Can PyCG optimize to flow sensitive?

[vitsalis]

Can PyCG optimize to flow sensitive?

[ashwinprasadme]

I propose making flow-sensitive analysis configurable and incorporating it accordingly. Initially, I can introduce incremental changes on my branch, and we can review each step along the way.

[ashwinprasadme]

I have made flow-sensitivity improvements in my extension of PyCG for the HeaderGen project. I would like to incorporate these changes into the main PyCG implementation moving forward. However, I believe it would be beneficial to discuss the design decisions and make it even more robust. I'm open to having a discussion with @vitsalis and @gdrosos to explore these ideas further.

[vitsalis]

How do you solve the above issue in the extension?

[ashwinprasadme]

I have incorporated Def-use chains generated by Beniget to reason about variable definitions and their uses.

call graph generated by HeaderGen looks like this:

{
    "main": [
        "main.func2:17"
    ],
    "main.nested_func:1": [],
    "main.nested_func2:5": [],
    "main.param_func:9": [
        "main.nested_func2:5"
    ],
    "main.func:13": [],
    "main.func2:17": [
        "main.param_func:9"
    ]
}

The difference in the assignment graph looks like this for var_a, var_b, var_c:

PyCG

"main.var_b": [
    "main.var_c",
    "main.func"
],
"main.var_c": [
    "main.func2"
],
"main.var_a": [
    "main.var_b"
]

HeaderGen

"main.var_b:21": [
    "main.func:13"
],
"main.var_c:22": [
    "main.func2:17"
],
"main.var_b:24": [
    "main.var_c:22"
],
"main.var_a:25": [
    "main.var_b:24"
]