Numbering sessions within events

[Laura Lee]

I have been working on this problem for a while, searching statalist and consulting chatgpt without success. I have a large dataset of events (with numeric IDs), some of which have sessions (stored as string variables). I need a way to generate session numbers for each distinct session *within each event*. In the example dataset below, I manually input the variable I would like to generate as "des_output". If there are no sessions, a session number should not be generated.

* Example generated by -dataex-. For more info, type help dataex
clear
input float event str6 session float des_output
25 "four"   1
25 "one"    2
25 "one"    2
25 "one"    2
25 "two"    3
40 "purple" 1
40 "purple" 1
40 "violet" 2
72 "smell"  1
72 "smell"  1
72 "smell"  1
72 "taste"  2
99 ""       .
99 ""       .
end

This command below

quietly by event session: gen sess_num = cond(_N==1,1,_n)

creates numbering *within sessions* for each combination of event number and string session name.
Other commands I tried counted sessions across the entire dataset. Does anyone know how to generate session numbers within events as shown in the variable des_output?

I greatly appreciate any guidance you can offer.

[Clyde Schechter]

by event (session), sort: gen wanted = sum(session != session[_n-1])
replace wanted = . if missing(session)

[Laura Lee]

Thank you, Clyde. This works perfectly!