Python binary detected multiple times when only installed once

[zhill]

What happened: Running syft 0.71.0 on docker.io/python:3.8.16-slim-bullseye shows two python binaries detected using metadata, but there is only one python package installed in the image.

Excerpt of the output:

   "id": "13cfac8a95b7e087",
   "name": "python",
   "version": "3.8.16",
   "type": "binary",
   "foundBy": "binary-cataloger",
   "locations": [
    {
     "path": "/usr/local/include/python3.8/patchlevel.h",
     "layerID": "sha256:39b5a69e8d44abf84aba019903294efc77e2e51ccabe6d80338ad11152e6bfb6"
    }
   ],
   "licenses": [],
   "language": "",
   "cpes": [
    "cpe:2.3:a:python_software_foundation:python:3.8.16:*:*:*:*:*:*:*",
    "cpe:2.3:a:python:python:3.8.16:*:*:*:*:*:*:*",
    "cpe:2.3:a:python:python:3.8.16:*:*:*:*:*:*:*"
   ],
   "purl": "pkg:generic/python@3.8.16",
   "metadataType": "BinaryMetadata",
   "metadata": {
    "classifier": "cpython-source",
    "realPath": "/usr/local/include/python3.8/patchlevel.h",
    "virtualPath": "/usr/local/include/python3.8/patchlevel.h"
   }
  },
  {
   "id": "b5e7b1f8cbfb8da1",
   "name": "python",
   "version": "3.8.16",
   "type": "binary",
   "foundBy": "binary-cataloger",
   "locations": [
    {
     "path": "/usr/local/lib/libpython3.8.so.1.0",
     "layerID": "sha256:39b5a69e8d44abf84aba019903294efc77e2e51ccabe6d80338ad11152e6bfb6"
    }
   ],
   "licenses": [],
   "language": "",
   "cpes": [
    "cpe:2.3:a:python_software_foundation:python:3.8.16:*:*:*:*:*:*:*",
    "cpe:2.3:a:python:python:3.8.16:*:*:*:*:*:*:*",
    "cpe:2.3:a:python:python:3.8.16:*:*:*:*:*:*:*"
   ],
   "purl": "pkg:generic/python@3.8.16",
   "metadataType": "BinaryMetadata",
   "metadata": {
    "classifier": "python-binary-lib",
    "realPath": "/usr/local/lib/libpython3.8.so.1.0",
    "virtualPath": "/usr/local/lib/libpython3.8.so.1.0"
   }
  },

However, looking in the image I see only one install of python:

root@bd9ea8bb8768:/# whereis python
python: /usr/local/bin/python /usr/local/bin/python3.8-config /usr/local/bin/python3.8 /usr/local/lib/python3.8
root@bd9ea8bb8768:/# whereis python3
python3: /usr/local/bin/python3.8-config /usr/local/bin/python3.8 /usr/local/bin/python3 /usr/local/lib/python3.8

What you expected to happen:
I expect the output to have a single "python" package binary detected using either or both of the pieces of evidence from the 'python-binary-lib' and 'cpython-source' classifiers if they are operating on the same package's metadata.

Something like:

 {
   "id": "b5e7b1f8cbfb8da1",
   "name": "python",
   "version": "3.8.16",
   "type": "binary",
   "foundBy": "binary-cataloger",
   "locations": [
    {
     "path": "/usr/local/bin/python3.8",
     "layerID": "sha256:39b5a69e8d44abf84aba019903294efc77e2e51ccabe6d80338ad11152e6bfb6"
    }
   ],
   "licenses": [],
   "language": "",
   "cpes": [
    "cpe:2.3:a:python_software_foundation:python:3.8.16:*:*:*:*:*:*:*",
    "cpe:2.3:a:python:python:3.8.16:*:*:*:*:*:*:*",
    "cpe:2.3:a:python:python:3.8.16:*:*:*:*:*:*:*"
   ],
   "purl": "pkg:generic/python@3.8.16",
   "metadataType": "BinaryMetadata",
   "metadata": {
    "classifier": "python-binary",
    "realPath": "/usr/local/bin/python3.8",
    "virtualPath": "/usr/local/bin/python3"
   }
  },

The above is just an example of something that would make sense, not exactly what I expect to see or even be possible.

I'm not sure how to make it clear which evidence (the lib or the cpython .h stuff) was used to determine the package, but the general approach I think could be to use those classifiers to determine the version of the python binary. I'm not sure how the paths could/should line up, but I think its better to be more narrowly constrained and correct than to report multiple entries.

Steps to reproduce the issue:
syft docker.io/python:3.8.16-slim-bullseye -o json=sbom.json

Anything else we need to know?:

Environment:

• Output of syft version:

Version:            0.71.0
JsonSchemaVersion:  6.2.0
BuildDate:          2023-02-09T19:40:41Z
GitCommit:          88c81d33edc67b0b9636d0046417a9373ac8b74f
GitDescription:     v0.71.0
Platform:           darwin/arm64
GoVersion:          go1.19.5
Compiler:           gc

• OS (e.g: cat /etc/os-release or similar):