971af6935c
We remove the left joins with inner joins and also
remove the coalesce on posts.DeleteAt which improves
the join plan from a nested loop to a hash join,
and also allows the DeleteAt to be used as part of the
index filter.
Old plan
```
Aggregate (cost=666288.49..666288.50 rows=1 width=8) (actual time=5114.328..5119.991 rows=1 loops=1)
-> Gather (cost=2880.69..666288.27 rows=87 width=8) (actual time=614.214..5119.575 rows=533 loops=1)
Workers Planned: 2
Workers Launched: 2
-> Nested Loop (cost=1880.69..665279.57 rows=36 width=8) (actual time=607.821..5099.078 rows=178 loops=3)
-> Hash Join (cost=1880.28..664932.93 rows=229 width=16) (actual time=224.374..1494.698 rows=1866781 loops=3)
Hash Cond: ((posts.channelid)::text = (channels.id)::text)
-> Parallel Seq Scan on posts (cost=0.00..662980.29 rows=27568 width=43) (actual time=201.053..1013.362 rows=4410819 loops=3)
Filter: (COALESCE(deleteat, '0'::bigint) = 0)
-> Hash (cost=1866.60..1866.60 rows=1094 width=27) (actual time=1.486..1.486 rows=1002 loops=3)
Buckets: 2048 Batches: 1 Memory Usage: 74kB
-> Index Scan using idx_channels_team_id_type on channels (cost=0.42..1866.60 rows=1094 width=27) (actual time=0.028..1.331 rows=1002 loops=3)
Index Cond: ((teamid)::text = '8ywxyw9ocp8smxrmjzrkqhrdwe'::text)
-> Index Scan using threadmemberships_pkey on threadmemberships (cost=0.41..1.50 rows=1 width=35) (actual time=0.002..0.002 rows=0 loops=5600344)
Index Cond: (((postid)::text = (posts.rootid)::text) AND ((userid)::text = 'tc3p1yqw67d8idcp3g98awexqe'::text))
Filter: (following AND (posts.createat > lastviewed))
Rows Removed by Filter: 0
Planning Time: 2.005 ms
JIT:
Functions: 59
Options: Inlining true, Optimization true, Expressions true, Deforming true
Timing: Generation 9.905 ms, Inlining 122.509 ms, Optimization 293.764 ms, Emission 186.064 ms, Total 612.241 ms
Execution Time: 5125.739 ms
```
New plan
```
Aggregate (cost=683129.17..683129.18 rows=1 width=8) (actual time=1650.746..1656.455 rows=1 loops=1)
-> Gather (cost=3766.65..683085.55 rows=17450 width=27) (actual time=641.073..1656.086 rows=533 loops=1)
Workers Planned: 2
Workers Launched: 2
-> Hash Join (cost=2766.65..680340.55 rows=7271 width=27) (actual time=617.096..1635.203 rows=178 loops=3)
Hash Cond: ((posts.rootid)::text = (threadmemberships.postid)::text)
Join Filter: (posts.createat > threadmemberships.lastviewed)
Rows Removed by Join Filter: 6
-> Hash Join (cost=1880.28..679334.08 rows=45742 width=16) (actual time=15.196..1242.832 rows=1866781 loops=3)
Hash Cond: ((posts.channelid)::text = (channels.id)::text)
-> Parallel Seq Scan on posts (cost=0.00..662980.29 rows=5513543 width=43) (actual time=0.028..775.118 rows=4410819 loops=3)
Filter: (deleteat = 0)
-> Hash (cost=1866.60..1866.60 rows=1094 width=27) (actual time=1.572..1.573 rows=1002 loops=3)
Buckets: 2048 Batches: 1 Memory Usage: 74kB
-> Index Scan using idx_channels_team_id_type on channels (cost=0.42..1866.60 rows=1094 width=27) (actual time=0.025..1.372 rows=1002 loops=3)
Index Cond: ((teamid)::text = '8ywxyw9ocp8smxrmjzrkqhrdwe'::text)
-> Hash (cost=823.41..823.41 rows=5037 width=35) (actual time=256.240..256.241 rows=5007 loops=3)
Buckets: 8192 Batches: 1 Memory Usage: 417kB
-> Bitmap Heap Scan on threadmemberships (cost=159.45..823.41 rows=5037 width=35) (actual time=253.011..255.255 rows=5007 loops=3)
Recheck Cond: ((userid)::text = 'tc3p1yqw67d8idcp3g98awexqe'::text)
Filter: following
Heap Blocks: exact=601
-> Bitmap Index Scan on idx_thread_memberships_user_id (cost=0.00..158.19 rows=5037 width=0) (actual time=0.627..0.627 rows=5037 loops=3)
Index Cond: ((userid)::text = 'tc3p1yqw67d8idcp3g98awexqe'::text)
Planning Time: 1.952 ms
JIT:
Functions: 83
Options: Inlining true, Optimization true, Expressions true, Deforming true
Timing: Generation 10.148 ms, Inlining 115.819 ms, Optimization 395.923 ms, Emission 244.130 ms, Total 766.021 ms
Execution Time: 1661.791 ms
```
No special test has been done on MySQL since these are pretty
basic optimizations and should apply to both equally.
