typing for NumpySegmentationExtractor

[bendichter]

roiextractors/src/roiextractors/extractors/numpyextractors/numpyextractors.py

Lines 168 to 325 in 22ccdad

	class NumpySegmentationExtractor(SegmentationExtractor):
	"""A Segmentation extractor specified by image masks and traces .npy files.
	NumpySegmentationExtractor objects are built to contain all data coming from
	a file format for which there is currently no support. To construct this,
	all data must be entered manually as arguments.
	"""
	extractor_name = "NumpySegmentationExtractor"
	installed = True # check at class level if installed or not
	is_writable = True
	mode = "file"
	installation_mesg = "" # error message when not installed
	def __init__(
	self,
	image_masks,
	raw=None,
	dff=None,
	deconvolved=None,
	neuropil=None,
	accepted_lst=None,
	mean_image=None,
	correlation_image=None,
	roi_ids=None,
	roi_locations=None,
	sampling_frequency=None,
	rejected_list=None,
	channel_names=None,
	movie_dims=None,
	):
	"""Create a NumpySegmentationExtractor from a .npy file.
	Parameters
	----------
	image_masks: np.ndarray
	Binary image for each of the regions of interest
	raw: np.ndarray
	Fluorescence response of each of the ROI in time
	dff: np.ndarray
	DfOverF response of each of the ROI in time
	deconvolved: np.ndarray
	deconvolved response of each of the ROI in time
	neuropil: np.ndarray
	neuropil response of each of the ROI in time
	mean_image: np.ndarray
	Mean image
	correlation_image: np.ndarray
	correlation image
	roi_ids: int list
	Unique ids of the ROIs if any
	roi_locations: np.ndarray
	x and y location representative of ROI mask
	sampling_frequency: float
	Frame rate of the movie
	rejected_list: list
	list of ROI ids that are rejected manually or via automated rejection
	channel_names: list
	list of strings representing channel names
	movie_dims: tuple
	height x width of the movie
	"""
	SegmentationExtractor.__init__(self)
	if isinstance(image_masks, (str, Path)):
	image_masks = Path(image_masks)
	if image_masks.is_file():
	assert image_masks.suffix == ".npy", "'image_masks' file is not a numpy file (.npy)"
	self.is_dumpable = True
	self._image_masks = np.load(image_masks, mmap_mode="r")
	if raw is not None:
	raw = Path(raw)
	assert raw.suffix == ".npy", "'raw' file is not a numpy file (.npy)"
	self._roi_response_raw = np.load(raw, mmap_mode="r")
	if dff is not None:
	dff = Path(dff)
	assert dff.suffix == ".npy", "'dff' file is not a numpy file (.npy)"
	self._roi_response_dff = np.load(dff, mmap_mode="r")
	self._roi_response_neuropil = np.load(neuropil, mmap_mode="r")
	if deconvolved is not None:
	deconvolved = Path(deconvolved)
	assert deconvolved.suffix == ".npy", "'deconvolved' file is not a numpy file (.npy)"
	self._roi_response_deconvolved = np.load(deconvolved, mmap_mode="r")
	if neuropil is not None:
	neuropil = Path(neuropil)
	assert neuropil.suffix == ".npy", "'neuropil' file is not a numpy file (.npy)"
	self._roi_response_neuropil = np.load(neuropil, mmap_mode="r")
	self._kwargs = {"image_masks": str(Path(image_masks).absolute())}
	if raw is not None:
	self._kwargs.update({"raw": str(Path(raw).absolute())})
	if raw is not None:
	self._kwargs.update({"dff": str(Path(dff).absolute())})
	if raw is not None:
	self._kwargs.update({"neuropil": str(Path(neuropil).absolute())})
	if raw is not None:
	self._kwargs.update({"deconvolved": str(Path(deconvolved).absolute())})
	else:
	raise ValueError("'timeeseries' is does not exist")
	elif isinstance(image_masks, np.ndarray):
	NoneType = type(None)
	assert isinstance(raw, (np.ndarray, NoneType))
	assert isinstance(dff, (np.ndarray, NoneType))
	assert isinstance(neuropil, (np.ndarray, NoneType))
	assert isinstance(deconvolved, (np.ndarray, NoneType))
	self.is_dumpable = False
	self._image_masks = image_masks
	self._roi_response_raw = raw
	assert self._image_masks.shape[-1] == self._roi_response_raw.shape[-1], (
	"Inconsistency between image masks and raw traces. "
	"Image masks must be (px, py, num_rois), "
	"traces must be (num_frames, num_rois)"
	)
	self._roi_response_dff = dff
	if self._roi_response_dff is not None:
	assert self._image_masks.shape[-1] == self._roi_response_dff.shape[-1], (
	"Inconsistency between image masks and raw traces. "
	"Image masks must be (px, py, num_rois), "
	"traces must be (num_frames, num_rois)"
	)
	self._roi_response_neuropil = neuropil
	if self._roi_response_neuropil is not None:
	assert self._image_masks.shape[-1] == self._roi_response_neuropil.shape[-1], (
	"Inconsistency between image masks and raw traces. "
	"Image masks must be (px, py, num_rois), "
	"traces must be (num_frames, num_rois)"
	)
	self._roi_response_deconvolved = deconvolved
	if self._roi_response_deconvolved is not None:
	assert self._image_masks.shape[-1] == self._roi_response_deconvolved.shape[-1], (
	"Inconsistency between image masks and raw traces. "
	"Image masks must be (px, py, num_rois), "
	"traces must be (num_frames, num_rois)"
	)
	self._kwargs = {
	"image_masks": image_masks,
	"signal": raw,
	"dff": dff,
	"neuropil": neuropil,
	"deconvolved": deconvolved,
	}
	else:
	raise TypeError("'image_masks' can be a str or a numpy array")
	self._movie_dims = movie_dims if movie_dims is not None else image_masks.shape
	self._image_mean = mean_image
	self._image_correlation = correlation_image
	if roi_ids is None:
	self._roi_ids = list(np.arange(image_masks.shape[2]))
	else:
	assert all([isinstance(roi_id, (int, np.integer)) for roi_id in roi_ids]), "'roi_ids' must be int!"
	self._roi_ids = roi_ids
	self._roi_locs = roi_locations
	self._sampling_frequency = sampling_frequency
	self._channel_names = channel_names
	self._rejected_list = rejected_list
	self._accepted_list = accepted_lst

The typing here is a bit confused and going in a few different directions.

• There are no typehints for the __init__
• The docstrings say that the types should be np.ndarray, but if you look at the code they are all clearly meant to be strings that are filepaths.
• This seems to be an attempt to be analogous to SpikeInterface NumpySorting and NumpyRecording, however the __init__s for both of these classes take in-memory np.ndarray objects (as the docstring types would suggest), not filepaths (as the code actually implements.