Spencer Russell

We’ll need to add a deprecation message to the 3-arg `conv2` anyways guiding people with how to change their code, so having them swap the argument order shouldn’t be a...

go for it!

I think a RecipesBase dependency is fine, and this doesn't need to be split out into a separate package.

It seems like either having `rect!`, `gaussian!`, `hanning!`, etc. window functions that accepted a destination array would be pretty straightforward, then the conversion would happen when the array was assigned...

See this [similar plot](https://github.com/JuliaDSP/DSP.jl/pull/286#issuecomment-484977111) (but for longer filters) that @galenlynch made that compares regular FFT processing to overlap-save. I think that the thing to do is to add another check...

@RoyiAvital I'm confused by something on your plots - I'd expect your first plot (up to length 2000) to be the top-left corner of your next plot (up to length...

This thread is getting a little unfocused. As I understand it the central claim is that there's a regime where direct convolution is faster than FFT or overlap-save (OVS). I...

I tend to think that it's useful to organize docs into a few standard sections: - User Guide: Guides the user through high-level concepts of the library, introduces how it...

I know that convolution is used in non-DSP contexts for other purposes (like multiplying huge numbers) and in those contexts true integer convolution makes sense, and should probably be implemented...

I'll preface by saying I haven't ever used any of the analog filter design stuff so I may be missing something. Given that `Lowpass`, `Highpass`, etc. don't actually do very...