Repository navigation
[FEATURE]: Support input_path in generated make() functions from mcstas-pygen #2554
Description
Is your feature request related to a problem? Please describe.
I use mcstas-pygen to turn an .instr file into a .py script so it can be wrapped in a python package to add a complex detector geometry. The generated script creates a make() function that instantiates McStas_instr with only hardcoded parameters.
The problem is that my package expects custom local components, so McStasScript needs to know where they are (input_path) right at instantiation time. Because make() takes no arguments, I'm forced to manually edit the generated python file to add input_path to the function signature and constructor call each time the base instrument file is edited and the python file is regenerated.
Describe the solution you'd like
It would be great if mcstas-pygen generated the make function with a default parameter: def make(input_path=None):.
This would pass input_path directly to McStas_instr(..., input_path=input_path). Since the constructor signature defaults to input_path=None (resolving to "." internally (see: here)), I expet this change to be completely backward-compatible and won't break any existing scripts calling make() with no arguments.
Describe alternatives you've considered
My alternative is to continue manually editing or to automate overriding :)
Additional context
Looking at the source code, I hope this would be a very quick change in mccode/src/pygen.c:
#if MCCODE_PROJECT == 1 /* neutron */
cout("# Python McStas instrument description");
cout("def make(input_path=None):");
coutf(" instr = ms.McStas_instr(\"%s_generated\", author = \"McCode Py-Generator\", origin = \"ESS DMSC\", input_path=input_path)",instr->name);
#elif MCCODE_PROJECT == 2 /* xray */
cout("# Python McXtrace instrument description");
cout("def make(input_path=None):");
coutf(" instr = ms.McXtrace_instr(\"%s_generated\", author = \"McCode Py-Generator\", origin = \"ESS DMSC\", input_path=input_path)",instr->name);
#endifNote:
I wouldn't mind having other McStas_instr arguments as well (I manually add output_path and NeXus, but these can be modified through .settings() after the initialisation, so they are not really important)
Activity
willend commentedon Jul 6, 2026
@MilanKlausz this can absolutely be done. https://github.com/mccode-dev/McCode/tree/test-pygenmod-2554 is under way.
For (even minor cogen-) implementing suggestions / changes we always welcome a PR? :-) (Use https://github.com/mccode-dev/McCode/wiki/Build-your-own-McStas-or-McXtrace-via-conda%E2%80%90forge to easily ‘roll your own’ locally in a conda-env.)
BTW as an alteratvie, I would think that once you are in Python / notebook you would always be able to use the
mcrun -I some_path_with_compsvia the McStasScript settingcustom_flags=‘-I some_path_with_comps’MilanKlausz commentedon Jul 6, 2026
Thank you!
I'll do PR the next time I see a small enough change ;)
The
-Isuggestion made me realise thatinput_pathis rather a working directory change then an include path... and I have runtime data files that are being copied to that "instrument_code" directory, so my notebook sitting in another directory would need more than just-Ito make it run, not just compile.Merge pull request #2555 from mccode-dev/test-pygenmod-2554
willend commentedon Jul 6, 2026
I’ve merged #2555 to main - can you try building from that to check that the adaptation does what you want?
mads-bertelsen commentedon Jul 7, 2026
Doing exactly this was on my todo list! Agree with doing this only for input_path, as it's the only setting that can not just be updated later.
As you say input_path only changes where mcrun is executed, not your work directory for the notebook, but typically data files for the simulation can just be in the input_path as mcrun then finds them.
One exception is if you are using the add_ncrystal_union_material tool in mcstasscript with some custom ncmat files, then they are needed both when creating the instrument file (in python work directory) and when executing mcrun (in input_path). I may update that tool so it looks in the input_path as well!
willend commentedon Jul 7, 2026
As I remember,
-Iformcrun(andmcstasitself) at least historically added to the “data load path” also… - Or could be made to do so...MilanKlausz commentedon Jul 7, 2026
I've built from main (it WAS super easy to ‘roll my own’ mcstas, nice!), used the resulting mcstas-pygen, and I got exactly what I wanted:
Thank you!
willend commentedon Jul 7, 2026
Great, I think that is worth an incremental release, https://github.com/mccode-dev/McCode/releases/tag/v3.7.12 is slowly but surely on the way to
conda-forgenow.willend commentedon Jul 7, 2026
And good you liked the new procedure for ‘rolling your own’ @MilanKlausz. 👍 Any feedback welcome if you find issues/annoyances later.
FYI the ‘main’ McCode CI stuff is now based on the same approach.
willend commentedon Oct 20, 2026
Implemented by PR #2555 (
daa92c59f, merged 2026-07-06; released in v3.7.12), confirmed above by @MilanKlausz.mccode/src/pygen.c.innow generatesdef make(input_path=None):, defaulting to the script's own directory. For other setups, the alternatives discussed above also apply, e.g. McStasScriptcustom_flags='-I <path>'. Closing.(Issue triage done by Claude on behalf of @willend.)