PR feature table

[Spedoske]

@tianyin @tylergu This is the features & changes table in plan 2 #235 (comment).

I think the first plan is hard because there are already a lot of code involved between the first commit and the first commit that passes the unit tests. These changes are necessary and atomic for the whole project, which make them really hard to be in different PRs. cfb5b0a...777a77d

Features & Changes	Files	Notes
Build Systems	.pre-commit-config.yaml, requirements.txt	Generate requirements.txt for backward compability
Build Systems	flake.nix, flake.lock, shell.nix, ray-dashboard.patch	Use nix to make sure Acto deployment is the same across nodes, used in Ray executor
Build Systems	pyproject.toml, poetry.lock	Use poetry to manage python dependencies
Checker Interface	acto/checker/*	Implement new checker interface
Checker Result Interface	acto/checker/*, acto/common.py	Implement new checker result interface
Collector	acto/kubectl_client/collector.py	Move Kubernetes cluster information collection into a standalone class
Config	config.yaml, config_post_diff_test.yaml, acto/config_schema.py, acto/config.py	Config files to run Acto
Coverage Collection	data/rabbitmq-operator-with-coverage/*	Example operator with coverage collection
Coverage Collection	acto/kubectl_client/collector.py	Implement coverage collection
Custom Checker Example	data/strimzi-kafka-operator/oracle.py	Implement Java Log Oracle
Deployment	scripts/ansible/*	Deployment Acto using Ansible, used in Ray executor
Deploy Interface	acto/deploy.py	Deploy Interface is changed because of the Runner Interface change
DIff tests	acto/post_diff_test.py, acto/post_process/*	rewrite Diff tests due to the change of Trials, Runners, and Deploy
Engine	acto/engine_new.py	Engine is rewritten because of the change of Runner, Deploy, Trial, collector Interface
Input	acto/input/*	move test cases partition out of the input module because the original methods is bound to thread local variable, which is not suitable now because executors can have different implementations.
Kafka Operator	data/strimzi-kafka-operator/*	Port Kafka operator
Parallel Executors	acto/ray_acto/*	It is NOT ray, it is an abstraction layer of parallel executors, with different backends including thread, process, and Ray.
Reproduce	acto/reproduce.py	Reproduce is rewritten because of the change of Runner, Deploy, Trial, Acto Interface
Runner	acto/runner/runner.py	An abstraction for an executor that accepts a trial and a function to collect snapshots, return the completed trial.
Snapshot	acto/snapshot.py	Add more fields and make it to able trace back previous snapshots.
Snapshot Collector	acto/runner/snapshot_collector.py	A set of functions to build functions that agree with the Snapshot collector function interface.
Tests	test/*	Fix tests for the API breaking changes
Trial	acto/runner/trial.py	Provide two layer of abstraction, the first layer is TrialInputIteratorLike, which decides what to do with the inputs when receiving different oracleControlFlow. The second layer is Trial, which decides how to interact with the snapshot collector.

[tianyin]

Add @tylergu and @whentojump to see how to merge.

[tianyin]

There are a lot of good code in the PR.

[tylergu]

@Spedoske I think we can start merging from the "build system". Then for the next step we can merge the "code coverage collection", then the "checker interface". These changes should be easy to separate out and not depend on other changes.