Bug 8231 – conv.to fails to convert string with trailing "u"

import std.conv; void main() { uint x = 3u; // ok uint y = to!uint("3u"); // fail } std.conv.ConvException@D:\DMD\dmd2\windows\bin\..\..\src\phobos\std\conv.d(1597): Unexpected 'u' when converting from type string to type uint If it can be used in D syntax it should be convertible by conv.to.

Commit pushed to master at https://github.com/D-Programming-Language/phobos https://github.com/D-Programming-Language/phobos/commit/9f5c180b667faa512dac2e01dff19490688c07ac Disable test for bug 8231 temporarily.

This is quite easily fixed with a little utility function that checks if the remainder after parsing is a floating point or integer suffix. However, it is potentially a nasty subtle breaking change as it's quite a common pattern to rely on functions like "to" to check for unusual input, which this would change the definition of. It also introduces a difference in attitude between "to" and "parse", which currently work together very neatly. On those grounds, I'm against it. To be acceptable it would have to be enabled with a template argument (an enum with names RAW and D_LITERAL_AWARE ?), or be in a function with a different name.

Both converting regular numbers and converting D literals to their values are important needs and use cases, but mixing them transparently is probably a not so good idea. So better to give it a different function name or to add some compile-time argument to tell apart the two cases, as Comment #2 says. D number literals also allow underscores: 1_000_000u

I agree that this is a bad idea. std.conv.to is intended for converting user data between different types, not for parsing D language constructs like uint literals. We should have a different function for that (maybe something from std.d.lexer, once we have that).

Yeah the suggestion was bad, if anything it might subtly allow buggy input.