https://community-daily.mattermost.com/boards/workspace/zyoahc9uapdn3xdptac6jb69ic/285b80a3-257d-41f6-8cf4-ed80ca9d92e5/495cdb4d-c13a-4992-8eb9-80cfee2819a4?c=c9aa17f3-b918-4a78-86aa-01562e418634
```release-note
NONE
```
Mattermost is an open source platform for secure collaboration across the entire software development lifecycle. This repo is the primary source for core development on the Mattermost platform; it's written in Go and React and runs as a single Linux binary with MySQL or PostgreSQL. A new compiled version is released under an MIT license every month on the 16th.
Useful Resources:
- Deployment Guide - Install, setup, and configure your own Mattermost instance.
- Admin and User Documentation - Learn how to run a Mattermost instance and take advantage of all the features.
- Developer Documentation - Contribute code to Mattermost or build an integration via APIs, Webhooks, slash commands, Apps, and plugins.
Table of Contents
- Try out Mattermost
- Install Mattermost
- Native Mobile and Desktop Apps
- Get Security Bulletins
- Get Involved
- Learn More
- Get the Latest News
- Contributing
Try out Mattermost
- Join the Mattermost Contributor's server to join community discussions about contributions, development and more
- Get started with Mattermost Cloud to try out Mattermost
Note: Heroku preview does not include email or persistent storage
Install Mattermost
- Quick Install Guide - Deploy in minutes via Mattermost Omnibus on Ubuntu.
- Run Mattermost via Docker - Launch a Mattermost server instantly to test functionality and build integrations.
- Developer Machine Setup - Follow this guide if you want to write code for Mattermost.
Other Install Guides:
Native Mobile and Desktop Apps
In addition to the web interface, you can also download Mattermost clients for Android, iOS, Windows PC, Mac OSX, and Linux.
Get Security Bulletins
Receive notifications of critical security updates. The sophistication of online attackers is perpetually increasing. If you are deploying Mattermost it is highly recommended you subscribe to the Mattermost Security Bulletin mailing list for updates on critical security releases.
Get Involved
- Contribute to Mattermost
- Find "Help Wanted" projects
- Join Developer Discussion on a Mattermost Server for contributors
- Get Help With Mattermost
Learn More
- API Options - webhooks, slash commands, drivers and web service
- See who's using Mattermost
- Browse over 700 Mattermost integrations
License
See the LICENSE file for license rights and limitations.
Get the Latest News
- Twitter - Follow Mattermost.
- Blog - Get the latest updates from the Mattermost blog.
- Facebook - Follow Mattermost.
- LinkedIn - Follow Mattermost.
- Email - Subscribe to our newsletter (1 or 2 per month).
- Mattermost - Join the ~contributors channel on the Mattermost Community Server.
- IRC - Join the #matterbridge channel on Freenode (thanks to matterircd).
Contributing
Please see CONTRIBUTING.md.
Any other questions, mail us at info@mattermost.com. We’d love to meet you!