Python multi-line with statement

As of Python 3.10, it is now possible to parenthesize the whole group of context managers, as you originally tried:

with (Dummy() as a, Dummy() as b,
      # comment about c
      Dummy() as c):
    pass

This is also technically possible in 3.9, but in a sort of semi-documented limbo.

On one hand, it's documented as new in 3.10, 3.9 wasn't supposed to introduce any features (like this one) that depend on the new parser implementation, and the 3.9 with docs forbid this form. On the other hand, the functionality ended up getting activated in the 3.9 CPython implementation, and the (mostly?) auto-generated 3.9 full grammar spec includes the parenthesized form.

On previous Python 3 versions, if you need to intersperse comments with your context managers, I would use a contextlib.ExitStack:

from contextlib import ExitStack

with ExitStack() as stack:
    a = stack.enter_context(Dummy()) # Relevant comment
    b = stack.enter_context(Dummy()) # Comment about b
    c = stack.enter_context(Dummy()) # Further information

This is equivalent to

with Dummy() as a, Dummy() as b, Dummy() as c:

This has the benefit that you can generate your context managers in a loop instead of needing to separately list each one. The documentation gives the example that if you want to open a bunch of files, and you have the filenames in a list, you can do

with ExitStack() as stack:
    files = [stack.enter_context(open(fname)) for fname in filenames]

If your context managers take so much screen space that you want to put comments between them, you probably have enough to want to use some sort of loop.

As Mr. Deathless mentions in the comments, there's a contextlib backport on PyPI under the name contextlib2. If you're on Python 2, you can use the backport's implementation of ExitStack.

Incidentally, the reason you can't do something like

with (
        ThingA() as a,
        ThingB() as b):
    ...

before the new parser implementation is because a ( can also be the first token of the expression for a context manager, and CPython's old parser wouldn't be able to tell what rule it's supposed to be parsing when it sees the first (. This is one of the motivating examples for PEP 617's new PEG-based parser.

This seems tidiest to me:

with open('firstfile', 'r') as (f1 # first
  ), open('secondfile', 'r') as (f2 # second
  ):
    pass

This isn't exactly clean, but you could do this:

with Dummy() as a, Dummy() as b, (
     #my comment
     Dummy()) as c:
    pass

There are no syntax errors, but it's not the cleanest. You could also do this:

with Dummy() as a, Dummy() as b, Dummy(
     #my comment
     ) as c:
    pass

Consider finding a way of doing this without using the comments in the middle of the with.

I would keep things simple and readable by adding the comment before the with statement, or on the line itself:

# my comment explaining why I wanted Dummy() as c
with Dummy() as a, Dummy() as b,\
     Dummy() as c: # or add the comment here
    pass

Like TigerhawkT3's answer, but with indenting that doesn't trigger pycodestyle's error E124:

with (
        open('firstfile', 'r')) as f1, (  # first
        open('secondfile', 'r')) as f2:  # second
    pass

IMO it's still ugly, but at least it passes the linter.

To avoid too long lines, you can do the following.

Instead of:

with open(filename, 'wb') as file, request.urlopen(image.url) as response:
    pass

I, do for example:

f = lambda: open(filename, 'wb')
r = lambda: request.urlopen(image.url)
with r() as response, f() as file:
    pass