Added an another engine for motus, using entropy instead of frequencies only

10 files changed, 121 insertions, 35 deletions

diff --git a/motus/bin/motus.ml b/motus/bin/motus.ml
index e32adca..a8a9188 100644
--- a/motus/bin/motus.ml
+++ b/motus/bin/motus.ml
@@ -69,11 +69,14 @@ let create_new_rules word result =
   !rules
 
 
-let rec run filters words =
+let rec run len filters words =
   let () = show_structure Format.std_formatter words filters in
+  (*
   let next =
     Freq_analysis.analyse words |> Freq_analysis.pick_next_word words
   in
+  *)
+  let _, next = Entropy.analyse len words in
 
   let () =
     Format.fprintf Format.std_formatter "Next word will be : %s@\n" next
@@ -83,14 +86,14 @@ let rec run filters words =
 
   (* if the input is empty, remove the word from the list and restart *)
   match String.equal String.empty input with
-  | true -> run filters (Wordlist.remove_word words next)
+  | true -> run len filters (Wordlist.remove_word words next)
   | false ->
       let new_rules =
         Criteria.merge_lists ~init:filters (create_new_rules next input)
         |> List.sort_uniq Stdlib.compare
       in
       let words = Wordlist.filter new_rules words in
-      run new_rules words
+      run len new_rules words
 
 
 let init_rule rules word =
@@ -119,4 +122,4 @@ let () =
   let words_channel = open_in (List.hd !dict) in
   let words = get_list words_channel initial_filter in
   close_in words_channel;
-  run initial_filter words
+  run !length initial_filter words
diff --git a/motus/js/initialize.ml b/motus/js/initialize.ml
index a088a0e..9d07a85 100644
--- a/motus/js/initialize.ml
+++ b/motus/js/initialize.ml
@@ -17,11 +17,17 @@ type t =
 
     The rule list is used to identify the letter already fixed by the previous
     results. *)
-let get_proposition : Wordlist.t -> Criteria.t list -> State.proposition option
-    =
- fun analysis rules ->
+let get_proposition :
+    int -> Wordlist.t -> Criteria.t list -> State.proposition option =
+ fun length wordlist rules ->
+  Printf.printf "Number of elements : %d\n" (Wordlist.list_size wordlist);
+
   let word =
-    Freq_analysis.analyse analysis |> Freq_analysis.pick_next_word analysis
+    if Wordlist.list_size wordlist > 2000
+    then Freq_analysis.analyse wordlist |> Freq_analysis.pick_next_word wordlist
+    else
+      let _, word = Entropy.analyse length wordlist in
+      word
   in
   match String.equal String.empty word with
   | true -> None
@@ -63,7 +69,7 @@ let process { sender; length; content; proposition } state =
         |> Wordlist.add_words rules
       in
 
-      let current_prop = get_proposition analysis rules
+      let current_prop = get_proposition length analysis rules
       and fields = FieldList.make length sender in
       ( match current_prop with
       | None -> state
diff --git a/motus/js/next.ml b/motus/js/next.ml
index 104b3e6..f3bb2fe 100644
--- a/motus/js/next.ml
+++ b/motus/js/next.ml
@@ -6,24 +6,24 @@ type t = unit
 
 let process : t -> State.state -> State.state =
  fun () state ->
-  (* Add the current proposition into the validated list *)
+  (* Get the news rules (and match the words only against the new ones )*)
   let rules = State.get_current_rules state.current_prop in
 
+  (* Update the word list with the new rules *)
+  let analysis = Motus_lib.Wordlist.filter rules state.analysis in
+
   let rules =
     Motus_lib.Criteria.merge_lists ~init:state.rules rules
     |> List.sort_uniq ~cmp:Stdlib.compare
   in
 
