[FAST_BUILD] Migrate runhooks to python

[yuvipanda]

This can't be merged without also migrating start.sh, but opening this primarily to
show the implementation of source. I think that is the biggest barrier to migrating start.sh
to python, and this approach solves it quite neatly - am pretty proud of it :D

You can test it by creating a folder like this:

test
test/01-set-variable.sh
test/02-get-variable.sh

And the contents of the files are:

$ cat test/01-set-variable.sh 
export HELLO=WORLD

$ cat test/02-get-variable.sh 
echo "What is the HELLO to? It is to $HELLO"

And then run:

$ ./run-hooks.py test
Running hooks in: test as 501 gid: 20
Sourcing shell script: test/01-set-variable.sh
Sourcing shell script: test/02-get-variable.sh
What is the HELLO to? It is to WORLD
Done running hooks in: test

you can see that 01-set-variable exports a variable, which is then read by 02-get-variable!

We can turn this into a unit test later

This is built on top of #2006, see yuvipanda/docker-stacks@start-notebook-python...yuvipanda:docker-stacks:runhooks-py for just the run-hooks change

[yuvipanda]

@mathbunnyru take a look at this for the source implementation :) To me this was the part of start.sh that needed solving, and I think this does a pretty good job. What do you think of this approach?

[mathbunnyru]

That's a nice trick, and I love it!
Fortunately, I added testing for run-hooks only a couple months ago, so as longs as the tests are ok, we should be fine 🙂
I have a few ideas and will write them, when we merge previous PR and this one will be easier to read/comment/change.

[mathbunnyru]

@yuvipanda I merged the previous PR, please rebase/merge/create new PR for start.sh and run-hooks.py (whichever you choose and makes git diff nice and small is ok for me).

[yuvipanda]

@mathbunnyru I've merged that in here.

Since start.sh uses source to call run-hooks, we will have to migrate start and run-hooks together. This PR migrates run-hooks, but of course start has more going on. I'll try work on it later this week, but next week is more likely.

[yuvipanda]

But do tell me your ideas, I'll revisit here wherever possible

[manics]

What happens with bash functions that are defined in one hook and used in another?

[mathbunnyru]

I agree with @yuvipanda on @manics issue.

Generally speaking, we haven't promised the order in which shell scripts are sourced/called (the current implementation is in order and tests use this though).
I think, if you have a.sh and b.sh, it's not that difficult to move the function to some common shell file and source it from both a.sh and b.sh. And, to be honest, I think there are not so many people using this.

[mathbunnyru]

@yuvipanda also, when you will be porting start.sh, take a look at plumbum.
We already use it in this repo for testing and tagging, so it won't increase the number of technologies used in this repo.

This package is not added in the images, but I am pretty sure it's tiny and without lots of additional dependencies.
And it makes calling executables easier and more pretty-looking in Python.

[yuvipanda]

I'm pacing myself on this one, and spent another 15 minutes on it today addressing some of the review feedback :) start.sh is the big one and to start next, but will take it slow.

[mathbunnyru]

Added a test for unset: #2022

[mathbunnyru]

Added a test for changing variable value: #2023

[mathbunnyru]

Unfortunately, we have to comment this line, otherwise, tests fail, because they do not expect warnings.

[mathbunnyru]

@yuvipanda I added some tests for run-hooks and rewrote the test file a little bit, to make adding new tests easier.
I believe these tests will make merging this PR easier (when you are done with start.sh).
I only added tests that should work after this PR as well (so, no functions), but now we check unset works and double export with value change also works fine.

[mathbunnyru]

@yuvipanda do you have plans to continue working on this? 🙂

[mathbunnyru]

Closing this PR, as there is no activity here