Compiling and testing custom version of NCrystal and plugin

[nicriz]

Hi @tkittel,

Following a request from @ramic_k, I tried to implement a theta min cut in the integration of the q*I(q) curve to calculate the total SANS cross section with the SANSND plugin. This would be useful to model transmission experiments where below a certain experimental theta min the emitted neutron is considered unperturbed and hence no SANS contribution should be evaluated. To do so I decided to try to add new features to IofQHelper and PointWiseDist.

I am developing the new features here https://github.com/nicriz/ncrystal/tree/thetaMin_transmission

I compiled the custom ncrystal_core and ncrystal_python in a new miniconda environment on SLES 15 :

(NCrystal_thetaMin) rizzi_n@login002:~/ncrystal (thetaMin_transmission) $ pip install ./ncrystal_core
Processing ./ncrystal_core
  Installing build dependencies ... done
  Getting requirements to build wheel ... done
  Installing backend dependencies ... done
  Preparing metadata (pyproject.toml) ... done
Building wheels for collected packages: ncrystal-core
  Building wheel for ncrystal-core (pyproject.toml) ... done
  Created wheel for ncrystal-core: filename=ncrystal_core-4.2.10-py3-none-linux_x86_64.whl size=2825373 sha256=641ea92c10d1e0b76e09f690d35f3c88f44d063cc9003cb7cfcfb4d992907ee8
  Stored in directory: /tmp/pip-ephem-wheel-cache-2gumq47m/wheels/77/13/17/4648dede77c7cd50183b54a187d931a76d05c005ee6307cd99
Successfully built ncrystal-core
Installing collected packages: ncrystal-core
  Attempting uninstall: ncrystal-core
    Found existing installation: ncrystal-core 4.2.10
    Uninstalling ncrystal-core-4.2.10:
      Successfully uninstalled ncrystal-core-4.2.10
Successfully installed ncrystal-core-4.2.10

But when I tried to compile the SANSND plugin, the compiler complains that it cannot find the new function I added:

(NCrystal_thetaMin) rizzi_n@login002:~/ncplugin-SANSND (thetaMin_transmission) $ pip install .
Processing /data/user/rizzi_n/ncplugin-SANSND
  Installing build dependencies ... done
  Getting requirements to build wheel ... done
  Installing backend dependencies ... done
  Preparing metadata (pyproject.toml) ... done
Requirement already satisfied: ncrystal-core>=3.9.85 in /data/user/rizzi_n/miniforge3/envs/NCrystal_thetaMin/lib/python3.11/site-packages (from ncrystal-plugin-SANSND==0.0.1) (4.2.10)
Collecting ncrystal-pypluginmgr>=0.0.3 (from ncrystal-plugin-SANSND==0.0.1)
  Using cached ncrystal_pypluginmgr-0.0.5-py3-none-any.whl.metadata (14 kB)
Using cached ncrystal_pypluginmgr-0.0.5-py3-none-any.whl (13 kB)
Building wheels for collected packages: ncrystal-plugin-SANSND
  Building wheel for ncrystal-plugin-SANSND (pyproject.toml) ... error
  error: subprocess-exited-with-error
  
