Homebrew PostgreSQL things
These formulae allow installing multiple versions of PostgreSQL in parallel. This is similar to what you can do on certain Linux distributions, for example Debian.
To install something, first
brew tap petere/postgresql and then
brew install <formula>. Since there might be name overlaps with core Homebrew formulae, you should use fully qualified formula names like
brew install petere/postgresql/postgresql-common.
Since PostgreSQL major releases have incompatible data directories and other occasional incompatibilities, it is useful for many developers to keep several major versions installed in parallel for development, testing, and production. So far, Homebrew had inconsistent support for that and did not provide the full range of supported major versions. This tap provides versioned formulae named
email@example.com, etc. that you can install in parallel. Technically, these are "keg-only", which has the nice side effect that they are automatically installed in side-by-side directories
To use the programs installed by these formulae, do one or more of the following, in increasing order of preference:
- Call all programs explicitly with
/firstname.lastname@example.org/bin/.... This will be boring in the long run.
- Add your preferred
/email@example.com/binetc. to your path. Preferably to the front, to come before the operating system's PostgreSQL installation. This will work alright, but depending on your setup, it might be difficult to get everything on the OS to see the same path.
brew link -fthe
firstname.lastname@example.org you prefer to use.
- Install the
postgresql-commonpackage (see below).
The versioned formulae can be installed alongside the main
postgresql formula in Homebrew. But there will be a conflict if you do
brew link -f or install
postgresql-common, so in those cases you have to uninstall the main
postgresql package first. This is not a problem, however, because the versioned packages provide the same functionality.
postgresql formula in Homebrew is missing a number of build options and also has a number of build options that I find useless. These formulae enable all
configure options that macOS can support, but also remove a number of Homebrew-level build options, to reduce complexity. I have also dropped supported for legacy macOS concerns, such as 32-bit Intel and PowerPC and really old macOS releases. Mainly because I can't test that anymore, YMMV.
postgresql-common cluster manager
postgresql-common is a port of the postgresql-common package from Debian, which contains programs that help manage these multiple versioned installations, and programs to manage multiple PostgreSQL instances (clusters). The port a bit experimental, but it works.
/usr/local/opt/postgresql-common/README.Debian to get started. If you have used Debian or Ubuntu before, you'll feel right at home (I hope).
The general idea is that for server-side operations you use the special wrapper scripts
pg_lsclusters instead of
pg_ctl. The scripts take version numbers and instance names (which map to directory names). For example:
pg_createcluster 9.6 test pg_ctlcluster 9.6 test start
See the respective man pages for details.
For client-side operations, to usual tools such as
pg_dump are wrapped to automatically use the right version for the instance they are connecting to, so you usually don't need to do anything special. See the man page
pg_wrapper for details.
To install extensions, I recommend Pex. It has support for multiple PostgreSQL installations and can easily support to the installation scheme used by these packages. Example:
pex -g /email@example.com install ip4r