Brad Roberts wrote:
On 1/11/2010 11:20 PM, grauzone wrote:
g wrote:
Hi
Is there a way to demangle .mangleof strings at CTFE
or at least know the fully qualified name of a class or template
instance.
If you have to use such dirty tricks, you probably should consider to
turn back NOW for your own good. D always lures you into doing tricky
template and CTFE stuff, and then you end up either in compiler bugs or
other dead ends.
And how do bugs get fixed? Step one is finding them.. which typically involves
writing code that uses the features. Secondarily, a good number of bugs have
been fixed over the last couple months, so re-exploring the area might well bear
good fruit.
Anyway, doom and gloom pronouncements like this aren't particularly helpful.
Some types of bugs just never seem to disappear: when one specific bug
got fixed, a regression occurs and you have a similar bug. (This
happened with forward references in the current dmd release.)
Nothing wrong with a warning.
That said, demangling a symbol and using that inside compile time expressions
does sound like one good definition of hell, regardless of how well it works.
Exactly. The thing is just (and that I wanted to say in my previous
post): you get easily fascinated by the possibilities, but then either
the language or dmd hit an unexpected barrier and fail. Then you start
hacks by throwing heaps of CTFE and mixins on the problem, or stuff like
parsing .mangleof. I believe choosing a simpler solution instead
(although it's boring in terms of hacking) is better.
Later,
Brad
