I see your point, but I think your Python's not terribly idiomatic and that's a big part of the problem.

This is easier to read and understand, only goes through the list twice, and loses nothing in terms of power:

    [j*j for j in [i+1 for i in a] if j%3 != 0]

(And for any given operation, there's very possibly a cleaner way to abstract out the inner list comprehension, which would again make it a lot nicer.)

In general, I don't see much of a reason to use filter/map/etc in Python: weak lambdas mean they're not terribly powerful. List/sequence comprehensions can do everything they can do with cleaner syntax and/or fewer operations.

This, I think, is actually at the core of this whole discussion. The Ruby/bash approach makes sense if you're used to working with sequences like Ruby/bash do. The Python approach is more natural to me though, because I've written a lot of Python. (And having spent the last year writing a lot of Ruby, I still find the Python approach cleaner/easier to understand at a glance.)