source_simple -> mcstas-pygen -> mcstasscript -> exception

[tkittel]

In a very simple instrument file with a Source_simple :

COMPONENT source = Source_simple(
    lambda0=lambda,
    dlambda=0.001,
    focus_yh=0.01,
    focus_xw=0.01
)
AT (0, 0, 0) RELATIVE PREVIOUS

I tried to export via the PyLab button in mcgui and run in mcstasscript. It mostly works, but when using

instr.show_diagram(analysis=True)

I end up with an exception:

 /opt/miniforge/envs/ESS/lib/python3.12/site-packages/mcstasscript/instrument_diagnostics/diagnostics_instrument.py:107, in DiagnosticsInstrument.correct_target_index(self)
    103 # Only here if target_index is used, correct it in modified instr
    104 
    105 # Find original index and the name of the original target
    106 original_comp_index = original_comp_names.index(comp.name)
--> 107 comp_target_name = original_comp_names[original_comp_index + int(comp.target_index)]
    109 # Find index of the original and target in modified instrument
    110 modified_comp_index = modified_comp_names.index(comp.name)

ValueError: invalid literal for int() with base 10: '+ 1'

It turns out that this is caused by the default value of a Source_simple parameter having a target_index=+1 as default value, which apparently gets converted to python as:

    # Comp instance source, placement and parameters
    source = instr.add_component('source','Source_simple', AT=['0', '0', '0'], AT_RELATIVE='origin', ROTATED=['0.0', '0.0', '0.0'], ROTATED_RELATIVE='origin')
    
    source.radius = '0.1'
    source.yheight = '0'
    source.xwidth = '0'
    source.dist = '0'
    source.focus_xw = '0.01'
    source.focus_yh = '0.01'
    source.E0 = '0'
    source.dE = '0'
    source.lambda0 = 'lambda'
    source.dlambda = '0.001'
    source.flux = '1'
    source.gauss = '0'
    source.target_index = '+ 1'

And apparently the space in between + and 1 in the last line gives issues.

[willend]

I dug in the comp repo, this list of McStas components look affected:

mcstas-comps/obsolete/ESS_moderator_short.comp:branchframe=0, int target_index=+1)
mcstas-comps/contrib/ISIS_moderator.comp:focus_yh = 0.01, xwidth = 0.074, yheight = 0.074, CAngle = 0.0,SAC= 1, Lmin=0, Lmax=0, int target_index=+1, verbose=0)
mcstas-comps/sources/Source_Maxwell_3.comp:int target_index=+1,lambda0=0, dlambda=0)
mcstas-comps/sources/Source_simple.comp:flux=1, gauss=0, int target_index=+1)
mcstas-comps/sources/Moderator.comp:SETTING PARAMETERS (radius=0.07, Emin, Emax, dist=0, focus_xw=0.02, focus_yh=0.02, t0=37.15, Ec=9.0, gamma=39.1, int target_index=+1, flux=1)
mcstas-comps/sources/Source_gen.comp:Lmin=0,Lmax=0,Emin=0,Emax=0,T2=0,I2=0,T3=0,I3=0,zdepth=0, int target_index=+1)
mcstas-comps/sources/Source_adapt.comp:  int target_index=+1, alpha=0.25, beta=0.25)

and none from McXtrace.

[willend]

@tkittel thanks for this one. My thinking is that perhaps the + is (wrongly) interpreted as "there is probably an expression, so use a string".

Will try to get this fix before next McStas release.

[willend]

Hmm. Have now identified where the space comes from:

McCode/mccode/src/cexp.c

Line 129 in 04c6062

separator = " "; /* Now use space separator for rest. */

... i.e. we are taking quite deep in the code generation engine, definition of "separators" between c-expression parts...

So in a way I am inclined to "fix" this in the related comps...

But just to get an idea of potential side-effects, I am experimenting a little over on https://github.com/mccode-dev/McCode/tree/issue-2175-fix - will start the test-suite and we may have a look...

[willend]

With the cexp.c correction, the test-suite looks like this: https://tmp.mcstas.org/run_test-suite_output.html I will double-check wrt. the nightlies tomorrow, but my feeling is they will align.