web.Application object have a boolean value of False

[Ricordel]

Long story short

Not a huge issue but a bit unsettling: the boolean value of web.Application() is False

Expected behaviour

import aiohttp.web
app = aiohttp.web.Application()
if app:
    print("true")
else:
    print("false")

should print true

Actual behaviour

Prints false

Steps to reproduce

Create an app like in the above code

Your environment

OS: Linux Ubuntu 19.04
python: Python 3.7.3
aiohttp and deps:

aiohttp==3.6.1
async-timeout==3.0.1
attrs==19.1.0
chardet==3.0.4
idna==2.8
multidict==4.5.2
pkg-resources==0.0.0
yarl==1.3.0

[asvetlov]

Why do you think so?

[Ricordel]

Because it seems to me that a boolean value of False should represent something that is nothing (None), logically or historically false (False, 0) or empty ("", [], {}). I don't think that a constructed and functional web.Application falls into any of those categories.

Also from experience, it seems to me that any random object from any random library has a boolean value of True unless there is a good reason not to (there may be one here I'm not aware of), so this goes against the principle of least astonishment. Python code like:

if subapp:
    app.add_subapp('/foo', app)

or

app = this_user_supplied_app or this_backup_app

will behave in a way that is unexpected for the random Python developer (which I am, I got bitten by the first one).

[asvetlov]

app is a mapping BTW (app["key"] = value). The idiomatic code should check for None:

if subapp is not None:
    app.add_subapp('/foo', app)

[Ricordel]

Yes I could figure that out.

I still find the behavior counter-intuitive compared to what all the rest of Python is usually doing, my arguments are in the previous comment. You are free to disagree, and considering your contributions to this project (which I thank you for) you are naturally the one with the final word, and I won't argue (especially since it's not a big problem and it's not preventing anyone to use aiohttp for any reasonable reason).

[asvetlov]

Ok. I forgot that we had the same problem for exceptions.

Would you want to make a pull request for web.BaseRequest and web.Application? The change can look like here.

A couple of trivial tests are required.

[Ricordel]

Ok, I'll try to do that in the coming days.

[asvetlov]

Fixed by #4124