-  (* Update the word list with the new rules *)
-  let analysis = Motus_lib.Wordlist.filter rules state.analysis in
-
   let propositions = state.current_prop :: state.propositions
   and current_prop = [] in
 
   let new_state = { state with propositions; current_prop; rules; analysis } in
 
   (* Get the new proposition if any *)
-  let current_prop = Initialize.get_proposition analysis rules in
+  let current_prop = Initialize.get_proposition state.length analysis rules in
   match current_prop with
   | None -> new_state
   | Some prop ->
diff --git a/motus/js/reload.ml b/motus/js/reload.ml
index 2756b74..f0b581f 100644
--- a/motus/js/reload.ml
+++ b/motus/js/reload.ml
@@ -22,7 +22,7 @@ let process : t -> State.state -> State.state =
   in
   (* Get the new proposition if any *)
   let current_prop =
-    Initialize.get_proposition new_state.analysis state.rules
+    Initialize.get_proposition state.length new_state.analysis state.rules
   in
   match current_prop with
   | None -> new_state
diff --git a/motus/lib/entropy.ml b/motus/lib/entropy.ml
new file mode 100644
index 0000000..5b86a9d
--- /dev/null
+++ b/motus/lib/entropy.ml
@@ -0,0 +1,55 @@
+type t = float * string
+
+let get_entropy max_element words_number arr =
+  let entropy = ref 0. in
+  for idx = 0 to max_element - 1 do
+    let content = Float.of_int (Bigarray.Array1.get arr idx) in
+    if content > 0.
+    then
+      let ratio = content /. words_number in
+      entropy := !entropy -. (ratio *. Float.log2 ratio)
+  done;
+  entropy
+
+
+let analyse : int -> Wordlist.t -> t =
+ fun base words ->
+  let max_element = Float.to_int @@ (Validity.elements ** Float.of_int base) in
+  let words_number = Float.of_int (Wordlist.list_size words) in
+
+  match Wordlist.pick words with
+  | None -> (0., "")
+  | Some v ->
+      (* Build the array  *)
+      Seq.fold_left
+        (fun (score, word) word_ref ->
+          (* Reinitialize the array (we use the same in the successive
+             iterations *)
+          let set_ref = String.to_seq word_ref |> Validity.CharSet.of_seq in
+
+          let arr =
+            Bigarray.Array1.create Bigarray.Int Bigarray.C_layout max_element
+          in
+
+          Seq.iter
+            (fun w2 ->
+              let result = Validity.compare_words ~ref:(word_ref, set_ref) w2 in
+              match result with
+              | None -> ()
+              | Some r ->
+                  let idx = Validity.index_of_result r in
+
+                  let content = Bigarray.Array1.get arr idx in
+                  Bigarray.Array1.set arr idx (succ content) )
+            (Wordlist.words words);
+
+          (* Now evaluate the entropy in the array *)
+          let entropy = get_entropy max_element words_number arr in
+
+          if !entropy > score
+          then (
+            Printf.printf "Entropy for selecting %s : %.2f\n" word_ref !entropy;
+            (!entropy, word_ref) )
+          else (score, word) )
+        (0., v)
+        (Wordlist.words words)
diff --git a/motus/lib/validity.ml b/motus/lib/validity.ml
index ae85f04..0fdc40c 100644
--- a/motus/lib/validity.ml
+++ b/motus/lib/validity.ml
@@ -12,25 +12,28 @@ type t =
 
 [@@@warning "+32"]
 
-let m = Float.of_int (1 + max)
+let elements = Float.of_int (1 + max)
 
 (** Get the index of a validity result *)
 let index_of_result : t array -> int =
  fun elems ->
-  let _, value =
-    Array.fold_left elems ~init:(0., 0.) ~f:(fun (pos, acc) content ->
-        let v = Float.of_int (to_enum content) in
-        let acc' = acc +. (v *. (m ** pos)) in
-        (pos +. Float.one, acc') )
-  in
-  Float.to_int value
+  (* This seems to be more efficient than a foldLeft *)
+  let value = ref 0. in
+  Array.iteri
+    ~f:(fun pos content ->
+      let pos = Float.of_int pos in
+      let v = Float.of_int (to_enum content) in
+      let acc' = !value +. (v *. (elements ** pos)) in
+      value := acc' )
+    elems;
+  Float.to_int !value
 
 
 let index_to_result : base:int -> int -> t array =
  fun ~base n ->
   let rec _f acc n i =
-    let next = Float.round (n /. m)
-    and rem = Float.(to_int @@ rem n m) in
+    let next = Float.round (n /. elements)
+    and rem = Float.(to_int @@ rem n elements) in
 
     match (rem, i) with
     | _, 0 -> Array.of_list acc
@@ -45,29 +48,33 @@ let index_to_result : base:int -> int -> t array =
 (** Build a sequence of all the possible status for a given number of letters *)
 let sequence : int -> t array Seq.t =
  fun base ->
-  let max_element = Float.to_int @@ (m ** Float.of_int base) in
+  let max_element = Float.to_int @@ (elements ** Float.of_int base) in
 
