On 17/06/2014 3:05 a.m., H. S. Teoh via Digitalmars-d wrote:
On Mon, Jun 16, 2014 at 12:44:05PM +0000, via Digitalmars-d wrote:On Monday, 16 June 2014 at 11:49:11 UTC, Rikki Cattermole wrote:I would go that far, when combining string mixins,As far as I can tell string mixins have the same bad properties that macros have. It makes automatic translation very difficult and makes reasoning about code more difficult. It is a cheap and effective solution, but without any trace of beauty... A design blunder IMHO.Actually, IIRC, string mixins were never designed to be nice -- they started as a kind of temporary last-resort kludge that got put in, in lieu of a true AST macro system, with the view that it would meet the current metaprogramming needs until the latter, ostensibly superior, solution came along. Unfortunately, AST macros never happened, and string mixins kinda took hold in the D codebase, so that's what we have now. I would personally avoid using string mixins unless there's absolutely no other way to achieve what you want -- they're a kind of last-resort nuclear warhead that you don't bring out unless all the other guns fail to win the battle.
I have a rule about string mixins which I believe to be a good one to follow: If using string mixins, don't expose it to client code unless they explicitly want it. Basically what this means is, use something to wrap it like a mixin template but don't just say hey call this function and use it as a string mixin! Its a little nicer. Also the explicit getting of the string is important for debugging. Sometimes its needed for the more complex cases in edge cases/broken functionality. I use it during development (a simple pragma msg inside the mixin template for example).
But in saying this, some of this could be handled by opDispatch. Its just a shame that both approaches currently aren't handled for auto-completion by any IDE's. I would expect the string mixin would be one day.
Having said that, though, proper use of string mixins with CTFE and
templates ('scuse me, *compile-time arguments* ;)) can be extremely
powerful, and one of the things that make D metaprogramming so awesome.
T
>
