ptrEnd for arrays

[monarchdodra]

Just a tiny something I've found to be missing: Getting the (one-past-the-) end pointer of an array (dynamic or static).

This is a tiny convenience/helper function, but is (IMO) quite useful.

No more writing arr.ptr + arr.length. Now, it's just arr.ptrEnd

Discussion here: http://forum.dlang.org/thread/kcfqyxbnlhxdkqibfpiu@forum.dlang.org

[alexrp]

Please add a big fat warning (with $(RED ...) saying that when the only reference to a GC-managed array is a one-past-the-end pointer, the array can be collected.

[monarchdodra]

Please add a big fat warning (with $(RED ...) saying that when the only reference to a GC-managed array is a one-past-the-end pointer, the array can be collected.

Is the message good? Prose has never been my strong suit.

[alexrp]

$(RED Warning: The pointer returned by this function will not be considered a live reference to the array by the garbage collector.) should be fine.

[monarchdodra]

Thanks

[andralex]

So we define what's supposed to be a convenience function, which comes with a big fat red warning. Doesn't sound like a great abstraction. I think we should just close this.

[alexrp]

@andralex This is pretty typical when doing pointer arithmetic. It just so happens that the garbage collector isn't very fond of this. So long as you keep a normal/interior reference to the array around on the stack, in registers, or in the heap, things will be fine.

[jmdavis]

Given that pointer arithmetic is supposed to be a bit of a rarity, I question that adding a helper function for this is justifiable, especially when you take the warning into account.

[alexrp]

It's a helper that can eliminate at least one source of human error when dealing with pointers.

I think the warning is completely irrelevant as to whether this should be merged or not. D is a garbage collected language at the same time as being a systems language, and this is just a result of that. In practice, I think people will very rarely iterate an array via pointer arithmetic without keeping a proper reference to it around.

[jmdavis]

But how often is pointer arithmetic even used in D? This just seems fairly pointless to me, cluttering the library with an extremely simple function for a rare use case.

[andralex]

The warning has relevance because it underlines the intended use and audience of the function. Even assuming the GC was good enough to mind pointers past end, I'd still be very convinced the function is not appropriate for inclusion into Phobos.

I will close this now. Thanks again for the contribution @monarchdodra - let's focus on the more fertile ones. Feel free to reopen if new insights come forward.

[monarchdodra]

I will close this now. Thanks again for the contribution @monarchdodra - let's focus on the more fertile ones. Feel free to reopen if new insights come forward.

No problem. Kind of wish I had closed it myself, but no problem.