Bug: nested functions do not get unnested into pipes

[emkguts]

Here are two test cases:

Test 1:

assert_style("c(b(a), d)", "a |> b() |> c(d)")

Output (while messing with a local copy):

  1) test pipifying pipifying (Styler.Style.PipesTest)
     test/style/pipes_test.exs:945
     Styling produced unexpected results
     expected:
       a |> b() |> c(d)
     styled:
       c(b(a), d)

     code: assert_style("c(b(a), d)", "a |> b() |> c(d)")
     stacktrace:
       test/style/pipes_test.exs:948: (test)

Test 2:

assert_style("c(b(a |> e), d)", "a |> e() |> b() |> c(d)")

Output:

  1) test pipifying pipifying (Styler.Style.PipesTest)
     test/style/pipes_test.exs:945
     Styling produced unexpected results
     expected:
       a |> e() |> b() |> c(d)
     styled:
       c(a |> e() |> b(), d)

     code: assert_style("c(b(a |> e), d)", "a |> e() |> b() |> c(d)")
     stacktrace:
       test/style/pipes_test.exs:948: (test)

[novaugust]

the first one's not a bug, styler intentionally doesn't pipe nested function calls - it just "unrolls" nested pipes.

the second one, on the other hand ... yup!

[emkguts]

the first one's not a bug, styler intentionally doesn't pipe nested function calls - it just "unrolls" nested pipes.

the second one, on the other hand ... yup!

Ah gotcha. Thanks for the clarification. And thanks for building such a dope tool! I've been playing with it, and...mad respect!

[novaugust]

it's crossed my mind to do but it'd get pretty ugly with DSLs. i figure if someone's writing elixir without pipes, they have a reason (hopefully that reason isn't just "what's a pipe?" ¯\_(ツ)_/¯)

the second one... is that a real thing in your codebase or are ya'll just coming up with edge cases 😂 either way i'll try to get to it at some point =)

and...mad respect

🤘 ❤

[emkguts]

Touche on rewriting no-pipe Elixir. As for the second point, we are about to style some large, old, gnarly code bases so was pressure testing a bit, but fair to call it a contrived edge-case 😂

[novaugust]

nah i get it, that's where styler came from too =) at least that'll get unpiped all the way with successive runs of mix format right? so shouldn't block converting a codebase