[FEATURE]: add Python binding

[1a1a11a]

We would like to provide the library for Python users.

[haochengxia]

I'm interested in building a Python package, starting with providing cachesim, which will allow us to run simulations easily using Python code.

e.g.,

from libcachesim import cachesim

results = cachesim(trace_path, trace_type, algos, cache_size_ratios, ...)

Would this be an appropriate approach for an initial Python binding?

[1a1a11a]

Hi Haocheng, this sounds good!
I would also like to expose the cache interface, e.g.,

cache = FIFO(cache_size, **kwargs)
for req in reader: 
    hit = cache.get(req)

We can leave it to later, but when you set up the binding, it would be good to think about how we can support this in the future.

[haochengxia]

Support exposing cache interface (example in example_cache.py)

$ cachesim cloudPhysicsIO.vscsi vscsi fifo 0.1
cloudPhysicsIO.vscsi FIFO cache size   194MiB,           113872 req, miss ratio 0.7999, throughput 3.40 MQPS

$ python example_cache.py
hit_cnt:  22789
miss_cnt:  91083
hit_rate:  0.7998717858648307

[1a1a11a]

If the Python binding grows large, we may consider creating a new repo so that it would be easier to maintain

[haochengxia]

Python Binding Roadmap

Current Status

We've made good progress on the Python bindings, supporting basic structures and eviction policy creation:

1. Core Simulation Structs: We now support Python bindings for essential simulation structures:

   • cache_t
   • reader_t
   • request_t

2. Eviction Policy Integration: You can create caches with specific eviction algorithms using two methods:

   • Python Classes (Limited but with Parameter Hints): Dedicated Python classes are available for the following algorithms, offering clear parameter hints:
     • ARC
     • FIFO
     • LRB
     • LRU
     • S3FIFO
     • Clock
     • Sieve
     • ThreeLCache
     • TinyLFU
     • TwoQ
   • Generic Python Method: The create_cache method uses cache_init from cache_sim and accepts the algorithm name as a string to invoke it.

---

Next Steps (Prioritized)

Here's what's next on our Python binding roadmap:

• Implement Basic Analysis Capabilities: Enable core analysis functionalities directly within the Python bindings.
• Full Feature Support for Basic Structs: Expand the bindings to include all features and functionalities of the cache_t, reader_t, and request_t structs.
• Add More Eviction Algorithm Classes: Introduce additional Python classes for other eviction algorithms, providing the same clear parameter hints as existing ones.
• Create a PyPI package for our python binding -> need to decide the name