Allow feeding Pandas DataFrames directly to CSV datasets

[holgerroth]

Is your feature request related to a problem? Please describe.
In some situations, the user has not access to the csv files but some already generated pandas objects only.

Describe the solution you'd like
In that case, it would be useful to create CSVDataset and CSVIterableDataset directly from the pandas DataFrames.

Describe alternatives you've considered
n/a

Additional context
n/a

[Nic-Ma]

Hi @holgerroth ,

Thanks for your feature request.
It makes sense to me, I will try to enhance the CSVDataset and CSVIterableDataset for it soon.

Thanks.

[Nic-Ma]

Hi @holgerroth ,

BTW, I think CSVIterableDataset is designed to partly load extremely big CSV file or streaming, if you already have the DataFrame in memory, I think no need to use this class, so I will not enhance it.
What do you think?

Thanks.

[holgerroth]

Good point. I still think that the current class only accepts a filename as input and tries to use pd.read_csv() from disk. It might be useful if it can accept the streaming object directly as well.

[Nic-Ma]

Hi @holgerroth ,

OK, it makes sense to accept streaming iterable objects here:
https://github.com/Project-MONAI/MONAI/blob/dev/monai/data/iterable_dataset.py#L211

Thanks.