Macros r() deleted when dropping a variable in Stata 15.1

[Volker Ludwig]

Dear Listers,

just stumbled over this:
When dropping a varaible in Stata 15.1, previously returned macros r() are deleted.
So, if I do the following code, Stata 15 displays a missing after the last line, whereas Stata 14 displays the mean value.

set obs 100
ge bla=runiform()
su bla
di r(mean)
drop bla
di r(mean)

Don't know if that's wanted, but might cause problems with older do-files (and ados).

Volker

[Clyde Schechter]

This is because, unlike earlier versions of Stata, current Stata returns r(k_drop), the number of variables dropped,when the -drop- command is run. In this regard, it is like any other command: if it returns something in r(), previous contents of r() go out of scope.

Older do-files and ados might indeed break over this, but not if they are run under version control. Try running the exact same code, preceded by -version 15.0- (or any older version) and you will see that the value of r(mean) remains available.

[Volker Ludwig]

Thanks Clyde, that does make sense.

[Nick Cox]

If you want to avoid invoking version control every time, try something like

program mydrop
    version 15
    drop `0'
end

for any suitable value of 15.