ShowEach

This package provides the @showeach macro which assists with poor man's debugging. In addition to @showing the annotated expression, a tree of every subexpression is printed (unlike @show, printing literals is omitted). Thus, @showeach helps to understand what is going on in powerful one liners that occur frequently in Julia.

Usage

Almost any valid Julia expression can be annotated with @showeach. For example,

julia> begin
          x = 1
          y = 2
          z = 3
          g(x) = 2x
          f(x, y) = x*y
       end

julia> @showeach z*f(g(x), y)
┌─ z = 3
│     ┌─ x = 1
│  ┌─ g(x) = 2
│  ├─ y = 2
├─ f(g(x),y) = 4
z * f(g(x),y) = 12

julia> @showeach a*f(rand(z), g(y))[x]
┌─ a = [1,2,3]
│        ┌─ z = 3
│     ┌─ rand(z) = [0.516954,0.854899,0.636338]
│     │  ┌─ y = 2
│     ├─ g(y) = 4
│  ┌─ f(rand(z),g(y)) = [2.06781,3.4196,2.54535]
│  ├─ x = 1
├─ (f(rand(z),g(y)))[x] = 2.06781
a * (f(rand(z),g(y)))[x] = [2.06781,4.13563,6.20344]

julia> @showeach sum(i for i = x:z if isodd(i))
┌─ x = 1
├─ z = 3
│  ┌─ i = 1
├─ isodd(i) = true
│  ┌─ i = 2
├─ isodd(i) = false
│  ┌─ i = 3
├─ isodd(i) = true
├─ i = 1
├─ i = 3
sum((i for i = x:z if isodd(i))) = 4

Annotating a function definition prints all subexpressions whithin the function's body every time the function is used, e.g.,

julia> @showeach h(x,y) = x*rand(y)

julia> @showeach z*h(x, y)
┌─ z = 3
│  ┌─ x = 1
│  ├─ y = 2
│  ├─ in function h(x,y):
│  │     ┌─ x = 1
│  │     │  ┌─ y = 2
│  │     ├─ rand(y) = [0.4288,0.781806]
│  │     x * rand(y) = [0.4288,0.781806]
├─ h(x,y) = [0.4288,0.781806]
z * h(x,y) = [1.2864,2.34542]

The output is suppressed after n expressions have been printed at the same level of a tree:

julia> @showeach f(x) = 2x

julia> @showeach map(f, 1:100)
┌─ f = f (generic function with 1 method)
├─ in function f(x):
│     ┌─ x = 1
│     2x = 2
├─ in function f(x):
│     ┌─ x = 2
│     2x = 4
├─ in function f(x):
│     ┌─ x = 3
│     2x = 6
├─ in function f(x):
│     ┌─ x = 4
│     2x = 8
├─ in function f(x):
│     ┌─ x = 5
│     2x = 10
├─  ⋮
map(f,1:100) = [2,4,6  …  196,198,200]

However, this only works within a @showeach expression. Calling map(f, 1:100) without @showeach will result in a long output. Unfortunately, due to a Julia issue, @showeach map(i->2i, 1:100) does not work yet. Instead, use a named function as above.

The maximum number of expressions can be set with set_loop_limit(n-1):

julia> @showeach calc_pi(M) = 4 * sum(rand()^2 + rand()^2 < 1 for i = 1:M) / M

julia> set_loop_limit(3)

julia> calc_pi(10^5)
in function calc_pi(M):
         ┌─ M = 100000
         │        ┌─ rand() = 0.841332
         │     ┌─ rand() ^ 2 = 0.707839
         │     │  ┌─ rand() = 0.164777
         │     ├─ rand() ^ 2 = 0.0271516
         │  ┌─ rand() ^ 2 + rand() ^ 2 = 0.734991
         ├─ rand() ^ 2 + rand() ^ 2 < 1 = true
         │        ┌─ rand() = 0.977022
         │     ┌─ rand() ^ 2 = 0.954572
         │     │  ┌─ rand() = 0.70994
         │     ├─ rand() ^ 2 = 0.504014
         │  ┌─ rand() ^ 2 + rand() ^ 2 = 1.45859
         ├─ rand() ^ 2 + rand() ^ 2 < 1 = false
         │        ┌─ rand() = 0.622204
         │     ┌─ rand() ^ 2 = 0.387138
         │     │  ┌─ rand() = 0.349908
         │     ├─ rand() ^ 2 = 0.122436
         │  ┌─ rand() ^ 2 + rand() ^ 2 = 0.509574
         ├─ rand() ^ 2 + rand() ^ 2 < 1 = true
         ├─  ⋮
      ┌─ sum((rand() ^ 2 + rand() ^ 2 < 1 for i = 1:M)) = 78498
   ┌─ 4 * sum((rand() ^ 2 + rand() ^ 2 < 1 for i = 1:M)) = 313992
   ├─ M = 100000
   (4 * sum((rand() ^ 2 + rand() ^ 2 < 1 for i = 1:M))) / M = 3.13992

Caveats

• @showeach does not produce a correct output with anonymous functions, let blocks, and other macros. Nevertheless, the return values of such expressions should be correct.
• If an expression throws an error, the indentation is not reset automatically and one has to call reset_indent() manually.
• The loop_limit suppresses any output even outside loops. This will be improved in the future.

Development

This package is in an early stage of development and things can break any time. If you find a bug or simply have a question, just open an issue.