CAST(CASE WHEN pc.contype IN ('f','u','p') THEN generate_series(1, array_upper(pc.conkey, 1)) ELSE NULL END AS int) AS position;
Above query is failing with "set-valued function called in context that cannot
accept a set". But if I remove the CASE from the query then it working just good.
Like:
SELECT *
FROM pg_constraint pc,
CAST(generate_series(1, array_upper(pc.conkey, 1)) AS int) AS position;
This started failing with 69f4b9c85f168ae006929eec44fc44d569e846b9. It seems
check_srf_call_placement() sets the hasTargetSRFs flag and but when the SRFs
at the rtable ofcourse this flag doesn't get set. It seems like missing something
their, but I might be completely wrong as not quire aware of this area.
I'm a bit surprised that your query actually works...and without delving into source code its hard to explain why it should/shouldn't or whether the recent SRF work was intended to impact it.
In any case the more idiomatic way of writing your query these days (since 9.4 came out) is:
SELECT *
FROM pg_constraint pc
LEFT JOIN LATERAL generate_series(1, case when contype in ('f','p','u') then array_upper(pc.conkey, 1) else 0 end) gs ON true;
generate_series is smart enough to return an empty set (instead of erroring out) when provided with (1,0) as arguments.