The last two weeks were dominated by upgrading to LALR(k) parser — I did very little from Skila TODO list, but something anyway. I almost finished dealing with methods inheritance – so far Skila supports the same model as in C#. I have one bug to hunt and one optimization to make (skipping creating empty virtual table). I don’t allow multiple class inheritance, however in general I would like to add it but without much expense. So this feature is on hold.
I rewrote generating classes and methods in the backend — now classes are pure structures (fields only) and methods are functions with extra argument
this passed. Of course I had to write custom dispatcher for virtual methods — it was easier than I thought, one step less for porting the backend to JS and R.
It was great opportunity to refresh my knowledge about C# — I didn’t know that interface method declaration is implemented as a method with empty body — and while doing so my patience for two C# books I have ran out. Each time I try to get definitive answer all I find is either the story from the same as me author who feels the urge to patronize me, or mumbling over 8 pages about the given concept. Polish editions of C# in Depth by Jon Skeet and Pro C# 2010 and the .NET 4 Platform by Andrew Troelsen are
for sale sold.
The incoming features are standalone functions and finding new angle for handling the data — something between C++, C# and Go. I am thinking of automatic de-/referencing the objects (this would mean no overloading based on pointer depth) with
struct defaulting to data and
class defaulting to non-null pointer to data — each default could be explicitly overridden though. We’ll see…