Push down quals more efficiently

[davidfetter]

There's got to be something libgit2 could do that's faster than this:

# EXPLAIN ANALYZE SELECT commit_date FROM master.master_repository WHERE sha1 = '7fee252f6fbf78ca5e50ee591573d59f98e75d37';
                                                          QUERY PLAN                                                          
══════════════════════════════════════════════════════════════════════════════════════════════════════════════════════════════
 Foreign Scan on master_repository  (cost=0.00..46668.15 rows=46092 width=8) (actual time=487.647..271489.852 rows=1 loops=1)
   Filter: (sha1 = '7fee252f6fbf78ca5e50ee591573d59f98e75d37'::text)
   Rows Removed by Filter: 46090
   Foreign Git Repository: /home/shackle/pggit/postgresql
   Foreign Git Branch: refs/heads/master
   Foreign Git Search Path: 
 Planning Time: 1522.944 ms
 Execution Time: 271498.346 ms
(8 rows)

Time: 273022.134 ms (04:33.022)

It's from the PostgreSQL repository.

[franckverrot]

This is bad. I'll take a look, thanks for reporting again!

[franckverrot]

Btw, I started working on that but being unfamiliar with PG's internals I couldn't get it done by the end of last year, and then moved on other things.

[davidfetter]

I suspect it's libgit2's APIs that are in play here. For example, there are ways to do range queries in it that just need to be surfaced to the FDW API.