diff --git a/doc/sql.extensions/README.aggregate_filter.md b/doc/sql.extensions/README.aggregate_filter.md
new file mode 100644
index 0000000000..2f33fe3356
--- /dev/null
+++ b/doc/sql.extensions/README.aggregate_filter.md
@@ -0,0 +1,27 @@
+# FILTER clause for aggregate functions
+
+The `FILTER` clause for aggregate function is a shortcut for when an aggregate function is used with some condition (decode, case, iif) to disconsider some values in the aggregation.
+
+The clause could be used with all aggregate functions in aggregate or windowed (`OVER`) calls, but cannot be used with window-only functions like `DENSE_RANK`.
+
+For example, if in the same query there is a need to count the number of status = 'A' and the number of status = 'E' as different columns, the old way to do it is:
+
+```sql
+select count(decode(status, 'A', 1)) status_a,
+       count(decode(status, 'E', 1)) status_e
+  from data;
+```
+
+With `FILTER` that conditions could be more explicit:
+
+```sql
+select count(*) filter (where status = 'A') status_a,
+       count(*) filter (where status = 'E') status_e
+  from data;
+```
+
+## Syntax
+
+```
+aggregate_function [FILTER (WHERE <condition>) [OVER (<window>)]]
+```