   Seq.unfold
     (fun n ->
-      if n < max_element then Some (index_to_result ~base n, n + 1) else None )
+      if n < max_element then Some (index_to_result ~base n, succ n) else None
+      )
     0
 
 
-let compare_words : string -> ref:string -> t array option =
+module CharSet = Set.Make (Char)
+
+let compare_words : string -> ref:string * CharSet.t -> t array option =
  fun w1 ~ref ->
+  let wordRef = fst ref in
   let l1 = String.length w1 in
-  if l1 <> String.length ref
+  if l1 <> String.length wordRef
   then None
   else
     let result =
       Array.init l1 ~f:(fun i ->
           let c1 = String.get w1 i
-          and c2 = String.get ref i in
+          and c2 = String.get wordRef i in
 
           let state =
             if Char.equal c1 c2
             then Wellplaced
-            else if String.contains ref c1
+            else if CharSet.mem c2 (snd ref)
             then Misplaced
             else Missing
           in
@@ -92,7 +99,7 @@ let to_criterias : string -> t array -> Criteria.t list =
   let l, _ =
     Array.fold_left t ~init:([], 0) ~f:(fun (acc, i) t ->
         let acc = to_criteria (String.get word i) i t acc in
-        (acc, i + 1) )
+        (acc, succ i) )
   in
 
   List.rev l
diff --git a/motus/lib/validity.mli b/motus/lib/validity.mli
index dfd876c..a3d8ae3 100644
--- a/motus/lib/validity.mli
+++ b/motus/lib/validity.mli
@@ -1,8 +1,13 @@
+module CharSet : Set.S with type elt = char
+
 type t =
   | Wellplaced
   | Misplaced
   | Missing
 
+val elements : float
+(** Number of elements in the sum type *)
+
 val sequence : int -> t array Seq.t
 (** Build a sequence of all the possible status for a given number of letters *)
 
@@ -11,7 +16,7 @@ val index_of_result : t array -> int
 
 val index_to_result : base:int -> int -> t array
 
-val compare_words : string -> ref:string -> t array option
+val compare_words : string -> ref:string * CharSet.t -> t array option
 
 val to_criteria : char -> int -> t -> Criteria.t list -> Criteria.t list
 
diff --git a/motus/lib/wordlist.ml b/motus/lib/wordlist.ml
index 7c400bb..4a84ab0 100644
--- a/motus/lib/wordlist.ml
+++ b/motus/lib/wordlist.ml
@@ -23,3 +23,5 @@ let words = S.to_seq
 let list_size = S.cardinal
 
 let remove_word t w = S.remove w t
+
+let pick = S.choose_opt
diff --git a/motus/lib/wordlist.mli b/motus/lib/wordlist.mli
index a56cab3..86881f2 100644
--- a/motus/lib/wordlist.mli
+++ b/motus/lib/wordlist.mli
@@ -9,6 +9,8 @@ val empty_data : unit -> t
 val words : t -> string Seq.t
 (** Load all the words *)
 
+val pick : t -> string option
+
 val list_size : t -> int
 (** Number of words in the list *)
 
diff --git a/motus/test/motus_test.ml b/motus/test/motus_test.ml
index 4ea952e..0586ffe 100644
--- a/motus/test/motus_test.ml
+++ b/motus/test/motus_test.ml
@@ -37,14 +37,20 @@ let tests =
          assert_equal 243 count )
        ; ( "Compare words 1"
          >:: fun _ ->
+         let w = "Test" in
+
+         let ref = (w, Validity.CharSet.of_seq (String.to_seq w)) in
+
          assert_equal
            (Some Validity.[| Wellplaced; Wellplaced; Wellplaced; Wellplaced |])
-           (Validity.compare_words "Test" ~ref:"Test") )
+           (Validity.compare_words "Test" ~ref) )
        ; ( "Compare words 2"
          >:: fun _ ->
+         let w = "ABC" in
+         let ref = (w, Validity.CharSet.of_seq (String.to_seq w)) in
          assert_equal
            (Some Validity.[| Missing; Misplaced; Wellplaced |])
-           (Validity.compare_words "DAC" ~ref:"ABC") )
+           (Validity.compare_words "DAC" ~ref) )
        ]