Bug 24323 – x[0 .. 2] cast to Ty[2], x[$-2 .. $] does not?

void fn(ref const ubyte[2]); void test(ubyte[] data) { fn(data[0 .. 2]); // <- call successfully fn(data[$-2 .. $]); // <- fails: not callable using argument types `ubyte[]` // this also fails: size_t i = data.length - 2; fn(data[i .. i+2]); // this works: fn(data[$-2 .. $][0 .. 2]); } I feel like the language should be able to determine that `$ - ($-2) == 2`. The workaround is pretty dumb. I can't possibly be the first one to report this, what's the reason? Just hard to implement the bit of algebra required? As long as there is just one variable term in a slice expression, and the rest of the terms are all known at compile time, this equation can be easily solved. I sure hope that $ symbol is only ever evaluated once inside a slice expression, even in the presence of a custom opDollar?

Actually, the number of variable terms is irrelevant; the equation just needs to be simplified and the variable terms will naturally be cancelled out.

> I can't possibly be the first one to report this, what's the reason? Just hard to implement the bit of algebra required? You're right: https://github.com/dlang/dmd/pull/4209 The difficulty is determining how this would fit with the rest of constant folding. *** This issue has been marked as a duplicate of issue 13700 ***

Difficult or otherwise, it's a pretty lame hole, and it's been that way for decades. It should be done.