Re: Scariest patch tournament, PostgreSQL 11 edition

On 2018-Jun-25, Jeff Janes wrote:

> Is there a summary of the results of the previous rounds? I didn't see any
> announcements of them. I've been trying to find some crash recovery
> torture testing to do for v11 release, but can't find features to focus on
> which might be scariest from a WAL perspective.

I processed the answers today after many delays. Here's [one way of
looking at] the results:

commit │ title │ committer │ points
─────────────┼────────────────────────────────────────────────────────────────┼──────────────────┼────────
499be013de6 │ Support partition pruning at execution time │ Alvaro Herrera │ 6
8b08f7d4820 │ Local partitioned indexes │ Alvaro Herrera │ 6
cc415a56d09 │ Basic planner and executor integration for JIT. │ Andres Freund │ 5
8561e4840c8 │ Transaction control in PL procedures │ Peter Eisentraut │ 5
86f575948c7 │ Allow FOR EACH ROW triggers on partitioned tables │ Alvaro Herrera │ 4
e2f1eb0ee30 │ Implement partition-wise grouping/aggregation. │ Robert Haas │ 4
3de241dba86 │ Foreign keys on partitioned tables │ Alvaro Herrera │ 4
2f178441044 │ Allow UPDATE to move rows between partitions. │ Robert Haas │ 4
f49842d1ee3 │ Basic partition-wise join functionality. │ Robert Haas │ 4
16828d5c027 │ Fast ALTER TABLE ADD COLUMN with a non-NULL default │ Andrew Dunstan │ 4
9fdb675fc5d │ Faster partition pruning │ Alvaro Herrera │ 4
1aba8e651ac │ Add hash partitioning. │ Robert Haas │ 4
6f6b99d1335 │ Allow a partitioned table to have a default partition. │ Robert Haas │ 4
eb7ed3f3063 │ Allow UNIQUE indexes on partitioned tables │ Alvaro Herrera │ 4
b96d550eb03 │ Support for optimizing and emitting code in LLVM JIT provider. │ Andres Freund │ 3
8224de4f42c │ Indexes with INCLUDE columns and their support in B-tree │ Teodor Sigaev │ 3
4b0d28de06b │ Remove secondary checkpoint │ Simon Riggs │ 3
9da0cc35284 │ Support parallel btree index builds. │ Robert Haas │ 2
1804284042e │ Add parallel-aware hash joins. │ Andres Freund │ 1

It's fair to say that anything in the list is scary regardless of score,
because:

1) Some respondents were vague and provided answers such as "Partition
improvements", which I attributed to several commits. In three cases
(all involving JIT) I added less commits than I probably should have,
but then there are so many of those commits that I'm not sure which are
relevant.

2) The mechanism to award points to each commit is quite unfair, but it
didn't seem important enough. We have too few answer anyhow :-(

Thanks to all respondents!

--
Álvaro Herrera https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services