Unable to use S3 presigned HTTP GET URL while attempting to create a dataframe from URL

[denismakogon]

Code Sample, a copy-pastable example if possible

def from_file(path_or_url: str, amplification_peak: np.float64):
    df = pd.read_csv(
        path_or_url,
        names=[TIMECODE, AMP_PEAK],
        dtype={TIMECODE: np.float64, AMP_PEAK: np.float64}
    )

where the URL is:

http://docker.for.mac.localhost:9000/ffmpeg/af503ec9-08a8-4aff-a3a9-c9c3b0d8475f.csv?X-Amz-Algorithm=AWS4-HMAC-SHA256\u0026X-Amz-Credential=admin%2F20181101%2Fus-east-1%2Fs3%2Faws4_request\u0026X-Amz-Date=20181101T084722Z\u0026X-Amz-Expires=3600\u0026X-Amz-SignedHeaders=host\u0026X-Amz-Signature=60de48f6828baa2f83f8528ead22b4346044913202d2bc19020a0a6043334a57"

Problem description

Well, starting 0.18 AFAIK pandas allows to read dataframes from the URLs. So, presigned S3 URLs must not be an exception.

Unfortunately, with S3 presigned URLs pandas fails to read the dataframe and instead of raising any kinds or an exceptions it just returns an empty dataset.
At first i thought that URLs is bad or expired, but no, it's not and using cURL or wget I still can get that whole file from there.

Expected Output

Expected output here would be a non-empty dataset.

Output of pd.show_versions()

0.23.4

[gfyoung]

That's really weird...I wonder if it's the combination of Docker and S3 that is causing the issue. Investigation would be useful to determine what the root cause is.

[denismakogon]

I can write a script that will test this behavior.