[postgis-devel] PostGIS 2.3 release plans

Regina Obe lr at pcorp.us
Thu Jul 21 23:11:25 PDT 2016


I think it's time we start putting a stake in the ground for PostGIS 2.3.

I would like us to release in mid September, primarily so we don't have to
support PostgreSQL 9.6 on PostGIS 2.2.  I'm hoping PostgreSQL 9.6 slated
release of September will be a little late as usual to buy us more time
which I think there is a good chance of given they are at beta2 (and still
have at least one rc to go).

I would also like to call feature freeze by early to mid August.  That said,
this is what I know we have in works or completed - much self-centered
around me cause I know more about what I have to do than what others have to
do.

1) Several functions already in place - our very first set of window
functions which I'm pretty excited about and other nice functions like
ST_Voronoi, ST_GeneratePoints
Plus some performance enhancements

http://postgis.net/docs/manual-dev/PostGIS_Special_Functions_Index.html#NewF
unctions_2_3

(so nothing to do here except more testing. Some I've already stress tested
a bit).


2) More testing on parallelism with PostgreSQL 9.6 (beta2 has a bunch of
fixes for parallel support I need to test). Marking costs on more functions.
Paul Ramsey already marked off PostGIS geometry/geography functions and Paul
Norman provided costs.
I marked off raster ones that I think are parallel safe.  I still need to
stress test and compare with old behavior on real work-loads and make sure
no crashing and burning and come up with costs for raster functions.

-- Things not yet committed but on the horizon slated for PostGIS 2.3 at
this point ( 8 pull requests - https://github.com/postgis/postgis/pulls )
3) Making PostGIS non-relocatable and schema qualifying function calls.
This I may need help with: As I stated in PostGIS 2.3, we need to schema
qualify our function calls primarily because setting search_path on
functions breaks inlining behavior needed for functions that utlize spatial
indexes and without schema qualification all sorts of things fall apart like

materialized views, backup / restore , logical decoding replication etc.
A good chunk of this was patched with optional helper script released in
PostGIS 2.2.2 to set function search_path (but functions that rely on
indexes could not be patched)


My proposal is outlined here:

https://trac.osgeo.org/postgis/ticket/3496

I could do it all myself if all we had to worry about was building for
extensions. To support the old-fashioned way, I may need some help.

The part I need help with is revising our perl scripts, to take in an extra
argument (whether to replace @extschema at . with the default schema specified
during configure (if any) or take it out if non-specified.

So I would change all the functions that call other functions to use
@extschema at .

For extension include, this will be unchanged, since extension plumbing
handles replace for that already.  For non-extension installs, this would
either be

replace @extschema at .  with nothing, or replaced with what a user passes in
for PostGIS install location during configure.

So our perl scripts need to take an additional parameter to denote if they
should strip out that, replace it, or do nothing with it.  I guess this can
be done after the script pass by adding another
Step in our generation.  

The extension install can no longer use what was already built for the
old-fashioned install since it would have the @extchema@ stripped or
hard-coded with some schema and extension script will need this untouched.

4) Tiger 2016 upgrade for postgis_tiger_geocoder - I expect new Tiger data
will be out in August and plan to patch the load scripts to support.

5) BRIN support -- I think some cleanup is still left with that patch being
worked on, but looks pretty close for commit. -
https://github.com/postgis/postgis/pull/106  So I would consider it a done
deal except, it's not in our code base yet.  I want this in before
mid-August so we can start stress testing it.

6) Precision Model support - Dan how much more do you have left before you
can commit? https://github.com/postgis/postgis/pull/100 (or do we need to
push to PostGIS 2.4?)
7) ST_Angle function - https://github.com/postgis/postgis/pull/97
8) Validity Flag - https://github.com/postgis/postgis/pull/99
9) ST_AsText -- adding optional precision argument --
https://github.com/postgis/postgis/pull/94
10) ST_AsGeoBuf -- https://github.com/postgis/postgis/pull/108  (still seems
a bit to go so may not make an August cut without some loving)

Feel free to add anything I missed and if I forgot your favorite patch, I'm
sorry in advance.

Thanks,
Regina








More information about the postgis-devel mailing list