rename directories.

wns/WN-0004/wn-0004.md → wns/WN-0004-sql-dimensions/wn-0004.md

@@ -10,17 +10,36 @@ LookML models are entirely based on this mechanism.  As much as possible, we sho
 
 SQL expressions are always dimensions.
 
-Let's assume a table orders with nested/repeated items.
+Assume a table with the following structure
+
+orders:
+```
+order_id
+order_time
+shipped_time
+status
+user_id
+items []
+  product_id
+  product_description
+  sale_price
+  is_returned
+  product_category
+```
 
 ## Coorelated Subquery Example
 Malloy currently does not support coorelated subqueries.  Coorelated subqueries on nested data are more efficient than joining nested data.
 
-The code below returns the count of items that have not been returned.
+The code below returns the count of items that have not been returned.  
+
+`${TABLE}` references the root alias for the source.  
+
+The `!number` lets us know that the return type from sql expression is a number.  
 
 ```
 source: orders is bigquery.table(...) extend {
 
-  dimension: item_count is sql("""
+  dimension: item_count is sql!number("""
     (
       SELECT count(*) from UNNEST(${TABLE}.items) 
       WHERE not is_returned

wns/WN-0007/wn-0007.md → wns/WN-0007-nested-computations/wn-0007.md

@@ -215,4 +215,29 @@ There are several intrinic functions that return nested sources.  These function
     group_by: word is description_words.value
     aggregate: order_count is count()
   }
+```
+
+## Re-nesting a nested Source
+If a nested computation doesn't have a 'return' block, it is assumed to be a calculation that returns multiple rows, and is of type 'nested source'  the result of which is treated as a ....
+
+
+```
+  run: orders -> {
+    select:
+      order_id
+      socks is items->{
+        project: *
+        where: product_catetory='SOCKS'
+      }
+  }
+
+```
+
+The SQL for this would be
+
+```
+  SELECT 
+    order_id,
+    ARRAY((SELECT * from UNNEST(items) WHERE product_category='SOCKS)) as socks
+  FROM 'orders.parquet'
 ```