open Lacaml.D

module Ds = DeterminantSpace
module De = Determinant
module Sp = Spindeterminant

type t = 
  {
    mo_basis  : MOBasis.t ;
    det_space : DeterminantSpace.t ;
    ci        : CI.t ;
    hf12_integrals : HF12.t ;
    eigensystem : (Mat.t * Vec.t) lazy_t;
  }

let ci t        =  t.ci
let mo_basis t  =  t.mo_basis
let det_space t =  t.det_space
let mo_class t  =  Ds.mo_class @@ det_space t
let eigensystem t =  Lazy.force t.eigensystem



let hf_ij_non_zero hf12_integrals deg_a deg_b ki kj =
  let integrals = [
    let one_e _ _ _ = 0. in
    let hf12_s, hf12_o, hf12m_s, hf12m_o = hf12_integrals in
    let kia = De.alfa ki and kib = De.beta ki
    and kja = De.alfa kj and kjb = De.beta kj
    in
    let mo_a =
      Bitstring.logand (Sp.bitstring kia) (Sp.bitstring kja)
      |> Bitstring.to_list
    and mo_b =
      Bitstring.logand (Sp.bitstring kib) (Sp.bitstring kjb)
      |> Bitstring.to_list
    in
    let two_e i j k l s s' =
      if s' = Spin.other s then
	hf12_o.{i,j,k,l}  -. 1. *. (
	(List.fold_left (fun accu m -> accu +. hf12m_o.{m,i,j,k,l}) 0. mo_a) +.
	(List.fold_left (fun accu m -> accu +. hf12m_o.{m,j,i,l,k}) 0. mo_b) 
        )
      else
	hf12_s.{i,j,k,l} -. 1. *. (
	(List.fold_left (fun accu m -> accu +. hf12m_s.{m,i,j,k,l}) 0. mo_a) +.
	(List.fold_left (fun accu m -> accu +. hf12m_s.{m,j,i,l,k}) 0. mo_b) 
        )
    in
    (one_e, two_e)
  ]
  in
  CIMatrixElement.non_zero integrals deg_a deg_b ki kj
  |> List.hd




let dressing_vector ~frozen_core hf12_integrals f12_amplitudes ci =

  if Parallel.master then
    Printf.printf "Building matrix\n%!";
  let det_space = 
    ci.CI.det_space 
  in

  let m_HF = 
    
    let f =
      match Ds.determinants det_space with
      | Ds.Arbitrary _ -> CI.create_matrix_arbitrary
      | Ds.Spin      _ -> CI.create_matrix_spin_computed
    in
    f (fun deg_a deg_b ki kj ->
         hf_ij_non_zero hf12_integrals deg_a deg_b ki kj
      ) det_space 

  in

  Matrix.mm (Lazy.force m_HF) (Matrix.dense_of_mat f12_amplitudes)


let make ~simulation ?(threshold=1.e-12) ~frozen_core ~mo_basis ~aux_basis_filename ?(state=1) () =

  let det_space = 
    DeterminantSpace.fci_of_mo_basis  mo_basis  ~frozen_core 
  in

  let ci = CI.make ~n_states:state det_space in

  let hf12_integrals =
    HF12.make ~simulation ~mo_basis ~aux_basis_filename ()
  in

  let ci_coef, ci_energy =
    let x = Lazy.force ci.eigensystem in
    Parallel.broadcast (lazy x)
  in

  let eigensystem = lazy (
    let m_H =
      Lazy.force ci.CI.m_H
    in

      
    let rec iteration ~state psi = 
      let column_idx = iamax (Mat.to_col_vecs psi).(state-1) in

      let delta = 
         (* delta_i = {% $\sum_j c_j H_{ij}$ %} *)
         dressing_vector ~frozen_core hf12_integrals psi ci
         |> Matrix.to_mat
      in

