On Thu, Apr 22, 2021 at 09:01:45PM +0000, Dukc via Digitalmars-d wrote:
> On Wednesday, 21 April 2021 at 07:50:35 UTC, rempas wrote:
> > I'm really wondering, what stops us about using mixins in run-time? I'm really interested on learning how they work. For my understating, the write the code for use in compile-time and then the whole things is compiled. However, I'm really wondering what stops us for writing the code and compiling it at run-time like it's an interpreter. Anyone that can explain how things work to me?
> 
> Having runtime mixins in the default language would the downside that the D runtime would have to include a D interpreter. No thanks.
> 
> Note though, I said in the default language. Having a separate D interpreter that would be called as an external program would be different thing altogether. To some extent, you can already do that by feeding your runtime scripts to rdmd.

I've written programs that invoke `dmd` to compile fragments of D code. It does work, and dmd is fast enough that the pause is often not noticeable (unless you're doing it in a tight inner loop).  The fragments are compiled into shared objects and dynamically loaded and executed by the program.  Ostensibly, one could employ a similar mechanism to dynamically compile parts of a program at runtime and integrate these bits of code into itself.

This does introduce a dependency on the availability of a D toolchain in the user's environment, though.  This may or may not be desirable, depending on what you're trying to do.  Plus, there may be licensing issues if you decide to distribute a dmd toolchain along with your program.  But it can probably be done, should you wish to do so.

Having said that, though, I'm also unsure about the wisdom of supporting this as a built-in feature in D. That would require shipping a runtime interpreter or a complete D toolchain in every program, which seems like overkill for a rather narrow use case.


T

-- 
Старый друг лучше новых двух.