merge strL - needs a better error message

[Doug Hemken]

This code

clear
sysuse auto
save test

rename make makeshort
generate strL make = makeshort

merge 1:1 make using test

produces this error message:

. merge 1:1 make using test
variable make does not uniquely identify observations in the master data
r(459);

The problem is the datatype, not uniqueness!

[Clyde Schechter]

Yes, and I can reproduce this error on my installation, Windows 7 64-bit, Stata v. 15.1 MP-2.

Doug, what is your setup? We need a bunch of people on different setups to identify if this is a pan-Stata issue or is specific to particular versions/platforms/flavors.

[Rich Goldstein]

#1 is not clear - when I do exactly what Doug shows I get the same message; however, test.dta is the data set in memory - so I agree that the message is not the best but the problem appears to be different

note that if I

sort make
save test, replace
sysuse auto
. merge 1:1 make using test
key variable make is strL in using data.
    The key variables -- the variables on which observations are matched -- can be str#, but
    they cannot be strLs.

then I get the message shown above

I am on a Mac with High Sierra using Stata 15.1, MP-4

[Doug Hemken]

Windows 7, Stata 15.1, SE-64
and
Windows server 2008, Stata 15.0, SE-64
and
Windows server 2008, Stata 15.0, MP-16 64bit
and
Linux version 2.6.32-642.15.1.el6.x86_64 (Red Hat 4.4.7-17) , Stata 15.0, MP-16

[Doug Hemken]

Rich, I found that too. It would be useful to have the same error message when strL is in master.