wiki:PostGISDevelopment2021

Version 18 (modified by robe, 3 years ago) ( diff )

Development Meeting - Virtual - 2021

Dates

Friday August 20th (estimated time period 2 hrs) - 3PM - 5PM UTC

Goals

The main aim is to begin thinking about current planned release of PostGIS 3.2.0 and major features to be attempted.

  • wiki:PostGIS32 topics
  • GSOC Gist building enhancements for PostgreSQL 14
  • PostGIS 3.2 ticket triage
  • Need for a postgis_c_api for projects like MobilityDB, h3_pg
  • Real World data testing
  • Vote: Should our master repo be master or main?
  • Add your own … ?

Attendance

PostGIS people who will be there

Put your name on the poll if you plan to attend

Venue


Meeting Notes

Welcome

  • Paul Ramsey (Crunchy Data) … coordination towards release time in fall
  • Mahmoud SAKR (MobilityDB project, Université Libre de Bruxelles) - Discussing the dependancy of MobilityDB on PostGIS
  • Darafei Praliaskouski (Kontur) - XMeans and future of parallel things, H3 things and creation of extensions depending on main PostGIS
  • Regina (self)
  • Raúl Marín (self)
  • Mohamed Bakli (MobilityDB)
  • Brian M Hamlin - OSGeo Live DVD
  • Bruce Rindahl (self) documentation
  • Giuseppe Broccolo - give updates about improvements in GiST build (with Han)
  • Han Wang - GiST with pre-sorting methods
  • Vicky Vergara (pgRouting & OSGeoLive & MobilityDB)
  • Maxime Schoemans (MobilityDB)
  • Esteban Zimanyi (MobilityDB)
  • Jignesh Shah (AWS) .. learning about PostGIS (More on managing multiple versions)
  • Jeremy Schneider (AWS) — PostGIS build considerations
  • Aliaksandr Kalenik (Kontur)

Agenda

  • GSOC Gist building enhancements for PostgreSQL 14
  • Need for a postgis_c_api for projects like MobilityDB, h3_pg
  • PostGIS 3.2 ticket triage
  • Real World data testing
  • Vote: Should our master repo be master or main?

GIST Building GSOC Project

  • Good PoC completed
  • Uses Pg14 pre-sort hooks in GIST to speed index build
  • Index build time is made much faster
  • Query performance may be slightly (?) degraded compared to old implementation w/o pre-sort
  • Pre-sort changes low-level page composition a little bit deleteriously
  • Need some feedback on query against real-world data sets
  • Is the sort key length (32bit) the problem? (Maybe not? Datum size is 64bit on most hardware)
  • Q: Do PgSQL implementations of presort also show a performance degredation in query?
  • Some degradation is expected. check first: https://observablehq.com/@mourner/hilbert-curve-packing

postgis_c_api

master or main

  • master - PR, Regina +1, Giuseppe, MS, Darafei +0 BH +1,JeremyVicky -0
  • main - BR, Regina -1, Darafei +1,Jignesh Vicky +0
  • Don't care: RM, Darafei +1 MD for now, but plan should be to change to main

Real World Data Testing

PostGIS 3.2 Ticket Triage

Alpha Release

Dependent on

  • Han's GIST sort merge
  • KMeans merge
  • Boundary merge
  • Paul's three feature tickets
  • Paul's crasher tickets
  • Sandro's cleanup of topology tickets and feature completion
  • Upgrade changes (Regina)
  • Address standardizer (Regina)

Target date?

  • August 34th

Next Meeting?

  • Maybe pre-release 3.2?
  • For sure mid-winter (Jan/Feb?) + 1
Note: See TracWiki for help on using the wiki.