follow symlinks for path_relative_to_include

[Patricol]

Closes #543

My use case involves wanting to prefix terragrunt dirs with numbers to make the stage order obvious, generate state keys based on path to ensure uniqueness, and use a fair number of terraform_remote_state objects.

Reordering or inserting stages is a time-consuming process involving moving many different state files, adjusting terraform_remote_state objects, and renaming folders.

Also, cding isn't as frictionless as it could be; with the extra often 0-padded prefixes.

Creating numbered symlinks to the terragrunt dirs (but having terragrunt ignore those symlinks) would be an ideal solution.

Things to consider:

• I think path_relative_to_include should take a boolean argument if possible; defaulting to the old behavior but allowing the use of this behavior.
• Should this be done for path_relative_from_include too?
• Haven't yet tested. Especially if not implementing the bool toggle; need to see how this behaves when a relative include path is valid from one but not both of the real and symlinked directories.
• Is it alright that this runs filepath.Clean on the path too?

[Patricol]

@brikis98 @yorinasub17

Happy to work on this more if I get confirmation that the 'help wanted' tag on #543 is still accurate. (Maybe after >2 years the change is no longer desired?)

[brikis98]

@Patricol Sorry for the delay. We are super, super buried right now. The PR is very much appreciated and we'll review it as soon as we can!

[brikis98]

Thanks for the PR and your patience as we got around to reviewing it!

Things to consider:

• I think path_relative_to_include should take a boolean argument if possible; defaulting to the old behavior but allowing the use of this behavior.

Unless you know of a specific use case where the current approach would break as a result, I'd rather avoid extra flags / branches.

• Should this be done for path_relative_from_include too?

Yes, probably.

• Is it alright that this runs filepath.Clean on the path too?

Sure.

Note that for symlinks to work, we may also need to solve #1611.