sebi06 / czitools Goto Github PK

Congrats on releasing!

Question... The license on this package is BSD but pylibcziRW and aicspylibczi are GPL. I assume license should be updated unless I just don't understand license stuff.

Hi Sebastian
Thanks for making this resource available and maintaining it.
I am trying to read nd arrays from large datasets with pylibczirw_tools.read_5darray,. I set this flag output_dask=True, so it doesn't read in the entire dataset. However, it takes a very long time...

czi_img = pylibczirw_tools.read_5darray(img_path,output_order="TCZYX",
                               output_dask=True,
                               remove_adim=True
                                )

Is this the right way to do it? It takes more than 1 hour..

However, if I use the following code to initialize a dask array it takes ~5 minutes for a 255GB dataset

from dask import delayed as da_delayed
import dask.array as da 
from pylibCZIrw import czi as pyczi
from tqdm import trange



czidoc = pyczi.CziReader(img_path)
temp_frame = czidoc.read()
print("Array Shape: ",temp_frame.shape)
total_bbox = czidoc.total_bounding_box
z_slices = total_bbox['Z'][1]
nchannels = total_bbox['C'][1]
total_time = total_bbox['T'][1]

#initialise empty list to hold the dask arrays
img=[]


for time in trange(total_time):
    ch_stack=[]
    for ch in range(nchannels):
        z_stack=[]
        for z in range(z_slices):
            z_slice = da.from_delayed(
                                da_delayed(czidoc.read)(plane={'C': ch, 'T': time,'Z':z}), shape=temp_frame.shape, dtype=temp_frame.dtype
                            )
            z_slice = da.squeeze(z_slice)
            #z_slice = z_slice.swapaxes(0,1)
            z_stack.append(z_slice)
        z_stack = da.stack(z_stack,axis=0)
        #break
        ch_stack.append(z_stack)
    img.append(ch_stack)

image_stack = da.stack(img,axis=0)

I think it would be a good idea the ability to install czitools with minimum requirements. My use case is the use of czitools for reading czi files so it is pretty basic but the installation of czitools bring so much with it from plotting, graphing, other reading libraries (aicsimageio[all] which brings even more stuff such as bfio).

Proposal

Have the ability to do something like this:
pip install czitools[min]
pip install czitools[all]

Additional context

It looks like the library already thinking of this. For example napari is not in the requirements but it is being used. I just want to go further than that and say that I only want stuff for reading using pylibczirw and metadata (no plotting, no aicsimageio).

p.s. I am liking this so far. I've learned a few things reading the code. I think I can contribute here and there. I will need to do some tests first.