Add option to output the number of unique dependency #386
Description
Describe the feature
This would give the number of library we could lose if we delete one dependency. For instance:
captum==0.7.0 (unique, 10 unique dependencies)
├── matplotlib [required: Any, installed: 3.8.4] (unique, 6 unique dependencies)
│ ├── contourpy [required: >=1.0.1, installed: 1.2.1] (unique)
│ │ └── numpy [required: >=1.20, installed: 1.24.4]
│ ├── cycler [required: >=0.10, installed: 0.12.1] (unique)
│ ├── fonttools [required: >=4.22.0, installed: 4.53.0] (unique)
│ ├── kiwisolver [required: >=1.3.1, installed: 1.4.5] (unique)
│ ├── numpy [required: >=1.21, installed: 1.24.4]
│ ├── packaging [required: >=20.0, installed: 24.1]
│ ├── Pillow [required: >=8, installed: 9.5.0] (unique)
│ ├── pyparsing [required: >=2.3.1, installed: 3.1.2] (unique)
│ └── python-dateutil [required: >=2.7, installed: 2.9.0.post0]
│ └── six [required: >=1.5, installed: 1.16.0]
├── numpy [required: Any, installed: 1.24.4]
├── torch [required: >=1.6, installed: 1.13.0] (unique, 1 unique dependency)
│ └── typing_extensions [required: Any, installed: 4.12.2] (unique)
└── tqdm [required: Any, installed: 4.66.4] (unique)
category-encoders==2.6.0 (unique, 8 unique dependencies) # patsy, pytz, pandas, scipy are only defined in this one
├── numpy [required: >=1.14.0, installed: 1.24.4]
├── pandas [required: >=1.0.5, installed: 1.5.3]
│ ├── numpy [required: >=1.21.0, installed: 1.24.4]
│ ├── python-dateutil [required: >=2.8.1, installed: 2.9.0.post0]
│ │ └── six [required: >=1.5, installed: 1.16.0]
│ └── pytz [required: >=2020.1, installed: 2024.1]
├── patsy [required: >=0.5.1, installed: 0.5.6]
│ ├── numpy [required: >=1.4, installed: 1.24.4]
│ └── six [required: Any, installed: 1.16.0]
├── scikit-learn [required: >=0.20.0, installed: 1.1.3] (unique, 2 unique dependencies)
│ ├── joblib [required: >=1.0.0, installed: 1.2.0] (unique)
│ ├── numpy [required: >=1.17.3, installed: 1.24.4]
│ ├── scipy [required: >=1.3.2, installed: 1.8.1]
│ │ └── numpy [required: >=1.17.3,<1.25.0, installed: 1.24.4]
│ └── threadpoolctl [required: >=2.0.0, installed: 3.5.0] (unique)
├── scipy [required: >=1.0.0, installed: 1.8.1]
│ └── numpy [required: >=1.17.3,<1.25.0, installed: 1.24.4]
└── statsmodels [required: >=0.9.0, installed: 0.14.2] (unique)
├── numpy [required: >=1.22.3, installed: 1.24.4]
├── packaging [required: >=21.3, installed: 24.1]
├── pandas [required: >=1.4,!=2.1.0, installed: 1.5.3]
│ ├── numpy [required: >=1.21.0, installed: 1.24.4]
│ ├── python-dateutil [required: >=2.8.1, installed: 2.9.0.post0]
│ │ └── six [required: >=1.5, installed: 1.16.0]
│ └── pytz [required: >=2020.1, installed: 2024.1]
├── patsy [required: >=0.5.6, installed: 0.5.6]
│ ├── numpy [required: >=1.4, installed: 1.24.4]
│ └── six [required: Any, installed: 1.16.0]
└── scipy [required: >=1.8,!=1.9.2, installed: 1.8.1]
└── numpy [required: >=1.17.3,<1.25.0, installed: 1.24.4]
It can be simply calculated : at each node, we take the set of libraries under the node and we subtract it to the set of libraries in the other nodes. The difference in length between the result of the operation and the length of the set of libraries under the node gives the number of unique dependencies.