When I was working on the Boundary deployment for Azure using Terraform, I noticed that the AWS implementation was using the templatefile function instead of the the template_file data source. I wondered why, and when I went to the official Terraform docs for template_file, I saw the following:

In Terraform 0.12 and later, the templatefile function offers a built-in mechanism for rendering a template from a file. Use that function instead, unless you are using Terraform 0.11 or earlier.

There appear to be two primary benefits of using the function instead of the data source.

1. The function is built into the Terraform binary, so it always matches the version of Terraform you are using. There is no need to get a plug-in, unlike the template_file data source, which uses the template provider plug-in.
2. The function can deal with complex data types, not just primitives. You can pass a set, list, map, etc. to templatefile as a value for a var, and it can interpolate it. The template_file data source can only accept strings, numbers, and boolean values.

Of course with the improvements in Terraform 0.12+, the need to use a template file has been reduced. You can now do much of what you did with template files directly in the configuration with stuff like for_each loops, for loops, and direct interpolation. The primary reason to use a template file is to improve the readability of the Terraform configuration file, and perhaps make reusable components. That's especially true when you're creating a cloud-init or custom_data file to be used with a VM deployment.

The rendering speed of templatefile and direct interpolation is markedly faster than using the data source. Terraform doesn't have to invoke a plugin or refresh the state of a data source. Even in this simple example, the difference is noticeable. If you are using a lot of templates, you could significantly reduce processing time by removing the template_file data source from all of your configurations.