Bug 23620 – 'alias this' is not respected in static array length position

struct Index { uint value; alias value this; } enum i = Index(); int[i] a; void main() { } Error: integer constant expression expected instead of `Index(0u)`

@SixthDot updated dlang/dmd pull request #14803 "fix issue 23620 - 'alias this' is not respected in static array lengt…" fixing this issue: - fix issue 23620 - 'alias this' is not respected in static array length position does the cast to size_t before calling toInteger(), which emitted the error https://github.com/dlang/dmd/pull/14803

dlang/dmd pull request #14803 "fix issue 23620 - 'alias this' is not respected in static array lengt…" was merged into stable: - 4ed4e2c0ff1f238a0d4cdbccc21ec3d50d8f2706 by Basile Burg: fix issue 23620 - 'alias this' is not respected in static array length position does the cast to size_t before calling toInteger(), which emitted the error https://github.com/dlang/dmd/pull/14803

dlang/dmd pull request #14813 "Merge Stable into master" was merged into master: - 1fe9ffb6722c902bf0a3fcd78cfb740bcbf1bc31 by Basile Burg: fix issue 23620 - 'alias this' is not respected in static array length position does the cast to size_t before calling toInteger(), which emitted the error https://github.com/dlang/dmd/pull/14813

This isn't a bug! void main() { struct Index { size_t value; alias value this; } enum size_t i = Index(); int[i] a;  int[cast(size_t)Index()] b;  static assert(a.length == b.length); }

Why? Explicit cast shouldn't be required.

// because : enum index1 = Index(); assert( is(typeof(index1) == struct) ); auto index2 = Index(); assert( !is(typeof(index1) == size_t) && !is(typeof(index2) == size_t) ); // but: enum index = Index(); int[index + 1] d; // automatic conversion d[0] = 42; // yeah SDB@79

(In reply to Salih Dincer from comment #6) > // because : > > enum index1 = Index(); > assert( > is(typeof(index1) == struct) > ); > > auto index2 = Index(); > assert( > !is(typeof(index1) == size_t) && > !is(typeof(index2) == size_t) > ); > > // but: > > enum index = Index(); > int[index + 1] d; // automatic conversion > d[0] = 42; // yeah > > SDB@79 I still don't understand. Static array declaration expects an integer. 'Index' implicitly converts to an integer. Why should casts or other hacks be required?

I don't understand too :) This code works me: struct Index(T) { T value; this(T v) { value = v; }; alias opCall this; @property opCall() inout { return value; } } enum one= Index!int(1)(); int[one] a = 1; int[Index!int(2)()] b = 2; static assert(a.length < b.length); Best regards... SDB@79

(In reply to Salih Dincer from comment #8) > I don't understand too :) > > This code works me: > > struct Index(T) > { > T value; > > this(T v) { value = v; }; > alias opCall this; > > @property opCall() inout { > return value; > } > > } > > enum one= Index!int(1)(); > int[one] a = 1; > > int[Index!int(2)()] b = 2; > static assert(a.length < b.length); > > Best regards... > > SDB@79 You don't need to mark opCall with @property for that code to work. 'alias this' is also unnecessary. You wrote "This isn't a bug!". Did you mean this bug report is invalid? If yes, it would be nice if you explained why you think the report is invalid.

(In reply to Salih Dincer from comment #4) > This isn't a bug! > > void main() > { > struct Index > { > size_t value; > alias value this; > } > > enum size_t i = Index(); > int[i] a; >  int[cast(size_t)Index()] b; >  static assert(a.length == b.length); > } before the fix DMD did a hidden cast a cast but too late.

(In reply to Max Samukha from comment #9) > (In reply to Salih Dincer from comment #8) > > I don't understand too :) > > > > This code works me: > > > > struct Index(T) > > { > > T value; > > > > this(T v) { value = v; }; > > alias opCall this; > > > > @property opCall() inout { > > return value; > > } > > > > } > > > > enum one= Index!int(1)(); > > int[one] a = 1; > > > > int[Index!int(2)()] b = 2; > > static assert(a.length < b.length); > > > > Best regards... > > > > SDB@79 > > You don't need to mark opCall with @property for that code to work. 'alias > this' is also unnecessary. > > You wrote "This isn't a bug!". Did you mean this bug report is invalid? If > yes, it would be nice if you explained why you think the report is invalid. Here's how I think (please let's discuss privately via email or here): First you create an object and it is a struct. There is a strange working alias in the build. The related problem may be a bug for you. But is there any reason not neomorph it? Because CTFE is involved. But what if we don't want the object to occur where it occurs (in a variable without CTFE) that it points to, not itself? SDB@79