OpSum to TTN refactor - Githubissues

ITensor / ITensorNetworks.jl

A package with general tools for working with higher-dimensional tensor networks based on ITensor.

MIT License

51 stars 12 forks source link

OpSum to TTN refactor #166

Closed emstoudenmire closed 1 month ago

emstoudenmire commented 2 months ago

This PR refactors the ttn_svd function into three functions, make_sparse_ttn, svd_bond_coefs, and compress_ttn.

Also the calc_qn function and associated cache has been turned into a "function object" called TermQN.

Other changes include:

remove some unnecessary helper data structures in opsum_to_ttn.jl
move code for MatElem and QNArrElem types into separate files
add more ITensors.Ops functions to using statement in opsum_to_ttn.jl
fix a missing keyword argument in src/apply.jl

emstoudenmire commented 1 month ago

Ok so the status of this PR is that I've made all the changes you recommended.

I have not done the following. I.e. these will be for future PR's:

have not replaced MatElem with FillArrays.OneElem (turned out to be a little complicated to do for reasons I can explain)
have not started using SparseArraysDOK yet (but that was for later anyway)
have not started using DataGraphs yet either (also pending more planning and design considerations for future PR's)

mtfishman commented 1 month ago

Sounds like a reasonable place to leave things, I'll merge once tests pass.

emstoudenmire commented 1 month ago

Great. Yes this PR was mostly just about splitting the functions. The excessive arguments ought to come down once the data structures and graph analysis patterns are better (in future PR's).