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.
Just brew tap petere/postgresql
and then brew install <formula>
.
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 did not support that (the bogus postgresql8
and postgresql9
formulae in homebrew-versions
notwithstanding). This tap provides versioned formulae named postgresql-9.1
, postgresql-9.2
, 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 /usr/local/opt/postgresql-9.1/
etc.
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
/usr/local/opt/postgresql-9.2/bin/...
. This will be boring in the long run. - Add your preferred
/usr/local/opt/postgresql-x.y/bin
etc. 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 -f
thepostgresql-x.y
formula you prefer to use.- Install the
postgresql-common
package (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.
The standard 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 OS X can support, but also remove a number of Homebrew-level build options, to reduce complexity. I have also dropped supported for legacy OS X concerns, such as 32-bit Intel and PowerPC and really old OS X releases. Mainly because I can't test that anymore, YMMV.
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.
See /usr/local/Cellar/postgresql-common/HEAD/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_createcluster
, pg_dropcluster
, pg_ctlcluster
, and pg_lsclusters
instead of initdb
and pg_ctl
. The scripts take version numbers and instance names (which map to directory names). For example:
pg_createcluster 9.2 test
pg_ctlcluster 9.2 test start
See the respective man pages for details.
For client-side operations, to usual tools such as psql
and 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.