[cfe-dev] String Interpolation for ObjC

[Alex L via cfe-dev]

Hi Uli,

Clang currently accepts `\(` as a valid escape for `(` (apparently it's a
GCC extension). It doesn't warn about this unless -Wpedantic is specified.
To avoid breaking existing code that relies on this Clang might have to use
another character pair (note that '[' and '{' have the same issue), or
something like '\@(expression)'.

On 26 July 2017 at 18:43, Uli Kusterer via cfe-dev <cfe-dev at lists.llvm.org>
wrote:

> Hi,
>
>  I have an idea on how one could add string interpolation to Swift. I
> wrote a tiny preprocessor as a proof-of-concept:
> https://github.com/uliwitness/ocpre
>
>  Since that doesn't actually parse the code, it can only handle object
> types so far, so I'm thinking I should try to add it to the Clang ObjC
> compiler proper, instead. But I've never done this, never even built clang.
> Is there a tutorial somewhere on how to do this, and pointers on where I
> would start hooking into string escape sequence parsing, and which other
> parts I'd have to touch?
>

I would recommend looking at the getting started guide first:
http://llvm.org/docs/GettingStarted.html. It contains instructions that
describe how to build LLVM and Clang. As well as the getting started guide,
I would recommend looking at Clang's internals manual before working on
Clang: https://clang.llvm.org/docs/InternalsManual.html. This document
contains a section that describes how one can add an new expression to
Clang.

I hope this helps,
Alex


>
>  My approach is basically a simple code transformation. I'd turn strings
> containing escape sequences and subexpressions:
>
>         NSLog( "This is \(NSFullUserName()), who is \([myDatabase
> ageForUser: NSUserName()]) years old." );
>
> into
>
>         NSLog( "This is %1$@, who is %1$d years old.", (NSFullUserName()),
> ([myDatabase ageForUser: NSUserName()]) );
>
> and that's it. So this *should* be a comparatively simple change, and,
> moreover should need no changes to any existing libraries, client source
> code etc. It would just work with existing variadic functions.
>
> Not quite sure how one would localize the format strings (which my
> preprocessor does), as it's kinda unusual for clang to generate strings
> files, but maybe one can just generate the code for the main file and
> output a status message with the generated string key and value, and then
> one could use that message's text to generate the initial strings file so
> it will actually work at runtime. But even support for interpolation at the
> level Swift has right now is an improvement, I suppose.
>
> Cheers,
> -- Uli Kusterer
> "The Witnesses of TeachText are everywhere..."
> http://orangejuiceliberationfront.com
>
>
>
>
>
> _______________________________________________
> cfe-dev mailing list
> cfe-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20170727/f86e0312/attachment.html>