Split of "" should be empty. Length of empty array should be 0

[bobtfish]

In place we want to do: count = "${length(split(",", var.list_of_things))}" and in some cases have an empty list (i.e. ""), and therefore get 0 resources.

To do this, splitting the string "" needs to return an empty array (rather than an array with one item which is empty), and the length function needs to do the right thing if it sees an empty array.

[radeksimko]

Hi Tom,
I was not testing your code thoroughly, but isn't this a duplicate of #2157 ?

[bobtfish]

@radeksimko - no, or at least not entirely if you add just my test changes to your branch, then the tests fail, like so:

--- FAIL: TestInterpolateFuncLength (0.00s)
interpolate_funcs_test.go:585: 5: bad output for input: ${length(split(",", ""))}

    Output: "1"
    Expected: "0"
--- FAIL: TestInterpolateFuncSplit (0.00s)
interpolate_funcs_test.go:585: 1: bad output for input: ${split(",", "")}

    Output: "B780FFEC-B661-4EB8-9236-A01737AD98B6B780FFEC-B661-4EB8-9236-A01737AD98B6"
    Expected: "B780FFEC-B661-4EB8-9236-A01737AD98B6"
FAIL
FAIL    github.com/hashicorp/terraform/config   0.040s

[phinze]

I went back and forth on this as I worked on the StringList refactor.

https://twitter.com/phinze/status/614116280480739328

I think this sums it up nicely:

it's more consistent to return [""], but more useful in most situations to return [].

So I figured we'd toss in a compact or without func to allow us to implement it it the consistent way, but I'd willing to be persuaded in the other direction.

Seems like a tough trade off between "one more interp func" and POLA. Thoughts?

[thegedge]

Although quite kludgy, I think one way to do this at the moment is to use a template_file that appends a comma to a string containing a comma-separated list, and do ${length(split(",", replace("/^,$/", "", template_file.foo.rendered))) - 1}.

My five cents: I think having split returning a unitary list for empty strings is expected for anyone coming from any (all?) other language, but it feels like it should have the alternative behaviour since we can't pass around lists "natively" (i.e., resorting to join/split). I'd lean towards having another interpolation function for now because it would lead to nicer configs, but would consider working towards better support for lists as variables/outputs which I think would alleviate the problems that led to this PR.

[apparentlymart]

Now that #3239 is merged, and thus we have the compact function, does that address the original need here?

In other words, it should now be possible to write the original example as count = "${length(compact(split(",", var.list_of_things)))}" and get 0 when the list_of_things is the empty string.

[bobtfish]

Sounds great to me.

I've added a test explicitly testing the combination of all 3 functions, as I didn't see that currently.

[phinze]

A PR that's iterated into a test addition. 💓 Thanks @bobtfish!

[elblivion]

🎉

[ghost]

I'm going to lock this issue because it has been closed for 30 days ⏳. This helps our maintainers find and focus on the active issues.

If you have found a problem that seems similar to this, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.