Type mismatch error when comparing a string variable across rows

[Amol Singh Raswan]

Hello, I am facing trouble in Stata 14.2 when comparing a string variable across rows (i.e., observations). Consider the following toy data:

clear all
set obs 3
gen string_var = "a"
gen numeric_var = 1

If I compare the values of the variables across the three observations, I face no problem for the numeric variable but get an error for the string variable. I run (and get):

count if numeric_var == numeric_var[_n+1] == numeric_var[_n+2]
 1

count if string_var == string_var[_n+1] == string_var[_n+2]
type mismatch
r(109);

Why am I getting the type mismatch error for the string variable?

[Nick Cox]

The problem arises because evaluations of the form

a == b == c

are resolved in this way. First a == b is evaluated, as 1 if true or 0 if false; either way the result is numeric and so if c is string, the result is a type mismatch.

 
 count if (string_var == string_var[_n+1]) &  (string_var[_n+1] == string_var[_n+2]_

[Amol Singh Raswan]

Ah, so my numeric variable spuriously gave me the correct result. Thanks very much, Nick!