step = linspace(5,0.125);
ZZZZZZZZZ_intermediates = start + cumsum([0 step]);
ZZZZZZZZZ_intermediates = ZZZZZZZZZ_intermediates(ZZZZZZZZZ_intermediates<=stop);
ZZZZZZZZZ_intermediates = ZZZZZZZZZ_intermediates(ZZZZZZZZZ_intermediates>=stop);
for loop_variable = ZZZZZZZZZ_intermediates
You might notice that the loop ended before 29.875 . You asked for the increments to be linspace(5,0.125) but remember by default linspace() is a vector of 100 values -- so the first increment would be 5, the second would be about 4.9508, the third is about 4.9501, and so on. Those quickly add up to more than the stop point.
Your requirement to be able to specify an expression for the increments, leads to the possibility that your increments might have a mix of positive and negative numbers. A decision had to be made about what to do for the case where the steps increment past the endpoint but that steps with the opposite sign then take the value back to below the endpoint. Clearly values beyond the end point should not be included... but should you stop as soon as you pass the endpoint the first time, or should you allow for the value to come back?
I decided that since it should be permitted to allow a mix of increments while you are in range, that it would make the most sense to allow values to return back to range. So what I do is filter out the values that go beyond the stop-point -- leaving any that come back to the range. So with positive increments 5 for a while, and negative increments (say) 7, a sequence with start 1 and stop 20 and increments [5 5 5 5 -7 -7] might go [1 6 11 16 (OMITTED 21) 14 7] leaving [1 6 11 16 14 7] as the loop control values.
It might also be reasonable to argue that the sequence should never emit values "before" the start after having gone past it -- so 1:[5 5 5 5 -7 -7 -7]:20 with the logic posted above would give [1 6 11 16 14 7 0] but it would be reasonable to filter on start as well as stop giving loop values [1 6 11 16 14 7] ...
