nonsensitive Function

nonsensitive takes a sensitive value and returns a copy of that value with the sensitive marking removed, thereby exposing the sensitive value.

Normally OpenTF tracks when you use expressions to derive a new value from a value that is marked as sensitive, so that the result can also be marked as sensitive.

However, you may wish to write expressions that derive non-sensitive results from sensitive values. For example, if you know based on details of your particular system and its threat model that a SHA256 hash of a particular sensitive value is safe to include clearly in OpenTF output, you could use the nonsensitive function to indicate that, overriding OpenTF's normal conservative behavior:

output "sensitive_example_hash" {
  value = nonsensitive(sha256(var.sensitive_example))
}

Another example might be if the original value is only partially sensitive and you've written expressions to separate the sensitive and non-sensitive parts:

variable "mixed_content_json" {
  description = "A JSON string containing a mixture of sensitive and non-sensitive values."
  type        = string
  sensitive   = true
}

locals {
  # mixed_content is derived from var.mixed_content_json, so it
  # is also considered to be sensitive.
  mixed_content = jsondecode(var.mixed_content_json)

  # password_from_json is derived from mixed_content, so it's
  # also considered to be sensitive.
  password_from_json = local.mixed_content["password"]

  # username_from_json would normally be considered to be
  # sensitive too, but system-specific knowledge tells us
  # that the username is a non-sensitive fragment of the
  # original document, and so we can override OpenTF's
  # determination.
  username_from_json = nonsensitive(local.mixed_content["username"])
}

When you use this function, it's your responsibility to ensure that the expression passed as its argument will remove all sensitive content from the sensitive value it depends on. By passing a value to nonsensitive you are declaring to OpenTF that you have done all that is necessary to ensure that the resulting value has no sensitive content, even though it was derived from sensitive content. If a sensitive value appears in OpenTF's output due to an inappropriate call to nonsensitive in your module, that's a bug in your module and not a bug in OpenTF itself. Use this function sparingly and only with due care.

nonsensitive will return an error if you pass a value that isn't marked as sensitive, because such a call would be redundant and potentially confusing or misleading to a future maintainer of your module. Use nonsensitive only after careful consideration and with definite intent.

Consider including a comment adjacent to your call to explain to future maintainers what makes the usage safe and thus what invariants they must take care to preserve under future modifications.

Examples​

The following examples are from opentf console when running in the context of the example above with variable "mixed_content_json" and the local value mixed_content, with a valid JSON string assigned to var.mixed_content_json.

> var.mixed_content_json
(sensitive value)
> local.mixed_content
(sensitive value)
> local.mixed_content["password"]
(sensitive value)
> nonsensitive(local.mixed_content["username"])
"zqb"
> nonsensitive("clear")

Error: Invalid function argument

Invalid value for "value" parameter: the given value is not sensitive, so this
call is redundant.

Note though that it's always your responsibility to use nonsensitive only when it's safe to do so. If you use nonsensitive with content that ought to be considered sensitive then that content will be disclosed:

> nonsensitive(var.mixed_content_json)
<<EOT
{
  "username": "zqb",
  "password": "p4ssw0rd"
}
EOT
> nonsensitive(local.mixed_content)
{
  "password" = "p4ssw0rd"
  "username" = "zqb"
}
> nonsensitive(local.mixed_content["password"])
"p4ssw0rd"