Get latest cumulative sales amount for various evaluation dates in SAS - sql

Question

I have a list of evaluation dates stored in a table, datelist. It's technically two columns, start_date and end_date, for each evaluation period. The end_date will definitely need to be used, but the start_date may not. I only care about periods that are completed, so, for example, the period from 2016-01-01 to 2016-07-01 is in progress but not complete. So, it's not in the table.
start_date end_date
2012-01-01 2012-07-01
2012-07-01 2013-01-01
2013-01-01 2013-07-01
2013-07-01 2014-01-01
2014-01-01 2014-07-01
2014-07-01 2015-01-01
2015-01-01 2015-07-01
2015-07-01 2016-01-01
I have a separate table that lists cumulative sales by customer, sales_table with three columns, customer_ID, cumul_sales, transaction_date. For example, let's say customer 4793 bought $100 worth of stuff on 2/14/2014 and $200 worth of stuff on 3/30/2014 and $75 on 7/27/2014, the table will have the following rows:
customer_ID cumul_sales transaction_date
4793 100 2014-02-14
4793 300 2014-03-30
4793 375 2014-07-27
Now, for each evaluation date and for each customer, I want to know what's the cumulative sales as of the evaluation date for that customer? If a customer hadn't purchased anything by an evaluation date, then I wouldn't want a row for that customer at all corresponding to said evaluation date. This would be stored in a new table, called sales_by_eval, with columns customer_ID, cumul_sales, eval_date. For the example customer above, I'd have the following rows:
customer_ID cumul_sales eval_date
4793 300 2014-07-01
4793 375 2015-01-01
4793 375 2015-07-01
4793 375 2016-01-01
I can do this, but I'm looking to do it in an efficient way so I don't have to read through the data once for each evaluation date. If there are a lot of rows in the sales_table and 40 evaluation dates, that would be a large waste to read through the data 40 times, once for each evaluation date. Would it be possible with only one read through the data, for example?
The basic idea of the current process is a macro loop that loops once per evaluation period. Each loop has a data step that creates a new table (one table per loop) to check each transaction to see if it has occurred before or on the end_date of that corresponding evaluation period. That is, each table has all the transactions that occur before or on that evaluation date but none of the ones that occur after. Then, a later data step uses "last." to get only the last transaction for each customer before that evaluation date. And, finally, all the various tables created are put back together in another data step where they are all listed in the SET statement.
This is in SAS, so anything SAS can do, including SQL and macros, is fine with me.

Answer 1

In SAS, when you use group by statement, you can still use not grouping variables in select statement, like this:
proc sql;
create table sales_by_eval as
select s.customer_ID, s.cumul_sales, d.end_date as eval_date
from datelist d
join sales_table s
on d.end_date > s.transaction_date
group by s.customer_ID, d.end_date
having max(s.transaction_date) = s.transaction_date
;
quit;
This mean that for each combination of selected variablem SAS will return rekord with measures summarized within defined group. To limit the result to the last state of transaction value, use having condition, where you select only those records that have transaction_date equal to max(transaction_date) within s.customer_ID, d.end_date group.