Add a pre- and post- hook for the process-compose wrapper

[johnalotoski]

• Allow prep and cleanup scripts, cmds or other customizations in the wrapper with pre- and post- process-compose hooks.
• The exec of the process-compose call line was removed to allow for a postHook.

[srid]

Curious what prompted this?

I considered adding a preHook equivalent (see initCommand commit in #55) but decided against doing it in favour of adding another one-off process which other processes can have a depends_on relation on. This way you get to see the output in the UI as well.

[johnalotoski]

What prompted the hook use was wanting to use shell variables obtained and/or set at runtime.

The simplest case was just wanting to use ${TMPDIR:=/tmp} as the base for a short postgres socket path which is then used in several process-compose processes, including postgres-init from services-flake.

When I first tried setting this as a process environment var in each process where it was needed, it evaluated it correctly in most places, but it fails to evaluate correctly in the probe commands, and then processes fails because the probes fail, the postgres ready probe in this case.

Using the hook approach with the shell var declared at the wrapper script level worked.

Here is an example preHook: link.

Example command for this process-compose stack (will download about ~1 GiB binaries, start syncing a public testnet Cardano chain, and extract chain data into postgres for analysis). Nix >= 2.17.0 and experimental-features = fetch-closure required:
nix run github:input-output-hk/cardano-parts/next-2024-01-10#run-process-compose-dbsync-sanchonet

[srid]

Cool, that makes sense. Nice to see the cardano project using this!