Skip to content

development.rst

Latest commit

 

History

History
276 lines (188 loc) · 12.1 KB

development.rst

File metadata and controls

276 lines (188 loc) · 12.1 KB

Development

Developer documentation with dependancies, build instructions, how to build a Docker image, update the documentation and run unit tests.

Requirements

  • PostgreSQL and PostgreSQL development packages must be installed (pg_config and server headers). For Red Hat and Ubuntu, the package names are usually postgresql-dev or postgresql-devel
  • LibXML2 development packages must be installed, usually libxml2-dev or libxml2-devel
  • CUnit packages must be installed
  • [Optional] laz-perf library may be installed for LAZ compression support (built from source)

Build/Install

After generating the configure script with autogen.sh, you can use ./configure --help to get a complete listing of configuration options.

$ ./autogen.sh
$ ./configure
$ make
$ sudo make install

Note

You can use --with-pgconfig on the ./configure command line if you have multiple PostgreSQL installations on your system and want to target a specific one. For example:

$ ./configure --with-pgconfig=/usr/lib/postgresql/12/bin/pg_config

Note

If qsort_r is not available on your system, then you can use the embedded implementation instead thanks to the next directive:

$ ./configure CFLAGS="-DNESTED_QSORT=1"

Tests

Unit tests

$ make check

Regressions tests

pgPointcloud includes SQL tests to run against an existing installation. To run the SQL tests:

$ sudo make install
$ PGUSER=a_user PGPASSWORD=a_password PGHOST=localhost make installcheck

This command will create a database named contrib_regression and will execute the SQL scripts located in pgsql/sql in this database.

Write a loading system

If you are writing your own loading system and want to write into Pointcloud types, create well-known binary inputs, in uncompressed format. If your schema indicates that your patch storage is compressed, Pointcloud will automatically compress your patch before storing it, so you can create patches in uncompressed WKB without worrying about the nuances of particular internal compression schemes.

The only issues to watch when creating WKB patches are: ensuring the data you write is sized according to the schema (use the specified dimension type); ensuring that the endianness of the data matches the declared endianness of the patch.

Documentation

Sphinx is used to build the documentation. For that, you have to install the next Python packages:

  • sphinx
  • sphinx_rtd_theme

Then:

$ cd doc && make html

The HTML documentation is available in doc/build/html.

Note

The documentation can be generated in another format like pdf, epub, ... You can use make to get a list of all available formats.

Docker Image

A Dockerfile is provided in the docker directory and based on the official PostgreSQL docker image available DockerHub. The image generated is based on PostgreSQL 14, PostGIS 3 and the laz-perf support is activated.

$ docker build --rm -t pgpointcloud docker/

Continuous Integration

pgPointcloud tests are run with Github Actions on several Ubuntu versions and with various PostgreSQL/PostGIS releases:

  w/o PostGIS PostGIS 3.3
PostgreSQL 13

13_33_jammy

13_33_noble

13_33_jammy

13_33_noble
PostgreSQL 14

14_33_jammy

14_33_noble

14_33_jammy

14_33_noble
PostgreSQL 15

15_33_jammy

15_33_noble

15_33_jammy

15_33_noble
PostgreSQL 16

16_33_jammy

16_33_noble

16_33_jammy

16_33_noble
PostgreSQL 17

17_33_jammy

17_33_noble

17_33_jammy

17_33_noble

Release

Steps for releasing a new version of Pointcloud:

  1. Add a new section to the NEWS file, listing all the changes associated with the new release.
  2. Change the version number in the README, Version.config and pgsql/expected/pointcloud.out files.
  3. Update the value of UPGRADABLE in pgsql/Makefile.in and pgsql_postgis/Makefile. This variable defines the versions from which a database can be upgraded to the new Pointcloud version.
  4. Create a PR with these changes.
  5. When the PR is merged create a tag for the new release and push it to GitHub:
$ git tag -a vx.y.z -m 'version x.y.z'
$ git push origin vx.y.z

Valgrind memcheck

For checking the memory management of pgPointcloud extension, Valgrind can be used with PostgreSQL in single-user mode.

But first, it's necessary to compile the extension with debug symbols and without compiler optimizations:

$ ./configure CFLAGS="-O0 -g"
$ make
$ sudo make install

Debug symbols may also be installed for PostgreSQL and PostGIS. For example for Debian based distributions with PostgreSQL 13 and PostGIS 3:

$ sudo apt-get install postgresql-13-dbgsym postgresql-13-postgis-3-dbgsym

And finally:

$ echo "select pc_transform(patch, 1) from patchs limit 1" | \
  valgrind -s --track-origins=yes --leak-check=yes \
    --show-leak-kinds=all --read-var-info=yes --log-file=/tmp/valgrind.log \
    /usr/lib/postgresql/13/bin/postgres --single -D /var/lib/postgresql/13/main \
    -c config_file=/etc/postgresql/13/main/postgresql.conf \
    mydatabase

Then Valgrind's report is available in /tmp/valgrind.log.

GDB interactive mode

GDB may be attached to a running session for debugging purpose:

$ psql mydatabase
psql (14.5)
Type "help" for help.

mydatabase=# select pid from pg_stat_activity where usename = 'pblottiere' and state = 'active';
  pid
-------
 34699
(1 row)
$ sudo gdb -p 34699
GNU gdb (GDB) 12.1
(gdb)

Then you can execute a SQL request in the corresponding session and use GDB as usual (step by step, backtrace, ...).