Re: SQL-standard function body

On Mon, Mar 15, 2021 at 04:03:44PM +0800, Julien Rouhaud wrote:
> On Mon, Mar 15, 2021 at 01:05:11AM -0500, Jaime Casanova wrote:
> > I found another problem when using CASE expressions:
> >
> > CREATE OR REPLACE FUNCTION foo_case()
> > RETURNS boolean
> > LANGUAGE SQL
> > BEGIN ATOMIC
> > select case when random() > 0.5 then true else false end;
> > END;
> >
> > apparently the END in the CASE expression is interpreted as the END of
> > the function
>
> I think that it's an issue in psql scanner. If you escape the semicolon or
> force a single query execution (say with psql -c), it works as expected.

Applying the following diff (not sending a patch to avoid breaking the cfbot)
the issue and doesn't seem to break anything else:

diff --git a/src/fe_utils/psqlscan.l b/src/fe_utils/psqlscan.l
index a492a32416..58026fe90a 100644
--- a/src/fe_utils/psqlscan.l
+++ b/src/fe_utils/psqlscan.l
@@ -871,7 +871,9 @@ other .

{identifier} {
cur_state->identifier_count++;
- if (pg_strcasecmp(yytext, "begin") == 0)
+ if ((pg_strcasecmp(yytext, "begin") == 0)
+ || (pg_strcasecmp(yytext, "case") == 0)
+ )
{
if (cur_state->identifier_count > 1)
cur_state->begin_depth++;