  × Building wheel for ncrystal-plugin-SANSND (pyproject.toml) did not run successfully.
  │ exit code: 1
  ╰─> [36 lines of output]
      2025-11-28 17:22:47,423 - scikit_build_core - INFO - RUN: /usr/bin/cmake -E capabilities
      2025-11-28 17:22:47,438 - scikit_build_core - INFO - CMake version: 3.28.3
      *** scikit-build-core 0.11.6 using CMake 3.28.3 (wheel)
      2025-11-28 17:22:47,439 - scikit_build_core - INFO - Implementation: cpython linux on x86_64
      2025-11-28 17:22:47,445 - scikit_build_core - INFO - Build directory: /tmp/tmpecm2ccor/build
      *** Configuring CMake...
      2025-11-28 17:22:47,467 - scikit_build_core - INFO - RUN: ninja --version
      2025-11-28 17:22:47,472 - scikit_build_core - INFO - Ninja version: 1.13.0
      2025-11-28 17:22:47,473 - scikit_build_core - INFO - RUN: /usr/bin/cmake -S. -B/tmp/tmpecm2ccor/build -DCMAKE_BUILD_TYPE:STRING=Release -C/tmp/tmpecm2ccor/build/CMakeInit.txt -DCMAKE_INSTALL_PREFIX=/tmp/tmpecm2ccor/wheel/platlib -DCMAKE_MAKE_PROGRAM=ninja
      loading initial cache file /tmp/tmpecm2ccor/build/CMakeInit.txt
      -- The CXX compiler identification is GNU 7.5.0
      -- Detecting CXX compiler ABI info
      -- Detecting CXX compiler ABI info - done
      -- Check for working CXX compiler: /usr/bin/g++ - skipped
      -- Detecting CXX compile features
      -- Detecting CXX compile features - done
      -- Configuring done (0.3s)
      -- Generating done (0.0s)
      -- Build files have been written to: /tmp/tmpecm2ccor/build
      *** Building project with Ninja...
      2025-11-28 17:22:47,847 - scikit_build_core - INFO - RUN: /usr/bin/cmake --build /tmp/tmpecm2ccor/build
      [0/2] Re-checking globbed directories...
      [1/6] Building CXX object CMakeFiles/NCPlugin_SANSND.dir/src/NCPluginBoilerPlate.cc.o
      [2/6] Building CXX object CMakeFiles/NCPlugin_SANSND.dir/src/NCPhysicsModel.cc.o
      FAILED: [code=1] CMakeFiles/NCPlugin_SANSND.dir/src/NCPhysicsModel.cc.o
      /usr/bin/g++  -pthread -B /data/user/rizzi_n/miniforge3/envs/NCrystal_thetaMin/compiler_compat -DNCPLUGIN_NAME=SANSND -DNCPlugin_SANSND_EXPORTS -DNCRYSTAL_NO_CMATH_CONSTANTS -I/data/user/rizzi_n/ncplugin-SANSND/src -isystem /tmp/pip-build-env-a9lq91mn/overlay/lib/python3.11/site-packages/_ncrystal_core/data/include -O3 -DNDEBUG -fPIC -MD -MT CMakeFiles/NCPlugin_SANSND.dir/src/NCPhysicsModel.cc.o -MF CMakeFiles/NCPlugin_SANSND.dir/src/NCPhysicsModel.cc.o.d -o CMakeFiles/NCPlugin_SANSND.dir/src/NCPhysicsModel.cc.o -c /data/user/rizzi_n/ncplugin-SANSND/src/NCPhysicsModel.cc
      /data/user/rizzi_n/ncplugin-SANSND/src/NCPhysicsModel.cc: In member function ‘double NCrystalPlugin_SANSND::PhysicsModel::calcCrossSection(double) const’:
      /data/user/rizzi_n/ncplugin-SANSND/src/NCPhysicsModel.cc:425:64: error: ‘const class NCrystal::IofQHelper’ has no member named ‘calcQIofQIntegralMin’; did you mean ‘calcQIofQIntegral’?
                   SANS_xs = 2 * NC::kPi / (k * k) * m_helper.value().calcQIofQIntegralMin(ekin);
                                                                      ^~~~~~~~~~~~~~~~~~~~
                                                                      calcQIofQIntegral
      [3/6] Building CXX object CMakeFiles/NCPlugin_SANSND.dir/src/NCTestPlugin.cc.o
      [4/6] Building CXX object CMakeFiles/NCPlugin_SANSND.dir/src/NCPluginFactory.cc.o
      ninja: build stopped: subcommand failed.
      
      *** CMake build failed
      [end of output]
  
  note: This error originates from a subprocess, and is likely not a problem with pip.
  ERROR: Failed building wheel for ncrystal-plugin-SANSND
Failed to build ncrystal-plugin-SANSND
error: failed-wheel-build-for-install

× Failed to build installable wheels for some pyproject.toml based projects
╰─> ncrystal-plugin-SANSND

When I check the ncrystal-config --show cmakedir I get correctly:

(NCrystal_thetaMin) rizzi_n@login002:~/ncplugin-SANSND (thetaMin_transmission) $ ncrystal-config --show cmakedir
/data/user/rizzi_n/miniforge3/envs/NCrystal_thetaMin/lib/python3.11/site-packages/_ncrystal_core/data/lib/cmake/NCrystal

What am I missing here?
Also feel free to comment with the actual implementation if it's flawed or can be improved

Cheers,
Nico

[tkittel]

Quick answer from my phone: You are 100% sure you do not have the NCrystal headers installed anywhere else on your system, in a location where the compiler might accidentally pick it up?

[tkittel]

Another point: What if you have multiple scatterings? Why should a secondary scattering use the same thetamin as the primary scattering?

[nicriz]

Yeah you are right but I think the main inteded use would be to retrieve the XS. Of course users may misuse this theta min in a Monte Carlo. The conceptual thing that interests me, is that for elastic SANS there is no clear boundary between transmitted and scattered beam if you just initialize the material and retrieve the XS. With theta min, people might stop for a second to think where is this boundary for their applications. If there is no clear need for one, theta_min=0.

[tkittel]

Well, the philosophy I am trying to put into the core NCrystal code (plugins are allowed to do otherwise), is that NCrystal should try to provide reasonable answers for any input neutron state and should not truncate output phase space either. Because doing otherwise always ends up adding a lot of requirements on how much the users have to understand in order to use the programme, and increases the possibility of someone getting misleading results (and in my subjective experience, users are often wrong about what they think is important - even "experts" like you and me often gets this wrong). And then we are back at "simulations are hard to configure" which is one of the things I would like to change with NCrystal 😄

[nicriz]

I see, those a really good points. Let's implement it in the plugin for now. My final point though will be that there is not a lot of ways to make SANS XS evaluation easy without putting some thinking (or parameters) into it. Even without theta min, there are already some quite obscure parameters. The safety plan would be to fall back on theta_min = 0 in any case, since that wouldn't really hurt that much.

[tkittel]

Absolutely, making things easy is hard ;-)

Agreed, put it in the plugin. But perhaps also emit a warning (with the NCPLUGIN_WARN macro) when thetamin is not 0, to remind the user to not use this in cases where multiple scattering is not negligible, or the geometrical layout not completely certain.