What causes this weird shape through frosted glass? I was looking at my frosted window and found that the light behind it from the neighbors house looked like this:

Behind the frosted glass is simply an ordinary CFL bulb. The frosting pattern on the glass is like a bunch of small squares:

What is causing this unusual shape to appear?
 A: What you are seeing is a direct result of refraction (not diffraction).
There is a repeating pattern in the "frosting" of the glass; this is like a bunch of small lenses. Now when a ray of light is incident on a particular point, it will be refracted through a certain angle, depending on the shape.
What you are seeing is the sum of these deflections - when you look a little bit down from the source, you see the bit of light that was refracted up; when you look up from the source, you will see the light that was shining down, etc.
One way you can convince yourself that this is the right explanation: take a laser pointer and shine it through the glass (preferably from a little distance, so the spot will be about as big as one "repeat pattern" of the frosting). Put a white sheet behind the glass, and take a picture of the pattern.
I predict it will look a lot like what you are seeing here (except it will all be one color - whatever the color of your laser pointer is). And if you repeat with a different color (wavelength), you will get the same pattern (shape and size) which confirms this is not diffraction. Diffraction would result in different size patterns for different wavelengths.
UPDATE
@Ruslan actually did the experiment I suggested, and asked that I edit the results into this answer. So with thanks, here it goes:
Using a piece of frosted glass similar to the one shown in the original post, he obtained three different pictures. The first shows what a white light looked like through the glass - clearly a similar pattern to the one OP saw:

The second was the result of shining a violet (405 nm) laser pointer through the glass: the refracted light shows a similar pattern:

Finally, using a red laser results in the same pattern (size) even though the wavelength is 50% longer. If this was diffraction, the pattern would have been correspondingly larger:

All this seems to confirm the explanation.
