Unexpected behavior of FVREVAR: bug or intended?

[Sebastian Kripfganz]

Dear Statalisters,
Before I bother the technical support of StataCorp, I would like to hear your opinion about the following behavior of fvrevar.

When I call fvrevar once with a variable name that contains time-series operators, it creates a new temporary variable with the same values as the original. This is expected.

When I subsequently call fvrevar again with the same original variable name, it does not create a new temporary variable but instead returns the name of the previously created temporary variable. This becomes problematic when I modify the content of this temporary variable before calling fvrevar the second time, or simply when I need two different temporary variables for subsequent computations.

Here is an example:

clear all
webuse abdata

program fvrevar_test
    syntax varname(num ts)

    // varlist refers to variable L.n
    // fvrevar creates temporary variable __000000 identical to L.n
    fvrevar `varlist'
    loc newvarname "`r(varlist)'"
    sum `varlist' `newvarname'

    // content of only the temporary variable __000000 is modified
    replace `newvarname' = 0

    // varlist still refers to original unchanged variable L.n
    // fvrevar attempts again to create the temporary variable __000000 (instead of __000001)
    // "new" temporary variable has the same values as the previously modified temporary variable (bug?)
    fvrevar `varlist'
    loc newvarname "`r(varlist)'"
    sum `varlist' `newvarname'
end

fvrevar_test L.n

The output is as follows:

. fvrevar_test L.n

    Variable |        Obs        Mean    Std. Dev.       Min        Max
-------------+---------------------------------------------------------
           n |
         L1. |        891    1.083518    1.338469  -2.095571   4.687321
             |
    __000000 |        891    1.083518    1.338469  -2.095571   4.687321
(1,031 real changes made)

    Variable |        Obs        Mean    Std. Dev.       Min        Max
-------------+---------------------------------------------------------
           n |
         L1. |        891    1.083518    1.338469  -2.095571   4.687321
             |
    __000000 |      1,031           0           0          0          0

I expected that the second call of fvrevar would create a new temporary variable _000001 with the values of the original variable L.n, not again _000000, because I want to do different things with the two temporary variables.

I would call this a bug, but maybe there is a reason for this behavior that I do not see?!

[daniel klein]

Not a bug; documented behavior in [R] fvrevar on p. 838, Techincal note. Implemented for good reasons, yet inconvenient at times, as reported here.

Best
Daniel

[Sebastian Kripfganz]

Thanks Daniel, I should have read the documentation more carefully. I still find this behavior of fvrevar annoying but there are of course ways to work around this problem now that I am aware of it.