put each dataset variant in its own module

lib/convergence_detection.ml

@@ -13,7 +13,7 @@ type det_out =
   | Diffsel_out
 
 type det_result = {
-  dataset : dataset ;
+  dataset : Dataset.t ;
   det_meth : det_meth ;
   det_result : det_out directory workflow ;
   }

lib/convergence_detection.mli

@@ -13,7 +13,7 @@ type det_out =
   | Diffsel_out
 
 type det_result = {
-  dataset : dataset ;
+  dataset : Dataset.t ;
   det_meth : det_meth ;
   det_result : det_out directory workflow ;
   }

lib/dataset.ml

+open Core
+open Bistro_utils
+
+type t = {
+  model_prefix: string ;
+  tree_prefix : string ;
+  dataset : Ready_dataset.t ;
+}
+
+
+
+let repo ~preview dataset_l =
+  List.map dataset_l ~f:(fun dataset ->
+      let model_prefix = dataset.model_prefix in
+      let tree_prefix = dataset.tree_prefix in
+      let repo =
+        if preview then
+          Ready_dataset.repo dataset.dataset
+        else
+          Raw_dataset.repo (Ready_dataset.to_raw dataset.dataset)
+      in
+      Repo.shift tree_prefix (Repo.shift model_prefix repo)
+    )
+  |> List.concat

lib/defs.ml

@@ -10,29 +10,6 @@ type output_parse_input_tree =
   | Tree4simu
   | Tree_diffsel
 
-type raw_dataset = {
-  input_tree: nhx workflow ;
-  fna: nucleotide_fasta workflow ;
-  }
-
-type ready_dataset = {
-  input_tree: nhx workflow ;
-  tree_dataset : [`tree_dataset] directory workflow ;
-  fna: nucleotide_fasta workflow ;
-  faa: aminoacid_fasta workflow ;
-  }
-
-type ('a,'b) w_dataset =
-  | Raw_dataset of 'a
-  | Ready_dataset of 'b
-
-
-type dataset = {
-  model_prefix: string ;
-  tree_prefix : string ;
-  ready_dataset : ready_dataset
-  }
-
 type det_meth =
   | Pcoc
   | Pcoc_gamma

lib/pipeline.ml

@@ -6,55 +6,6 @@ open File_formats
 open Defs
 open Convergence_detection
 
-let ready_dataset_of_raw_dataset raw_dataset =
-    let input_tree = raw_dataset.input_tree in
-    let fna = raw_dataset.fna in
-    let tree_dataset = Tree_dataset.prepare input_tree in
-    let faa = Bppsuite.fna2faa ~fna in
-    { input_tree; tree_dataset ; fna; faa}
-
-let raw_dataset_of_ready_dataset ready_dataset =
-    let input_tree = ready_dataset.input_tree in
-    let fna = ready_dataset.fna in
-    { input_tree; fna}
-
-let repo_of_ready_dataset (rd : ready_dataset) =
-  Repo.[
-    item ["input_tree.nhx"] rd.input_tree ;
-    item ["tree.H0.node_ids" ] (Tree_dataset.nodes rd.tree_dataset H0) ;
-    item ["tree.Ha.node_ids" ] (Tree_dataset.nodes rd.tree_dataset Ha) ;
-    item ["tree.only_convergent_tags.nhx" ] (Tree_dataset.tree rd.tree_dataset `Detection) ;
-    item ["tree.only_node_ids.nhx" ] (Tree_dataset.tree rd.tree_dataset `Simulation) ;
-    item ["tree.diffsel" ] (Tree_dataset.diffsel_tree rd.tree_dataset) ;
-    item ["simulated_sequences.fna"] rd.fna ;
-    item ["simulated_sequences.faa"] rd.faa ;
-  ]
-  |> Repo.shift "ready_dataset"
-
-let repo_of_raw_dataset (raw_dataset:raw_dataset) =
-  Repo.[
-    item ["input_tree.nhx"] raw_dataset.input_tree ;
-    item ["simulated_sequences.fna"] raw_dataset.fna ;
-  ]
-  |> Repo.shift "raw_dataset"
-
-
-let repo_of_dataset_l ~preview dataset_l =
-  List.map dataset_l ~f:(fun (dataset : Defs.dataset) ->
-    let model_prefix = dataset.model_prefix in
-    let tree_prefix = dataset.tree_prefix in
-    let w_dataset =
-      if preview then
-        Ready_dataset dataset.ready_dataset
-      else Raw_dataset (raw_dataset_of_ready_dataset dataset.ready_dataset) in
-    let repo_d = match w_dataset with
-      | Ready_dataset d -> repo_of_ready_dataset d
-      | Raw_dataset d ->  repo_of_raw_dataset d
-    in
-      Repo.shift tree_prefix (Repo.shift model_prefix repo_d)
-    )
-    |> List.concat
-
 let derive_from_model ~model ~tree ~tree_dataset ~tree_prefix ~profile_f ~preview =
   let model_prefix = Convergence_hypothesis.string_of_model model in
   let nb_sites = if preview then 20 else 1000 in
@@ -70,8 +21,8 @@ let derive_from_model ~model ~tree ~tree_dataset ~tree_prefix ~profile_f ~previe
   let fna = Bppsuite.bppseqgen_multi_profiles ~descr ~nb_sites ~tree ~config:config_p ~profile_f in
 
