Could someone explain more why this should be a language feature?

My understanding of template strings is they are like f-strings but don't do the interpolation bit. The name binding is there but the values are not formatted into the string yet. So effectively this provides a "hook" into the stringification of the interpolated values, right?

If so, this seems like a very narrow feature to bake into the language... Personally, I haven't had issues with introducing some abstraction like functions or custom types to do custom interpolation.

It reminds me of javascripts template literals (https://developer.mozilla.org/en-US/docs/Web/JavaScript/Refe...) and .nets FormattableString (https://learn.microsoft.com/en-us/dotnet/api/system.formatta...)

The best use case I know of for these kinds of things is as a way to prevent sql injection. SQL injection is a really annoying attack because the "obvious" way to insert dynamic data into your queries is exactly the wrong way. With a template string you can present a nice API for your sql library where you just pass it "a string" but it can decompose that string into query and arguments for proper parameterization itself without the caller having to think about it.

You can see some motivations for it further down the document. Basically it allows libraries to access the ease of use of f-strings for more than just formatting a string according to what the language allows. Structured logging is one area where I would like to use this.

It's useful for doing things like writing parameterized SQL queries that avoid injection attacks. Or anything that quacks like that proverbial duck, where you are going to emit some plaintext and need to validate it as it is filled.

Because python dominates AI, and python is dominating because of AI. And prompting really, really benefits from f-strings and templated strings. LLMs as a whole means the rise of unstructured data, and flexible string manipulation is really important for handling that.