      Printf.printf "Cmax : %e\n" psi.{column_idx,state};
      Printf.printf "Norm : %e\n" (sqrt (gemm ~transa:`T delta delta).{state,state});

      let f = 1.0 /. psi.{column_idx,state} in
      let delta_00 =
        (* Delta_00 = {% $\sum_{j \ne x} delta_j c_j / c_x$ %} *)
        f *. ( (gemm ~transa:`T delta psi).{state,state} -.
          delta.{column_idx,state} *. psi.{column_idx,state} )
      in
      Printf.printf "Delta_00 : %e %e\n" delta.{column_idx,state} delta_00;

      delta.{column_idx,state} <- delta.{column_idx,state} -. delta_00;


      let eigenvectors, eigenvalues = 

(* Column dressing 
*)
        let delta = lacpy delta in
        Mat.scal f delta;
        for k=1 to state-1 do
          for i=1 to Mat.dim1 delta do
                delta.{i,k} <- delta.{i,state}
          done;
        done;
        let diagonal =
          Vec.init (Matrix.dim1 m_H) (fun i ->
            if i = column_idx then
              Matrix.get m_H i i +. delta.{column_idx,state}
            else
              Matrix.get m_H i i
            )
        in

        let matrix_prod c = 
          let w = 
            Matrix.mm ~transa:`T m_H c
            |> Matrix.to_mat
          in
          let c = Matrix.to_mat c in

          for k=1 to state do
            for i=1 to (Mat.dim1 w) do
              w.{i,k} <- w.{i,k} +.  delta.{i,k} *. c.{column_idx, k} ;
              w.{column_idx,k} <- w.{column_idx,k} +.  delta.{i,k} *. c.{i,k};
            done;
            w.{column_idx,k} <- w.{column_idx,k} -.
                delta.{column_idx,k} *. c.{column_idx,k};
          done;

          Matrix.dense_of_mat w
        in


(* Diagonal dressing *)
(*
        let diagonal =
          Vec.init (Matrix.dim1 m_H) (fun i ->
            Matrix.get m_H i i +.
              if (abs_float psi.{i,state} > 1.e-8) then
                delta.{i,state} /. psi.{i,state}
              else  0.
            )
        in

        let matrix_prod c = 
          let w = 
            Matrix.mm ~transa:`T m_H c
            |> Matrix.to_mat
          in
          for i=1 to (Mat.dim1 w) do
            w.{i,state} <- w.{i,state} +. delta.{i,state}
          done;
          Matrix.dense_of_mat w
        in
*)


        Parallel.broadcast (lazy (
          Davidson.make ~threshold:1.e-9 ~guess:psi ~n_states:state diagonal matrix_prod
          ))

(*
        let m_H = Matrix.to_mat m_H |> lacpy in
*)

(* DIAGONAL TEST 
        for i=1 to Mat.dim1 m_H do
          if (abs_float psi.{i,state} > 1.e-8) then
            m_H.{i,i} <- m_H.{i,i} +. delta.{i,state} /. psi.{i,state};
        done;
*)


(* COLUMN TEST
        for i=1 to Mat.dim1 m_H do
          let d = delta.{i,state} /. psi.{column_idx,state} in
          m_H.{i,column_idx} <- m_H.{i,column_idx} +. d;
          if (i <> column_idx) then
            begin
              m_H.{column_idx,i} <- m_H.{column_idx,i} +. d;
              m_H.{column_idx,column_idx} <- m_H.{column_idx,column_idx} -.
                d *. psi.{i,state}
            end
        done;
*)



(*
        let m_v = syev m_H in
        m_H, m_v
*)
      in


      Vec.iter (fun energy -> Printf.printf "%g\t" energy) eigenvalues;
      print_newline ();

      let conv =
        1.0 -. abs_float ( dot
          (Mat.to_col_vecs psi).(0)
          (Mat.to_col_vecs eigenvectors).(0) )
      in
      if Parallel.master then
        Printf.printf "F12 Convergence : %e  %f\n" conv (eigenvalues.{state} 
          +. Simulation.nuclear_repulsion simulation);

      if conv > threshold then
        iteration ~state eigenvectors
      else
        let eigenvalues =
          Vec.map (fun x -> x +. ci.CI.e_shift) eigenvalues
        in
        eigenvectors, eigenvalues
    in
    iteration ~state ci_coef

  )
  in
  { mo_basis ; det_space ; ci ; hf12_integrals ; eigensystem }