iohub cli convert making dtype to np.float64

[edyoshikun]

I have noticed this with the ndtiffs in mantis datasets
/hpc/instruments/cm.mantis/2023_08_03_argolight_resolution/argolight_patterns_7/argolight_patterns_lightsheet_1/

Where after using iohub version0.1.0.dev3, the conversion returns:

Raw data:

(pyplay) [eduardo.hirata@cpu-b-1 argolight_patterns_7]$ iohub info -v /hpc/instruments/cm.mantis/2023_08_03_argolight_resolution/argolight_patterns_7/argolight_patterns_lightsheet_1/
Reading file:    /hpc/instruments/cm.mantis/2023_08_03_argolight_resolution/argolight_patterns_7/argolight_patterns_lightsheet_1/
Dataset opened                 
=== Summary ===
Format:          ndtiff
Positions:       1
Time points:     1
Channels:        1
Channel names:   ['GFP EX488 EM525-45']
(Z, Y, X):       (601, 700, 2048)
Z step (um):     0.01

Zarr Converted data:

iohub info -v argolight_patterns_lightsheet_1.zarr/
Reading file:    argolight_patterns_lightsheet_1.zarr/
Zarr hierarchy:
/
 └── 0
     └── 0
         └── 0
             └── 0 (1, 1, 601, 700, 2048) float64

=== Summary ===
Format:          omezarr v0.4
Axes:            T (time); C (channel); Z (space); Y (space); X (space); 
Channel names:   ['GFP EX488 EM525-45']
Row names:       ['0']
Column names:    ['0']
Wells:           1

[ziw-liu]

What's the NDTiff version in your env?

[ziw-liu]

I was able to reproduce with iohub==0.1.0dev3 and ndtiff==1.9.0, but iohub==0.1.0dev4 and ndtiff==2.2.1 converts to uint16 correctly.

[ziw-liu]

My intuition is that #145 makes the difference. It may just because ndtiff v1.x does not handle V3 correctly.

[edyoshikun]

Most likely need to bump recOrder to depend on iohub==0.1.0dev4 as this iohub version does convert properly. Closing this issue.