File object support in the Python API (fsspec)

[daweim0]

Preliminary Checks

• This issue is not a duplicate. Before opening a new issue, please search existing issues.
• This issue is not a question, bug report, or anything other than a feature request directly related to this project.

Proposal

Currently, pyzed functions involving files require a path to a file physically on disk. For example, the following code is valid:

init_params = sl.InitParameters()
init_params.set_from_svo_file("path/to/svo.svo")

However, this chunk of code is not

with open("path/to/svo.svo", "rb") as svo_file_obj:
    init_params = sl.InitParameters()
    init_params.set_from_svo_file(svo_file_obj)

My request is to modify set_from_svo_file() to accept file objects, not just paths to files.

Use-Case

This is extremely useful when SVO files are stored remotely. For example, I record SVO files on a computationally limited device, then save the SVO files to S3. Many Python libraries (including Numpy, Pandas, and PyAV) can read/write files directly from S3 like so

import fsspec
fs = fsspec.filesystem("filecache", target_protocol="s3")
with fs.open("s3://bucket/path/to/numpy/archive.npz", "rb") as numpy_archive:
    numpy_data = np.load(npz)

This is very convenient in two ways:

1. I (the user) don't have to download files I need ahead of time
2. Numpy maintainers don't have to write any s3 specific code. They add ~5 lines to the start of their functions involving files and suddenly they support every blob store under the sun.

I currently download SVO files before processing them, but I work with all other files (mp4, csv, ...) directly in s3.

Anything else?

I don't think SVO streaming solves my problem, since S3 servers aren't going to run ZED-specific code.