gegen max does not properly evaluate string expressions

[adamreir]

I just updated gtools to version 1.8.1, and some code that used to work now issues an error message.

The command

gegen newvar=(stringvar=="value"), by(group)

Now gives the error message

warning: gegen is NOT parsing the expression '(stringvar==value")' by group." invalid name

Here is a minimal example that produces the error:

clear
set seed 1
set obs 10 
g cat=round(runiform()*5)
g stringvar = "1"
replace stringvar="2" if runiform()>.7
gegen newvar = max(stringvar=="2"), by(cat)

I've managed to circumvent the problem by creating a temporary variable and evaluating that in egen max. But if it indicates some more significant issue I thought you should know.

(and since I'm here: thanks for making this awesome package)

Version info

• OS: Windows 10 and Red Hat Linux version 7.9
• Version: 1.8.1 27Aug2021

[mcaceresb]

@adamreir I put that warning there to make it plain this was not parsed in the same way as by varlist: egen (i.e. it's parsed by creating the temporary variable first). However, I forgot about this Stata quirk with strings, which has gotten me many times over the years. You can replace lines 646 and 647 in egen.ado with

                mata printf("{bf:warning}: gegen is {bf:NOT} parsing the expression '%s' by group.\n", st_local("args"))
                mata printf("To parse this expression by group, call gegen using the -by:- prefix.\n")

or delete them if you understand the warning.

[adamreir]

Oh, right. I didn't consider the possibility the bug was in the error message and not in the parsing. That does fix the issue.

Thanx again!