Obsoleted FTS trick

[obartunov]

Many years I recommended to use subselects to avoid extra ts_headline() evaluation:

select ts_headline(body,to_tsquery('supernovae & x-ray')), rank from (
select body, ts_rank(fts,to_tsquery('supernovae & x-ray')) as rank
from apod
where fts @@ to_tsquery('supernovae & x-ray')
order by rank desc limit 5
) as foo;

But modern version of Postgres 9.6 is wise enough (commit 9118d03a8cca3d97327c56bf89a72e328e454e63: When appropriate, postpone SELECT output expressions till after ORDER BY) and now one could write much simple sql:

select ts_headline(body,to_tsquery('supernovae & x-ray')), ts_rank(fts,to_tsquery('supernovae & x-ray')) as rank
from apod
where fts @@ to_tsquery('supernovae & x-ray')
order by rank desc limit 5

I wrote simple tsheadline wrapper and learned that ts_headline() in the last sql query was called exactly 5 times.

=================================================================================
Now, it'd be great if postgres handles even simpler version of sql:

select ts_headline(body,q),ts_rank(fts,q) as rank
from apod, to_tsquery('supernovae & x-ray') q
where fts @@ q
order by rank desc limit 5;

UPDATE:PG13 allows use such a query with immutable form of to_tsquery (always specify FTS configuration name).

Unfortunately, this form of query could be very slow, since postgres will use join (nested loop), so, don't use this!

Limit (actual time=0.457..1.153 rows=5 loops=1)
-> Result (actual time=0.456..1.150 rows=5 loops=1)
-> Sort (actual time=0.284..0.285 rows=5 loops=1)
Sort Key: (ts_rank(apod.fts, q.q)) DESC
Sort Method: top-N heapsort Memory: 32kB
-> Nested Loop (actual time=0.085..0.267 rows=36 loops=1)
-> Function Scan on to_tsquery q (actual time=0.022..0.022 rows=1 loops=1)
-> Bitmap Heap Scan on apod (actual time=0.045..0.074 rows=36 loops=1)
Recheck Cond: (fts @@ q.q)
Heap Blocks: exact=34
-> Bitmap Index Scan on gin_apod_fts_idx (actual time=0.037..0.037 rows=36 loops=1)
Index Cond: (fts @@ q.q)
Planning time: 0.079 ms
Execution time: 1.182 ms
(14 rows)