   let faa = Bppsuite.fna2faa ~fna in
-  let ready_dataset = { input_tree = tree ; tree_dataset ; fna; faa} in
-    { model_prefix; tree_prefix; ready_dataset }
+  let ready_dataset = { Ready_dataset.input_tree = tree ; tree_dataset ; fna; faa} in
+  { Dataset.model_prefix; tree_prefix; dataset = ready_dataset }
 
 let derive_from_tree ~tree_dir ~tree ~profile_f ~preview =
   let tree_prefix = Filename.chop_extension tree in
@@ -126,12 +77,12 @@ let repo_of_dataset_results_l ~dataset_results_l =
   )
   |> List.concat
 
-let derive_from_det_meth ~det_meth ~dataset ~preview =
-  let faa = dataset.ready_dataset.faa in
-  let fna = dataset.ready_dataset.fna in
+let derive_from_det_meth ~det_meth ~(dataset : Dataset.t) ~preview =
+  let faa = dataset.dataset.faa in
+  let fna = dataset.dataset.fna in
   let phy_n = Bppsuite.fa2phy ~fna in
-  let pcoc_tree = Tree_dataset.tree dataset.ready_dataset.tree_dataset `Detection in
-  let diffsel_tree = Tree_dataset.diffsel_tree dataset.ready_dataset.tree_dataset in
+  let pcoc_tree = Tree_dataset.tree dataset.dataset.tree_dataset `Detection in
+  let diffsel_tree = Tree_dataset.diffsel_tree dataset.dataset.tree_dataset in
   let w_every = if preview then 1 else 10 in
   let n_cycles = if preview then 100 else 1000 in
   let det_result = match det_meth with
@@ -151,8 +102,8 @@ let derive_from_dataset ~dataset ~preview =
     ) in
   let merged_results = merge_results ~res_by_tools in
   let tsv = merged_results in
-  let faa = dataset.ready_dataset.faa in
-  let tree = Tree_dataset.tree dataset.ready_dataset.tree_dataset `Detection in
+  let faa = dataset.dataset.faa in
+  let tree = Tree_dataset.tree dataset.dataset.tree_dataset `Detection in
   let plot_merged_results = plot_merge_results ~res_by_tools ~tsv ~faa ~tree in
   let model_prefix = dataset.model_prefix in
   let tree_prefix = dataset.tree_prefix in
@@ -172,7 +123,7 @@ let main ~outdir ?(np = 2) ?(mem = 2) ~tree_dir ~profile_fn ~preview () =
   let dataset_l = derive_sim ~tree_dir ~trees ~profile_fn ~preview in
   let dataset_results_l = derive_det ~dataset_l ~profile_fn ~preview in
   let repo = [
-    repo_of_dataset_l dataset_l ~preview  ;
+    Dataset.repo dataset_l ~preview  ;
     repo_of_dataset_results_l ~dataset_results_l;
   ]
   |> List.concat

lib/raw_dataset.ml

+open Bistro.Std
+open File_formats
+open Bistro_utils
+
+type t = {
+  input_tree: nhx workflow ;
+  fna: nucleotide_fasta workflow ;
+}
+
+let repo rd =
+  Repo.[
+    item ["input_tree.nhx"] rd.input_tree ;
+    item ["simulated_sequences.fna"] rd.fna ;
+  ]
+  |> Repo.shift "raw_dataset"

lib/ready_dataset.ml

+open Bistro.Std
+open Bistro_utils
+open File_formats
+
+type t = {
+  input_tree: nhx workflow ;
+  tree_dataset : [`tree_dataset] directory workflow ;
+  fna: nucleotide_fasta workflow ;
+  faa: aminoacid_fasta workflow ;
+}
+
+
+let of_raw (raw_dataset : Raw_dataset.t) =
+  let input_tree = raw_dataset.input_tree in
+  let fna = raw_dataset.fna in
+  let tree_dataset = Tree_dataset.prepare input_tree in
+  let faa = Bppsuite.fna2faa ~fna in
+  { input_tree; tree_dataset ; fna; faa}
+
+let repo rd =
+  Repo.[
+    item ["input_tree.nhx"] rd.input_tree ;
+    item ["tree.H0.node_ids" ] (Tree_dataset.nodes rd.tree_dataset H0) ;
+    item ["tree.Ha.node_ids" ] (Tree_dataset.nodes rd.tree_dataset Ha) ;
+    item ["tree.only_convergent_tags.nhx" ] (Tree_dataset.tree rd.tree_dataset `Detection) ;
+    item ["tree.only_node_ids.nhx" ] (Tree_dataset.tree rd.tree_dataset `Simulation) ;
+    item ["tree.diffsel" ] (Tree_dataset.diffsel_tree rd.tree_dataset) ;
+    item ["simulated_sequences.fna"] rd.fna ;
+    item ["simulated_sequences.faa"] rd.faa ;
+  ]
+  |> Repo.shift "ready_dataset"
+
+let to_raw { input_tree ; fna } =
+  { Raw_dataset.input_tree ; fna }