for those who don't know: it's standard (required?)...

for those who don't know: it's standard (required?) for C++'s std::vector<bool> to be specialized to actually bitpack the booleans, getting you 8x the booleans per allocated byte (nice!)

...at the cost of any API that returns T* or the like now being literally impossible to satisfy, silently deleting random APIs from the std::vector interface (oh no!)

aiui this makes code that generically works with vectors now randomly incompatible with that particular substitution (unless you very carefully avoid those APIs)

Like 15 July at 16:49 | Open on toot.cat

4 comments

Nicole

@Gankra I don't believe it's required to bitpack, but the APIs are required not to return a pointer so there's no reason not to.

15 July at 16:58 | Open on mstdn.social

Nicole

@Gankra fwiw, my opinion is "it's overhyped how terrible vector<bool> is, but it's certainly not good"

15 July at 17:01 | Open on mstdn.social

Gankra

@streganil i think that's where i fall too, yeah. vector<bool> is relatively rare, so it won't come up much, but it's unfortunate that a desire to optimize the concrete case of vector<bool> made the whole vector type just a little less useful in generic contexts

15 July at 17:10 | Open on toot.cat

Ed Page

@Gankra @streganil

> a desire to optimize the concrete case

And it wasn't even created as an optimization but for pedagogical purposes.

> The vector<bool> specialization was intentionally put into the standard to provide an example of how to write a proxied container.

http://www.gotw.ca/publications/mill09.htm

16 July at 1:50 | Open on hachyderm.io