17:08:36Z

I just cloned the PR branch after merging develop, built preCICE (to work with MPICH), compiled the MATLAB bindings and the solverdummy works without a problem.

That's good news.

As soon as @fsimonis has aproved the PR, we can merge it. There are some comments in his review that we should either directly address in this PR or at least transfer to issues, such that we do not forget.

MakisH · 2019-10-08T15:25:35Z

This pull request has been mentioned on preCICE. There might be relevant details there:

https://precice.discourse.group/t/i-was-asked-to-try-a-pull-request-how-can-i-do-that/38/2

BenjaminRodenberg · 2019-10-24T10:46:53Z

I would like to merge this PR into a new feature branch precice:draft_MATLAB_bindings on the preCICE repository, now. Some issues still remain open end will be solved by another PR, before we merge everything into precice:develop. Using this approach, we can clearly identify @Dominanz contribution to the code and let @gilbertolem add his contributions on his own fork.

From my point of view the following issues still remain open:

put everything into extras/bindings see Add MATLAB Bindings #494 (review)
use enums for a clearer mapping of cpp to matlab API functions in the preciceGateway

@fsimonis do you agree? If yes, you can just merge this PR from my point of view.

Moved from https://github.com/precice/precice/tree/draft_MATLAB_bindings. Main contributions contained in this PR: * precice/precice#494 * precice/precice#580 Co-authored-by: Dominik <volland@ma.tum.de> Co-authored-by: Gilberto Lem <35875991+gilbertolem@users.noreply.github.com>

Dominik and others added 27 commits March 7, 2019 16:30

Added the matlab solverdummy

6154109

Merge branch 'develop' of https://github.com/precice/precice into dev…

102acaf

…elop

Added MATLAB bindings.

ea19897

Note: Bindings not finished, currently only those required for solverdummy are implemented.

Merge branch 'develop' of https://github.com/precice/precice into dev…

ec4da3a

…elop

Continued MATLAB bindings.

6ec91c8

Added bindings for three more functions. Fixed a minimal typo in solverdummy.

Merge branch 'develop' of https://github.com/precice/precice into dev…

bc4ed34

…elop

Added support for MATLAB strings

7242cf0

The MATLAB bindings now support both character arrays and strings for inputs where the C++ API requires strings.

Merge https://github.com/precice/precice into develop

7b66c95

Merge branch 'develop' of https://github.com/precice/precice into dev…

d215a65

…elop This includes the const-correctness which is valuable for continuing the MATLAB bindings

Merge branch 'develop' of https://github.com/precice/precice into dev…

423c18b

…elop Adds the remaining changes for const correctness.

Merge branch 'develop' of https://github.com/precice/precice into dev…

71d2348

…elop

Used workaround by @uekerman to fix static IDs

873a99f

Merge branch 'develop' of https://github.com/precice/precice into dev…

01f6f39

…elop

Largely refactored matlab bindings

1b85ba5

Merge branch 'develop' of https://github.com/precice/precice into dev…

88f04c2

…elop

Added most remaining API functions to the matlab bindings.

f728a6e

Merge branch 'develop' of https://github.com/precice/precice into dev…

16ec264

…elop

Removed constants in accordance with constants cleanup and fixed a typo

8e8cfa9

Allowed for out-of-process execution in the solver dummies.

2ad6620

Fixed problems with constants cleanup.

05f19ca

Removed another wrong inclusion

6c52fbe

Merge branch 'develop' of https://github.com/precice/precice into dev…

bbacd8d

…elop

Cleaned up matlab bindings and improved documentation

31bc248

Added OOP variant for MATLAB bindings

1768660

BenjaminRodenberg added the enhancement A new feature, a new functionality of preCICE (from user perspective) label Aug 29, 2019

BenjaminRodenberg self-assigned this Aug 29, 2019

fsimonis self-requested a review September 3, 2019 08:53

BenjaminRodenberg mentioned this pull request Sep 3, 2019

Implement a Matlab API #102

Closed

BenjaminRodenberg added this to the Version 1.6.0 milestone Sep 3, 2019

fsimonis reviewed Sep 4, 2019

View reviewed changes

fsimonis requested changes Sep 4, 2019

View reviewed changes

BenjaminRodenberg modified the milestones: Version 1.6.0, Version 1.x.x Sep 9, 2019

BenjaminRodenberg modified the milestones: Version 2.x.x, Version 2.0.0 Sep 12, 2019

Dominanz added a commit to Dominanz/precice that referenced this pull request Sep 12, 2019

Implemented further changes as suggested by @fsimonis in precice#494

c5f9d69

- Switched to an enum class for the function IDs - Undid unrelated changes - Some minor fixes and changes

Dominanz added 2 commits September 12, 2019 22:22

Cleaned up code, made some changes as suggested by @fsimonis

a505303

Implemented further changes as suggested by @fsimonis in precice#494

09dfc79

- Switched to an enum class for the function IDs - Undid unrelated changes - Some minor fixes and changes

BenjaminRodenberg commented Sep 26, 2019

View reviewed changes

src/precice/impl/SolverInterfaceImpl.cpp Outdated Show resolved Hide resolved

src/precice/SolverInterface.hpp Outdated Show resolved Hide resolved

src/precice/SolverInterface.cpp Outdated Show resolved Hide resolved

BenjaminRodenberg mentioned this pull request Sep 26, 2019

Use const double& for writeScalarData #530

Closed

BenjaminRodenberg added 4 commits September 26, 2019 13:28

Merge branch 'develop' into PR_draft_MATLAB_bindings

e922868

Remove changes for these files from this PR, if changes are necessary…

0615ee2

…, the will be introduced via precice#530

rename matlab-legacy

ed808fe

Provide information on workaround in documentation

40b16c7

BenjaminRodenberg changed the title ~~WIP: Add MATLAB Bindings~~ Add MATLAB Bindings Sep 26, 2019

Add Information on MATLAB bindings to Changelog

b3d7ede

BenjaminRodenberg changed the base branch from develop to draft_MATLAB_bindings October 24, 2019 10:40

fsimonis merged commit 2a23bb1 into precice:draft_MATLAB_bindings Oct 25, 2019

BenjaminRodenberg mentioned this pull request Dec 20, 2019

Add implementation of matlab-bindings precice/matlab-bindings#5

Merged

Uh oh!

Add MATLAB Bindings #494

Add MATLAB Bindings #494

Uh oh!

Conversation

BenjaminRodenberg commented Aug 29, 2019 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Requirements

Todo

Uh oh!

fsimonis left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

fsimonis left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

fsimonis Sep 4, 2019

Choose a reason for hiding this comment

Uh oh!

Dominanz Sep 12, 2019

Choose a reason for hiding this comment

Uh oh!

fsimonis Sep 4, 2019

Choose a reason for hiding this comment

Uh oh!

Dominanz Sep 12, 2019

Choose a reason for hiding this comment

Uh oh!

Uh oh!

BenjaminRodenberg commented Sep 9, 2019

Uh oh!

Dominanz commented Sep 12, 2019 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

BenjaminRodenberg left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

BenjaminRodenberg commented Sep 26, 2019

Uh oh!

MakisH commented Oct 8, 2019

Uh oh!

BenjaminRodenberg commented Oct 24, 2019

Uh oh!

Uh oh!

BenjaminRodenberg commented Aug 29, 2019 •

edited

Loading

Dominanz commented Sep 12, 2019 •

edited

Loading

BenjaminRodenberg left a comment •

edited

Loading