Cryptic "undefined function arg0/0" error message in EEx templates

What version of Elixir are you testing it on? Do you get the same error with multiple different versions?

Hi @axelson! I’m using Elixir 1.10.04:

Erlang/OTP 23 [erts-11.0.3] [source] [64-bit] [smp:2:2] [ds:2:2:10] [async-threads:1] [hipe]

Elixir 1.10.4 (compiled with Erlang/OTP 22)

Haven’t tested with a different version tbh.

I’ve tested with more versions and it still doesn’t compile…

• 1.10.04-otp-22
• 1.10.04-otp-23
• 1.11.03-otp-22
• 1.11.03-otp-23

I have no clue what this could be… Do you have any other suggestions?

Do you get the same error with

<%= component Slot do %>
  <p>How are you?</p>
<% end %>

in your test?

Hey, @sanswork! The error seems to happen only when I’m using the slot/2 function inside a component/2 block inside an EEx template… In other words, your example would succeed (at least at compilation time).

Should I open an issue on Phoenix’s repository perhaps?

Update: Cryptic “undefined function arg0/0” error message in EEx templates · Issue #4200 · phoenixframework/phoenix · GitHub

The AST from that issue pretty-prints to:

[arg0 = case(slot(:greet) do
  {:safe, ["    <strong>Hello!<strong>\n"]}
end) do
  {:safe, data} ->
    data
  bin when is_binary(bin) ->
    Plug.HTML.html_escape_to_iodata(bin)
  other ->
    Phoenix.HTML.Safe.to_iodata(other)
end, arg1 = case(slot(:name) do
  {:safe, ["    <span>John</span>\n"]}
end) do
  {:safe, data} ->
    data
  bin when is_binary(bin) ->
    Plug.HTML.html_escape_to_iodata(bin)
  other ->
    Phoenix.HTML.Safe.to_iodata(other)
end, {:safe, [arg0, arg1, "  <p>How are you?</p>\n"]}]

or distilled to its essence:

[a = 1, b = 2, a + b]

This fragment doesn’t compile either:

** (CompileError) iex:5: undefined function a/0
    (stdlib 3.13.2) lists.erl:1358: :lists.mapfoldl/3

The issue is that a list of AST nodes is not the same as a block.

This version of the simplified example does compile (and returns 3):

(a = 1; b = 2; a + b)

(or you can use newlines instead of ;)

The difference is clear in their respective ASTs:

iex(9)> quote do: [a = 1, b = 2, a + b] 
[
  {:=, [], [{:a, [], Elixir}, 1]},
  {:=, [], [{:b, [], Elixir}, 2]},
  {:+, [context: Elixir, import: Kernel], [{:a, [], Elixir}, {:b, [], Elixir}]}
]

iex(10)> quote do: (a = 1; b = 2; a + b)
{:__block__, [],
 [
   {:=, [], [{:a, [], Elixir}, 1]},
   {:=, [], [{:b, [], Elixir}, 2]},
   {:+, [context: Elixir, import: Kernel], [{:a, [], Elixir}, {:b, [], Elixir}]}
 ]}

Yes, I noticed something was wrong when I actually printed that AST. I couldn’t quite put my finger on it though. One thing that I noticed is that this compiles:

<%= component Slot do %>
  <% slot :greet do %>
    <strong>Hello!<strong>
  <% end %>
  <% slot :name do %>
    <span>John</span>
  <% end %>
  <p>How are you?</p>
<% end %>

And this does not:

<%= component Slot do %>
  <%= slot :greet do %>
    <strong>Hello!<strong>
  <% end %>
  <%= slot :name do %>
    <span>John</span>
  <% end %>
  <p>How are you?</p>
<% end %>

Notice the <% versus <%=, which generates a completely different AST (much more complicated one) from the previous example. I guess because it involves actually outputting something on Phoenix’s EEx Engine side.

Since I was counting on having specific elements while expanding the block, I didn’t expect that the block would be different (but obviously it is).