Skip initdb during cargo pgx init if $USER == "root"

[anth0nyleung]

Notes

In some use case, we would set up rust toolchain and pgx as root user, we do not necessarily need pgx to spin up running postgres instance. However during cargo pgx init, as pgx is trying to initdb for postgres, it will fail to do so since initdb cannot be executed as root user.

Changes

If current user is root user, skip doing initdb.

Testing

cargo test --features pg14 --no-default-features passed

> ls -l ~/.pgx   
total 0

> cargo pgx init --pg14 /usr/local/pgsql/bin/pg_config 
   Validating /usr/local/pgsql/bin/pg_config
 Initializing data directory at /Users/antholeu/.pgx/data-14

> ls -l ~/.pgx                                        
total 8
drwx------  24 antholeu  staff   768B Aug 25 16:36 data-14
-rw-r--r--   1 antholeu  staff    48B Aug 25 16:36 config.toml

> ls -l ~/.pgx                                             
total 0

> sudo cargo pgx init --pg14 /usr/local/pgsql/bin/pg_config
   Validating /usr/local/pgsql/bin/pg_config
   Skipping initdb as current user is root user

> ls -l ~/.pgx                                             
total 8
-rw-r--r--  1 root  staff    48B Aug 25 16:37 config.toml

[workingjubilee]

This looks good per se! I just have some design-level questions so I can better understand why to add this, and not other solutions.

[workingjubilee]

Currently we avoid running initdb if pg_config.data_dir() exists.

But presumably, in the case that the user has specified a given path, the database should in fact always not be initialized by us, as the user wants a specific database to be used and likely has it configured the way they like it, or will want to do so themselves?

It may be that we will not be able to execute successive commands using cargo pgx if initdb has not been run, but we will run into that issue if this flag is passed as well, so we must handle that case anyways. After all, cargo pgx cannot simply assume that the system state has remain unchanged between runtime sessions. This is especially true if cargo pgx is pointed at a specific PostgreSQL database and so is not "in control" of it.

Is this hypothesis incorrect?

[workingjubilee]

This hypothesis is incorrect! PGX expects a specific path for the data_dir to exist, it's not actually controlled by pg_config, see: https://github.com/tcdi/pgx/blob/1a89778046547c52c714fe6885069eab5638f965/pgx-utils/src/pg_config.rs#L199-L202

[anth0nyleung]

datadir is currently hardcoded to data-<major-version>, user is not allow to specify this.

[workingjubilee]

Thank you!