yeah one way of thinking about the DFT on an N-sample window/signal is it does N filters independently and the FFT is an optimisation that reduces a lot of redundant work between the filters. filters is a pretty broad term but for linear filters and dsp and stuff this isn't an analogy and is exactly what is happening. since sines/cosines/complex eponentials go on forever theyre a little unusual but in terms of transfer functions* they are very easy to interpret
*wikipedia not linked to preserve the innocent
i hate looking at wiki articles for something like this because it will shit out everything and you're forced to look at it and understand it