Migrate to pymmcore plus

[jeskesen]

This PR is to address issue #170 .

[tlambert03]

quick note: mmc.run_mda is a convenience method for starting a thread and calling mmc.mda.run in that thread. If you just want to block immediately and wait, you can call mmc.mda.run(mda) directly and avoid the thread altogether.

[jeskesen]

Status 05/12/25

TLDR

• acquisition engine runs on pymmcore_plus using demo config
• file output is handled using acquire-zarr directly (no position handling yet).

Detail

As has been noted before, I'm trying to make the port as minimally invasive as I can. Thus, where possible, have just tried to use analogous data structures between pycromanager & pymmcore_plus. The main data structure conversions are:

• pycromanager.Core -> pymmcore-plus.CMMCorePlus
• pycromanager.Acquisition -> None
  • File writing functionality handled by acquire_zarr.ZarrStream
  • The Core runs the sequence of events directly
• pycromanager.multi_d_acquisition_events -> useq.MDASequence or Iterable[useq.MDAEvent]
• pycromanager.Studio -> TBD

The above handled most of the engine but there were a couple of modifications that needed to be made: both related to the looping structure, where the outer time & position loops are handled by ShrimPy, while the inner channel & z loops are handled by the engine, which allows autofocus and autoexposure to happen at each new position.

Previously, a copy of the acquisition_event instance was made and the copy was then modified to have the correct indices. That couldn't be exactly replicated using pymmcore-plus because MDAEvent inherits from a read-only class. To get around this, I took advantage of the fact that one can use a list of MDA Events as if it were a MDASequence, so I simply regenerate the sequence by iterating over it, and appending a new object to a list created with modified constructor args.

[jeskesen]

Closing this PR as I've created a duplicate with a different source branch at #177