ok    1  filter(scores, keep): the values whose mask is 1
ok    2  a comparison mask: filter(scores, scores >= 60)
ok    3  nothing selected: an empty 1D array
ok    4  a SHORT mask (2 vs 5) raises ARRAYSHAPE by default
ok    4.5  partial: true -- a short mask selects nothing beyond its end
ok    5  a LONG mask (8 vs 5) raises ARRAYSHAPE by default
ok    5.5  partial: true -- extra mask elements are ignored
ok    5.6  partial: allow? (a variable) with equal counts is the plain result
ok    6  a 2D values array: storage order, result 1D
ok    7  a 2D mask of a different shape: positions pair in storage order
ok    8  stats$mean(filter(...)) -- a function argument
ok    9  size(filter(...))
ok    10  print filter(scores, scores >= 60)
ok    11  filter nests: filter(filter(...), ...)
ok    12  an array expression as the values (scores * 2), the result piped into a routine
ok    13  filter(...) |> total_of(): the 1D result by reference
ok    14  filter(...) inside a larger array expression
ok    15  two filter results of the same length add element-wise
ok    16  string values: who$ = filter(names$, keep)
ok    17  string values with a comparison mask (the mask is a hidden REAL temp)
ok    18  string filter over reshaped values and mask (storage order)
ok    19  a string filter result piped into a routine
ok    20  scalar-context filter(values, mask) raises catchable ARRAYRESULT (the atom registration)

passed:  23   failed:  0 
