Handle Situation where V1 Visium tissue_positions_list.csv Has a Header

[jeffquinn-msk]

I've been given a new dataset where the positions file is in the V1 location spatial/tissue_positions_list.csv, but it has a header. Really annoying I know. This ends up causing an exception in the Visium loader as the coordinates are loaded as string type by pandas. You might see similar data eventually as well I'm guessing..

Could we add something to the code to check for header in both locations? This is the function I wrote to deal with this:

def read_with_maybe_header(path, header):
    df = pd.read_csv(path)
    seen_header = df.columns
    if set(seen_header).issubset(set(header)):
        return df
    else:
        return pd.read_csv(path, names=header)

[LucaMarconato]

Hi @jeffquinn-msk @inodb, thanks for reporting. Do you know any publicly available dataset that presents this structure?

[LucaMarconato]

To reproduce I am taking the visium_1.3.0_io dataset from the sandbox and replacing the tissue_positions_list.csv file (=> no header) with the tissue_positions.csv file (=> with header) from the visium_2.1.0_1_io dataset from the sandbox, and renaming it to tissue_positions_list.csv. In this way we have a < 2.0.0 filename, but with a header (as found with 2.x datasets)

[LucaMarconato]

Btw, this issue reminds me of this one (Xenium), where data processed with an old version of Xenium Onboard Analysis is partly reprocessed with a recent version of SpaceRanger, creating confusion when parsing. Anyway, the case that you reported seems easy to address, I'll work on a fix now.

[LucaMarconato]

Hi made a fix in the linked PR. Some details:

• the 1.3.0 dataset with the borrowed tissue position file from 2.1.0 could be read if the new name was kept in the 2.1.0 format (i.e. tissue_positions.csv; the problem (now fixed) was when the name was changed to the old format.
• I tested the new PR also on standard versions, running this test here: https://github.com/giovp/spatialdata-sandbox/blob/e818c976bd49a0c75de496fc49569f74b3ddd695/tests/test_to_zarr_visium.py.