Move panda flyer

[Thomas Hopkins (thopkins32)]

Summary

Primarily does two things:

• Ports the HXNFlyerPanda device + dependencies from hxn-profile-collection
• Fixes the descriptor shapes declared by the device

I made these two changes as separate commits so you can clearly see where the shape fix comes into play.

Motivation / Context

While doing an audit of descriptor data across NSLS-II beamlines, I discovered that there are 240 data keys that describe their data shape incorrectly, all stemming from the HXNFlyerPanda device. In this case, the declared shape was simply 10x what it was supposed to be.

I intentionally port the device definition here so we can write proper tests against this device, such that no reversion can easily occur. I intentionally left the device code unchanged, although it may be prudent to adopt the ophyd-async version in the near future and remove this code entirely. How to handle SIS data and Xps3 ROI data with the ophyd-async device is an open question...I think we will have to rewrite how this is done as well to make it compatible.

Evidence

• Keys (all in the fly-scan stream, n_events = 1):
  • Det{1,2,3}_<element>[_L|_M] — per-detector element maps (e.g. Det1_Fe declared [49000], stored [4900])
  • Det4_* — declared [64000], stored [6400]; Det4_Pt_L [100] → [10]
  • sclr1_ch1 … sclr1_ch16 — declared [49000], stored [4900]
  • xspress3_channel{1,2,3}_rois_roi01 … roi16 — declared [1000], stored [100]
• Stored lengths are scan pixel counts (4900 = 70², 3136 = 56², 6400 = 80², 2500 = 50², 10000 = 100², 22500 = 150², 200 = 20×10 …). Declared = pixels × 10.

[Copilot]

Copilot review overview

🟡 Changes recommended

Acquisition correctness issues and missing CI/lockfile integration must be addressed.

Get a fresh assessment by requesting another Copilot review.

Review effort: Balanced
Findings: 5 · 1 · 1

Open (7)

• uv.lock is stale and missing declared development dependencies · New
• MCA length check truncates short arrays instead of padding · New
• Timeout uses wrong units and reports partial acquisition as success · New
• Event timestamps incorrectly use datum IDs instead of numeric times · New
• External dataset descriptors omit the float32 dtype · New
• CI does not run pytest regression tests · New
• Command requests undefined test extra instead of the dev group · New

What changed in this PR

Ports the PandA flyer into hxntools and corrects oversampled scaler/ROI descriptor shapes.

Changes:

• Adds HXNFlyerPanda and HDF5 exporters.
• Adds a Tiled-backed shape regression test.
• Adds test dependencies.

File	Description
src/​hxntools/​panda_flyer.py	Adds flyer acquisition and descriptor logic.
tests/​test_panda_flyer.py	Tests declared and stored array shapes.
pyproject.toml	Adds integration-test dependencies.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

Copilot review overview

🟡 Changes recommended

The flyer has data-integrity, timeout, multi-detector, timestamp, and CI-lint issues that should be fixed first.

Get a fresh assessment by requesting another Copilot review.

Review effort: Balanced
Findings: 2 · 1

Open (3)

• Remove unused dataset assignment · New
• Replace fieldless f-string with plain string · New
• Track exporter state per Xspress3 device · New

Resolved since last review (7)

• External dataset descriptors omit the float32 dtype
• Event timestamps incorrectly use datum IDs instead of numeric times
• Timeout uses wrong units and reports partial acquisition as success
• MCA length check truncates short arrays instead of padding
• uv.lock is stale and missing declared development dependencies
• CI does not run pytest regression tests
• Command requests undefined test extra instead of the dev group

[Thomas Hopkins (thopkins32)]

Waiting to test at beamline before we merge.

[Thomas Hopkins (thopkins32)]

Dan Henriksen (@dihenriksen) good to merge?