@whale-av No, the math I wrote is the math I meant. It's just an expression to access the previous sample.
@solipp Yes, this example is functionally pointless because I've removed everything that wasn't necessary to generate that message. In the original I'm reading from and writing to a 1 sample table to reduce an audio feedback delay to 1 sample.
so I think I've met the requirement. Plus it seems to work, though admittedly my test is weak. And when I think about it, if fexpr~ allows access to the previous sample, then even with a 64 sample vector it would have to access the last sample of the previous vector while processing the first sample of the current vector. It's the same as when the vector size is 1. So I read the restriction on the index values to mean "Sorry, I'm not gonna buffer an arbitrarily large number of past vectors for you. Be happy with the one I'm giving you."