Trolling a bit: Should this invariant always hold?

assert(equal(range.retro.array, range.array.retro));

Can you find a definition of range, using only std.algorithm verbs (ie. no custom ranges), that violates it?

(And no peeking at my recently-filed bug reports!)

On Monday, 17 October 2022 at 11:02:07 UTC, FeepingCreature wrote:

No, it shouldn't. The first expression might consume range. So it might be empty in the second expression.

On Monday, 17 October 2022 at 11:46:35 UTC, ag0aep6g wrote:

Ah, good idea!

assert(equal(range.save.retro.array, range.save.array.retro));

On Monday, 17 October 2022 at 11:02:07 UTC, FeepingCreature wrote:

Posting this without having looked at other replies.

No. The range could be a reference range, in which case this does not hold unless the range is already empty.

The invariant would have to be assert(equal(range.save.retro.array, range.save.array.retro)); instead.

Also this won't hold if the range contains floating point NaNs, or other elements that are not equal to themselves.

There may be other corner cases too I didn't think of.

On Monday, 17 October 2022 at 11:02:07 UTC, FeepingCreature wrote:

Solution, or rather, Phobos code that violates it (even the strengthened assert(equal(range.save.retro.array, range.save.array.retro))):

rot13: nhgb enatr = bayl(1, 2).havd!"gehr"; // havd onpxjneq vgrengvba purpxf ryrzragf sbe havdhrarff va gur onpxjneq beqre bs gur haqreylvat enatr.