Strings and things

[Brian Goetz]

OK, so I guess there is a third interpretation of your comment that I 
didn't realize the first time, which is about compile-time _validation_ 
of the format string, so the user can get errors at compile time.  (It 
really seemed like you were going for more control over lexing.)

Yes, there is a viable path to this, though we surely won't pave it all 
in one go.  As we outlined in 
https://www.youtube.com/watch?v=iSEjlLFCS3E, we have explored 
opportunistic compile-time constant folding.  This is complicated for a 
lot of reasons, but not impossible.  If all the pieces could be made to 
line up, then it is theoretically possible that we could get 
compile-time validation of the format string.  Its a long string to 
pull, but its not a dead end like library control of lexing.

On 9/17/2021 8:28 AM, Stephen Colebourne wrote:
> On Thu, 16 Sept 2021 at 19:10, Brian Goetz <brian.goetz at oracle.com> wrote:
>>> I believe however that there is a missing element - pre-processing at
>>> compile-time.
>> The other interpretation is that you are asking for a way to let the
>> library plug into the parser (the lexer, actually!) so that you're not
>> stuck with the language's notion of "string with holes."
> I'm open to any approach that can produce errors/exceptions before
> runtime when the template is invalid (preferably at compile-time). The
> discussion in the doc about safety feels a bit incomplete without
> early validation. Anyway, thanks for the reply.
>
> Stephen