New FUNCTION documentation, and a minor test case modification.

Docs/sp-imp-spec.txt:
  Updated docs about stored FUNCTIONs.
Docs/sp-implemented.txt:
  Updated docs about stored FUNCTIONs.
mysql-test/r/sp.result:
  Changed the ifac test example into a combination of a procedure and a function.
mysql-test/t/sp.test:
  Changed the ifac test example into a combination of a procedure and a function.

1 files changed, 19 insertions, 6 deletions

diff --git a/Docs/sp-implemented.txt b/Docs/sp-implemented.txt
index b3b12b7edb8..5173606e95c 100644
--- a/Docs/sp-implemented.txt
+++ b/Docs/sp-implemented.txt
@@ -3,7 +3,6 @@ Stored Procedures implemented 2003-02-02:
 
 Summary of Not Yet Implemented:
 
-  - FUNCTIONs
   - Routine characteristics
   - External languages
   - Access control
@@ -26,14 +25,14 @@ Summary of what's implemented:
 
 List of what's implemented:
 
-  - CREATE PROCEDURE name ( args ) body
+  - CREATE PROCEDURE|FUNCTION name ( args ) body
     No routine characteristics yet.
 
-  - ALTER PROCEDURE name ...
+  - ALTER PROCEDURE|FUNCTION name ...
     Is parsed, but a no-op (as there are no characteristics implemented yet).
     CASCADE/RESTRICT is not implemented (and CASCADE probably will not be).
 
-  - DROP PROCEDURE name
+  - DROP PROCEDURE|FUNCTION name
     CASCADE/RESTRICT is not implemented (and CASCADE probably will not be).
 
   - CALL name (args)
@@ -45,7 +44,7 @@ List of what's implemented:
 	  will either generate an error message, or it might even work to
 	  call all procedures from the top-level.
 
-  - Procedure body:
+  - Function/Procedure body:
     - BEGIN/END
       Is parsed, but not the real thing with (optional) transaction
       control, it only serves as block syntax for multiple statements (and
@@ -82,9 +81,23 @@ Closed questions:
   Answer: Same tables, with an additional key-field for the type.
 
 
-Open questions:
+Open questions/issues:
 
   - SQL-99 variables and parameters are typed. For the present we don't do
     any type checking, since this is the way MySQL works. I still don't know
     if we should keep it this way, or implement type checking. Possibly we
     should have optional, uset-settable, type checking.
+
+  - FUNCTIONs do not work correctly in all circumstances yet.
+    For instance a function like:
+      create function s() returns int
+      begin
+        declare s int;
+        select sum(test.t.y) into s from test.t;
+        return s;
+      end
+    do not work. Invoking this in queries like "SELECT * FROM t2 WHERE x = s()"
+    will make things very complicated. And, in fact, even "SET @s=s()" and
+    "SELECT s()" fail, although the exact reasons in these cases are a bit
+    obscure; part of the problem might be the way the lex structure is
+    bit-copied (a not completely sound thing to do).