Dedup corner case

[LinjianMa]

Currently, the dedup function regards T.einsum('cb,ca->ab',B,B) and T.einsum('cb,ca->ba',B,B) as different expressions. However, their results are the same and should be viewed as duplicated.

[ByzanTine]

So for me, this sounds like a general case where you can logically evaluate the two einsums and they shall be the same.

Should we then put an evaluation engine that put common sense there?

Like einsum('.., ..->..', A, A) = A ? ('..' is any characters)

[LinjianMa]

I think the common sense you mentioned is the property of einsum when symmetry is involved. There're too many potential cases so I think we may need to transfer the einsum expression back to some union find data structure to detect all the cases. Seems it's not an easy case for me

[LinjianMa]

Should be fixed by #142