Use ChatGPT to help understand complex code

[#1]

Ever open a piece of complex code someone else wrote and think:

Like you’re really in for it?

Ever look at your own old code and think exactly the same thing?

If you said yes — to either (or both) — you’re in good company. Hackers gotta hack, and sometimes we could all use a little help making sense of what’s in front of us.

That’s where ChatGPT comes in again — as your coding partner.

Suppose you inherit a block of SAS code that looks something like this:

*-------------------------------------------------------------------------------------*
|    	   						  Collapse Data 									  | 
|	                    	Produce State-Level Estimates          	          		  |
*-------------------------------------------------------------------------------------*;

********************************************************  By State ;
proc sql;
	create 	table hhs.covid_labor_supply as 
	select	distinct state_fip, state_name, 
            year(yearquarter) as Year format 9.,
			
/*******************************************************************  Labor Force Status | All  */
			sum( ( unemp=1 ) * WTFINL ) 											/ sum( ( in_LF=1 ) *   	WTFINL )									as UE_Women				label="Unemployment Rate"	format percent9.1 		,
			sum( ( in_LF=1 ) * WTFINL ) 											/ sum(  				WTFINL )									as LFP_Women			label="LFP Rate"			format percent9.1 		,


/*******************************************************************  Labor Force Status | By Education  */

			/*******************************************************  Unemployment */
			sum( ( educ_ltd="High School Diploma" ) * ( unemp=1 ) * WTFINL ) 		/ sum( ( educ_ltd="High School Diploma" ) * ( in_LF=1 ) * WTFINL )	as UE_Women_HS			label="EDUC <= HS" 		format percent9.1 		,
			sum( ( educ_ltd="Some College" ) * ( unemp=1 ) * WTFINL ) 				/ sum( ( educ_ltd="Some College" ) * ( in_LF=1 ) * WTFINL ) 		as UE_Women_SCollege	label="Some College"	format percent9.1 		,
			sum( ( educ_ltd="College +" ) * ( unemp=1 ) * WTFINL ) 					/ sum( ( educ_ltd="College +" ) * ( in_LF=1 ) * WTFINL ) 			as UE_Women_CollegeP	label="College +" 		format percent9.1 		,

			/*******************************************************  LFP */
			sum( ( educ_ltd="High School Diploma" ) * ( in_LF=1 ) * WTFINL ) 		/ sum( ( educ_ltd="High School Diploma" ) * WTFINL ) 				as LFP_Women_HS			label="EDUC <= HS" 		format percent9.1 		,
			sum( ( educ_ltd="Some College" ) * ( in_LF=1 ) * WTFINL ) 				/ sum( ( educ_ltd="Some College" ) * WTFINL ) 						as LFP_Women_SCollege	label="Some College" 	format percent9.1 		,
			sum( ( educ_ltd="College +" ) * ( in_LF=1 ) * WTFINL ) 					/ sum( ( educ_ltd="College +" ) * WTFINL ) 							as LFP_Women_CollegeP	label="College +" 		format percent9.1 		,


/*******************************************************************  Labor Force Status | By Child Status  */

			/*******************************************************  Unemployment */
			sum( ( child_status="No Children" ) * ( unemp=1 ) * WTFINL ) 			/ sum( ( child_status="No Children" ) * ( in_LF=1 ) * WTFINL ) 		as UE_Women_NoKids		label="No Children" 	format percent9.1 		,
			sum( ( child_status="Older Children" ) * ( unemp=1 ) * WTFINL ) 		/ sum( ( child_status="Older Children" ) * ( in_LF=1 ) * WTFINL ) 	as UE_Women_OlderKids	label="Older Children" 	format percent9.1 		,
			sum( ( child_status="Child < 5" ) * ( unemp=1 ) * WTFINL ) 				/ sum( ( child_status="Child < 5" ) * ( in_LF=1 ) * WTFINL ) 		as UE_Women_YoungKids	label="Young Children"	format percent9.1 		,

			/*******************************************************  LFP */
			sum( ( child_status="No Children" ) * ( in_LF=1 ) * WTFINL ) 			/ sum( ( child_status="No Children" ) * WTFINL ) 					as LFP_Women_NoKids		label="No Children" 	format percent9.1 		,
			sum( ( child_status="Older Children" ) * ( in_LF=1 ) * WTFINL ) 		/ sum( ( child_status="Older Children" ) * WTFINL ) 				as LFP_Women_OlderKids	label="Older Children" 	format percent9.1 		,
			sum( ( child_status="Child < 5" ) * ( in_LF=1 ) * WTFINL ) 				/ sum( ( child_status="Child < 5" ) * WTFINL ) 						as LFP_Women_YoungKids	label="Young Children"	format percent9.1 		


	from 	hhs.hhs_otj_raw
	group	by 1,2,3 
	order	by 1,2,3 ;
quit;

A ChatGPT prompt:

The response:

Ha: the “Kid level idea”!

I love how ChatGPT took that scary-looking SQL code and made it instantly more approachable with a jellybean analogy. 

And if jellybeans aren’t your thing – or you’d rather think like an 8-year-old instead of a 10-year-old – you’re just one prompt refinement away from a different example and a fresh perspective.

Once again, this shows how dynamic your conversation can be with your research partner in crime: ChatGPT.

(Editor’s note: Crime is bad. Don’t use ChatGPT for that.)

The “explain this code for me” trick is powerful, so I’ll just keep this post short.  But use it – seriously – use it.

And I’ll see you in the final post for this series: Use ChatGPT to document your code today. Trust me, this will make future you happy.