SQL Server - conditional aggregation with correlation

Question

Background:

The original case was very simple. Calculate running total per user from highest revenue to lowest:

CREATE TABLE t(Customer INTEGER  NOT NULL         


        

Answer 1

There is no such limitation in the SQL standard for LATERAL. CROSS APPLY is a vendor-specific extension from Microsoft (Oracle adopted it later for compatibility) and its limitations are obviously not owed to the ISO/IEC SQL standard, since the MS feature pre-dates the standard.

LATERAL in standard SQL is basically just a modifier for joins to allow lateral references in the join tree. There is no limit to the number of columns that can be referenced.

I wouldn't see a reason for the odd restriction to begin with. Maybe it's because CROSS APPLY was originally intended to allow table-valued functions, which was later extended to allow sub-SELECTs.

The Postgres manual explains LATERAL like this:

The LATERAL key word can precede a sub-SELECT FROM item. This allows the sub-SELECT to refer to columns of FROM items that appear before it in the FROM list. (Without LATERAL, each sub-SELECT is evaluated independently and so cannot cross-reference any other FROM item.)

The Postgres version of your query (without the more elegant window functions) can be simpler:

SELECT c.*
     , round(revenue        / c2.sum_total, 2) END AS percentage 
     , round(c2.running_sum / c2.sum_total, 2) END AS running_percentage 
FROM   t c, LATERAL (
   SELECT NULLIF(SUM(revenue), 0)::numeric AS sum_total  -- NULLIF, cast once
        , SUM(revenue) FILTER (WHERE revenue >= c.revenue) AS running_sum
   FROM   t
   WHERE  "User" = c."User"
   ) c2
ORDER  BY c."User", c.revenue DESC;

• Postgres 9.4+ has the more elegant aggregate FILTER for conditional aggregates.

• NULLIF makes sense, I only suggest a minor simplification.

• Cast sum_total to numeric once.

• Round result to match your desired result.