Render user-friendly string representation of open files

[piskvorky]

Currently, an open S3 object prints like this:

>>> f = smart_open.open('s3://…')
>>> print(f)
<smart_open.s3.SeekableBufferedInputBase object at 0x10ad02b38>

This is not useful, and we can do much better: we have full info about the concrete bucket, key, mode, etc. Compare to boto3's nice native display:

>>> print(f._object)
s3.Object(bucket_name='some-bucket', key='test/bank_form.pdf')

Task: render open objects more nicely in __str__. This is great help during logging and debugging (I consider a human-friendly str() best practice, and define them religiously on custom classes).

It seems easy enough to do for S3, but I'm not sure how much work for our other supported storages (HTTP, FTP or whatever), for consistency.

[mpenkov]

We could implement __str__ and __repr__ for our custom classes, which would solve the problem partially.

What should we do for non-custom wrapper classes? e.g.

import sys
from smart_open import open

print('str read_text: ' + str(open(sys.argv[1])))

gives

str read_text: <encodings.utf_8.StreamReader object at 0x7fa0b1c14eb8>

We don't really have control of that object. What's the best thing to do there?

[piskvorky]

I'd leave builtins as-is. "Falling back to built-in where possible" is a well-communicated property of smart_open. A nicer display would be welcome, but not essential.

Also, I never bother overriding __repr__ (and never miss it), but I'm open to better approaches :)

[zlatsic]

Hi!

I'm new to OSS and am looking to contribute somewhere, mind if I give this a shot?

Just to clarify, the behavior you're looking for is to related to the __str__ and __repr__ methods of objects created with the s3://... descriptor? Which properties would you like to be shown in the representations? Just the bucket and key or additionally, some of the other, optional properties?

[mpenkov]

Closed via #359