RANK OVER Function - Proc SQL

[Adir92]

Hello to all, hope you are doing well.
I have a database TEST that has 5 columns. Here is an example:

ID_Customer	ID_Project	ID_Application	Nbre_Application	Creation_Application
200200	1000	267	1	20JANV2020
200200	1000	256	1	10JANV2020
200200	1000	456	1	15JANV2020
400400	2000	120	2	10FEB2020
400400	2000	101	1	05JANV2020

For each client, I would like to classify his applications, from the oldest to the most recent with the number 1 for the oldest, 2 for the most recent...

Like this : 

ID_Customer	ID_Project	ID_Application	Nbre_Application	Creation_Application
200200	1000	256	1	1
200200	1000	456	1	2
200200	1000	267	1	3
400400	2000	101	1	1
400400	2000	120	2	2

I know there's a proc rank procedure but here how can I use it ? And is there a way with proc sql ? 

Thank you !

[PaigeMiller]

proc rank data=have out=want;
    by id_customer;
    var creation_application;
    ranks ranked_creation_application;
run;
proc sort data=want;
     by id_customer ranked_creation_application;
run;

Or alternatively

proc sort data=have;
    by id_customer creation_application;
run;
data want;
    set have;
    by id_customer;
    if first.id_customer then ranked_creation_application=0;
    ranked_creation_application+1;
run;

Really no need for SQL here.

[Adir92]

Thank you !

[ballardw · Accepted solution]

That would not be a "rank" , that would be a sequence number and not ideal for Proc SQL.

The question might be what happens if a single ID_customer value has multiple Id_project values and how that gets treated. The example below resets the sequence for each project. If that is not desired use First.id_customer

Proc sort data=test;
   by Id_customer id_project creation_application;
run;

data want;
   set test;
   by id_customer id_project;
   retain sequence;
   if first.id_project then sequence=1;
   else sequence+1;
run;

The BY statement creates boolean (true/false or 1/0) values for each variable that on the statement indicating whether the specific record is the first or last of a value group. These are accessed using the First. (see the dot) and Last. notation preceding the variable name.

Retain creates a variable whose value is kept across iteration boundaries of a data step so can be used to do sequences. But you need to reset it to a starting value when needed.

Bad juju results from changing the values of variables like Creation_application from a date to numeric, or was that just an accident in your "want"?

[Adir92]

Yes I see, no It was an accident, its better with a created variable. Thank you so much ! I will read in detail your explanation to understand it well but the important thing is that it works!