Ensure S3 tests are mocked

[TomAugspurger]

Anaconda owns the pandas-test S3 bucket that we use for some of our tests. They're doing some spring cleaning, and would like to remove this bucket sometime next week.

We'll need to make sure that all of our tests are mocked.

From what I can tell, pandas/tests/io/parser/test_network::TestS3 is not mocked.

We also have an ASV benchmark that hits S3, whose goal is to ensure that we don't read the whole file when doing read_csv(..., nrows=10). I think we could remove that and replace it with a mocked fixture that asserts something about the number of bytes read.

[jreback]

we could also sign up for an aws account with NuMFOCUS. would be minimal cost I think.

[TomAugspurger]

Yeah, it'd be good to have one non-mocked test under a bucket we have control over.

…

On Wed, Feb 21, 2018 at 5:48 PM, Jeff Reback ***@***.***> wrote: we could also sign up for an aws account with NuMFOCUS. would be minimal cost I think. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#19825 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABQHIlXgnYIhoxcCDsVONxpkKOOsxOt0ks5tXKtCgaJpZM4SOS16> .

[TomAugspurger]

We've temporarily set the pandas-test bucket to be unreadable, and it'll be going away entirely soon. Tests still pass for me locally, so hopefully things are OK.

There's one reference to s3://pandas-test in the docs, but that's in a non-executed code-block.