How to access external data when making an analysis?

[ChijinZ]

Hi, I'm now implementing a project related to tensor computation graph rewriting. I would like to attach custom data (e.g., output tensor shape) with each eclass. For example:

define_language! {
    pub enum MyLanguage {
        "tensor"     = Tensor(Id),
        "transpose"  = Transpose(Id),
    }
}

#[derive(Default, Debug, Clone)]
pub struct MyAnalysis {}

#[derive(Default, Debug, Clone)]
pub struct TensorShape {
    pub shape: Vec<usize>,
}

impl Analysis<MyLanguage> for MyAnalysis {
    type Data = TensorShape;
    fn make(egraph: &EGraph<MyLanguage, Self>, enode: &MyLanguage) -> Self::Data {
        let x = |i: &Id| egraph[*i].data;
        match enode {
            MyLanguage::Tensor(id) => {
                // how can I know the shape of this tensor?
            },
            MyLanguage::Transpose(id) => {
                // trasfer shape
                unimplemented!()
            }
        }
    }
    fn merge(&mut self, a: &mut Self::Data, b: Self::Data) -> DidMerge {
        unimplemented!()
    }
}

fn main() {
    let expr = "(transpose (tensor 0))";
    let tensorshapes = vec![TensorShape { shape: vec![1, 2, 3] }];
    let mut runner = Runner::<MyLanguage, MyAnalysis>::default();
    runner.run(..);
}

However, I cannot find a good way to access the tensorshape of MyLanguage::Tensor(id) in the MyAnalysis::make. My solution for now is to change the MyLanguage to something like MyLanguage::Tensor(Id, Vec<Id>), and embeding the tensor shape into the Vec<Id>. But it seems not a good solution.

Is there a more elegant way to achieve this goal?