digitalmars.D.learn - Error: field r must be initialized in constructor, because it is
- =?UTF-8?B?QWxpIMOHZWhyZWxp?= (28/28) Jun 03 2015 Pretty standard thing doesn't work and I can't find it on bugzilla:
- anonymous (18/36) Jun 03 2015 You can shut the compiler up by doing `this.r = R.init;` in the
Pretty standard thing doesn't work and I can't find it on bugzilla: import std.algorithm; struct Foo(R) { R r; this(R r) // <-- Compilation error {} } auto foo(R)(R r) { return Foo!R(r); } void main() { auto arr = [1]; arr.map!(i => i).foo; } I was actually trying to dispatch the construction to a separate function: this(R r) // <-- Still error { init(r); } void init(R r) { this.r = r; } (I remember this latter issue having been discussed recently.) Ali
Jun 03 2015
On Wednesday, 3 June 2015 at 23:27:31 UTC, Ali Çehreli wrote:Pretty standard thing doesn't work and I can't find it on bugzilla: import std.algorithm; struct Foo(R) { R r; this(R r) // <-- Compilation error {} } auto foo(R)(R r) { return Foo!R(r); } void main() { auto arr = [1]; arr.map!(i => i).foo; }You can shut the compiler up by doing `this.r = R.init;` in the constructor. Searching for "must be initialized" yields issue 13945 which calls for better documentation: https://issues.dlang.org/show_bug.cgi?id=13945 Not covered there is why `map!(i => i)` results in a nested struct. This is the compiler being conservative/stupid, I guess. The lambda could require a context pointer. It doesn't, but the compiler isn't smart enough to see that. Some variants that don't trip the compiler up: arr.map!((int i) => i).foo; static auto identity(T)(T x) {return x;} arr.map!identity.foo; arr.map!"a" It would be nice if the compiler would take this hint, but it doesn't: arr.map!(function (i) => i).foo;
Jun 03 2015