On Wednesday, 12 February 2014 at 01:48:24 UTC, Manu wrote:
The other major hurdle is a proper parser usable for tooling.
Many are
trying to reinvent the wheel, and nothing short of the DMD
front-end itself
is really capable of properly parsing D code.
The biggest missing component I'm aware of is this
DMD-frontend-as-a-library idea that is always being discussed,
but never
seems to be happening. If I had to nominate a single critical
goal for the
ecosystem for 2014, that would be it.
To be precise, it's not the parsing that is the issue. Creating a
fully compliant (with many updates to the spec needed along the
way) and robust parser is not particularly hard and is happening
with stuff like SDC, and more recently, DScanner.
It is the semantic analysis that needs a *proper* engine such as
that of a compiler, in which case DMD's front-end and SDC are the
only real options, not the best-effort, good-enough crap is
currently used by IDEs.
The best we can manage with the current approach is something
close to what Visual Studio has for C++, but not quite even at
that level (because D's generic additions and CTFE mix things up
further), and even then it would be years away in terms of
development time.