[ESSNMX] Mandi

[YooSunYoung]

@aaronfinke I could use most of the mcstas xml helpers and essnmx-reduce code to make the mandi reduction tool.

There are still a few todos, like reducing monitor data but the detector banks could be saved in the NXLauetof file just like NMX files. Here is the command if you want to try:

essmandi-reduce \
  --input-file MANDI_10383.nxs.h5 \
  --verbose  #  --overwrite  # if you want to ignore safety check of the output file name.

Then it'll save the output to scipp_mandi_output.h5, if you don't specify the output.

[aaronfinke]

Binned data look good, but the geometries are not in the binned file. From Claude:

The bug

src/ess/mandi/_idf_helper.py, DetectorDesc.pixel_positions:

return (
(pixel_n_slow * slow_axis_steps)
+ (pixel_n_fast * fast_axis_steps)
+ self.rotation_matrix * sc.vector([self.start_x, self.start_y, 0.0], ...)
)

self.position — declared on line 61 as # 'x', 'y', 'z' — is never added. It's used only for its .unit, on the line above. Same in assemble_detector_metadata, which takes a transformation parameter and reads only transformation.value.unit.

Adding + self.position should give origin = the bank's xyz (0.4138 m for bank1, matching the IDF exactly), and distance then becomes a real per-bank value. The sample is at (0,0,0) per the function's own comment, so relative and absolute coincide.

[aaronfinke]

For mandi it would be good to include the instrument_xml data in the binned output, since mandi is still generally dependent on the mantid IDF. should be propagated throughout the workflow and included in the output.

[YooSunYoung]

@aaronfinke thank you! Forgot the pixel offset.
Now I could reconstruct the geometry from the reduced file:

Is this correct...?

[aaronfinke]

Looks right to me. Also loaded it in my new NMX 3D viewer (out soon!) and geometries all look correct. Excellent!

[YooSunYoung]

I think the implementation itself is okay based on the reduced data it produces.
I need add tests and such...

@aaronfinke can I use these Mandi file in our CI?
(most of events will be stripped out due to size issue)

Also need more fixes for monitors and crystal rotation later...