I have a Node-based script runner with two modes of operation: process and
require. It has a set of scripts with their schedules and runs them based on
those schedules. Some scripts are invoked using the process mode - cp. Some
are invoked by require-ing them.
I want to transition all scripts to the require-based mode, however, it is
harder to make isolated. For scripts which run using require, I override their
console methods to capture as their output (as opposed to stdout and stderr
of the cp based mode) and try-catch their execution to capture their
errors. I also have to reset process.argv and walk into their directory using
process.chdir and back out after their execution so that from their perspective
it is as if they were ran using node . in their directory and not require in
the runner's directory.
This feels brittle and I'm looking for a better way and I figured Node's VM
might be it. However, as this code showcases, that's not the case. Not only
would I have to redefine all the APIs the scripts might use (which currently I
just need to sanitize them as the script might use them and I need to fool it
into thinking they are the same as if it was run directly and not required),
but there is a problem with require in the VM basically undoing the whole
context rendering the whole exercise useless. Without require in the script
I would have the send in the script content as text, which is fine, but the
script itself also might have dependencies and that's where it all breaks down.
node . test