Sebastian Krantz

Thanks @traversc for this note. Your decision is very respectable, although I would urge to keep `qs` on CRAN as long as possible. Many packages have not, or cannot, move...

@traversc perhaps one more note here, the [CRAN policy](https://cran.r-project.org/web/packages/policies.html) suggests that only major x.y.0 updates may be forced to fix all issues. So it should be possible to keep `.qs`...

Thanks @andrewGhazi, I'll look into this as I find time! In general though if you want speedy standard deviations, do use the option `stable.algo = FALSE` to `fsd()`. You can...

Thanks, I just observe the bug doesn't occur with `fsum()`, which you should be using if you care about performance with *collapse*.

@r-barnes do you want to remain the CRAN maintainer? Otherwise I'd be willing to take that over (assuming that @edzer is not interested?) to handle smaller releases like the present...

I'd be surprised if adding a zero breaks any code - except for this particular issue which is clearly not meant to be as far as R is concerned. But...

Adding zero also works and is faster, e.g.: https://github.com/SebKrantz/collapse/compare/master...zero_dups However, I haven't implemented this either yet, as the speed penalty is noticeable... But yes, we should probably do this.

I wonder if there is a way to do this using bitwise operations – that would probably be the cleanest and fastest way.

No idea, but it works on clang (Mac).

Thanks. In general `set_collapse(mask = "all")` is risky because it replaces basic functions like `max()` with `fmax()`, which is S3 generic. For example `max(matrix)` gives a number, but `fmax(matrix)` gives...