More bugs with stratification fixed

datasail/cluster/clustering.py

@@ -153,18 +153,16 @@ def finish_clustering(dataset: DataSet) -> None:
     """
     # compute the weights and the stratification for the clusters
     dataset.cluster_weights = {}
-    if dataset.stratification is not None:
-        dataset.cluster_stratification = {}
+    dataset.cluster_stratification = {}
 
     for key, value in dataset.cluster_map.items():
         if value not in dataset.cluster_weights:
             dataset.cluster_weights[value] = 0
         dataset.cluster_weights[value] += dataset.weights[key]
 
-        if dataset.stratification is not None:
-            if value not in dataset.cluster_stratification:
-                dataset.cluster_stratification[value] = np.zeros(len(dataset.classes))
-            dataset.cluster_stratification[value] += dataset.strat2oh(name=key)
+        if value not in dataset.cluster_stratification:
+            dataset.cluster_stratification[value] = np.zeros(len(dataset.classes))
+        dataset.cluster_stratification[value] += dataset.strat2oh(name=key)
 
 
 def additional_clustering(

datasail/cluster/diamond.py

@@ -69,7 +69,5 @@ def run_diamond(dataset: DataSet, threads: int = 1, log_dir: Optional[Path] = No
     shutil.rmtree(result_folder, ignore_errors=True)
 
     dataset.cluster_names = table.index.tolist()
-    print(dataset.cluster_names)
     dataset.cluster_map = {n: n for n in dataset.cluster_names}
     dataset.cluster_similarity = table.to_numpy()
-    dataset.cluster_weights = {n: 1 for n in dataset.cluster_names}

datasail/cluster/vectors.py

@@ -149,8 +149,8 @@ def run_vector(dataset: DataSet, method: SIM_OPTIONS = "tanimoto") -> None:
     else:
         raise ValueError(f"Unknown method {method}")
     fps = [dataset.data[name] for name in dataset.names]
-    run(dataset, fps, method)
 
+    run(dataset, fps, method)
     dataset.cluster_names = copy.deepcopy(dataset.names)
     dataset.cluster_map = {n: n for n in dataset.names}
 

datasail/reader/utils.py

@@ -269,7 +269,7 @@ def read_data(
     Returns:
         A dataset storing all information on that datatype
     """
-    # parse the protein weights
+    # parse the weights
     if isinstance(weights, Path) and weights.is_file():
         if weights.suffix[1:].lower() == "csv":
             dataset.weights = dict((n, float(w)) for n, w in read_csv(weights, ","))
@@ -286,13 +286,31 @@ def read_data(
     elif inter is not None:
         dataset.weights = dict(count_inter(inter, index))
     else:
-        dataset.weights = dict((p, 1) for p in list(dataset.data.keys()))
+        dataset.weights = {k: 1 for k in dataset.data.keys()}
 
-    dataset.classes, dataset.stratification = read_stratification(strats)
+    # parse the stratification
+    if isinstance(strats, Path) and strats.is_file():
+        if strats.suffix[1:].lower() == "csv":
+            dataset.stratification = dict(read_csv(strats, ","))
+        elif strats.suffix[1:].lower() == "tsv":
+            dataset.stratification = dict(read_csv(strats, "\t"))
+        else:
+            raise ValueError()
+    elif isinstance(strats, dict):
+        dataset.stratification = strats
+    elif isinstance(strats, Callable):
+        dataset.stratification = strats()
+    elif isinstance(strats, Generator):
+        dataset.stratification = dict(strats)
+    else:
+        dataset.stratification = {k: 0 for k in dataset.data.keys()}
+
+    # .classes maps the individual classes to their index in one-hot encoding, important for non-numeric classes
+    dataset.classes = {s: i for i, s in enumerate(set(dataset.stratification.values()))}
     dataset.class_oh = np.eye(len(dataset.classes))
     dataset.num_clusters = num_clusters
 
-    # parse the protein similarity measure
+    # parse the similarity or distance measure
     if sim is None and dist is None:
         dataset.similarity, dataset.distance = get_default(dataset.type, dataset.format)
         dataset.names = list(dataset.data.keys())
@@ -312,37 +330,6 @@ def read_data(
     return dataset
 
 
-def read_stratification(strats: DATA_INPUT) -> Tuple[Dict[Any, int], Optional[Dict[str, np.ndarray]]]:
-    """
-    Read in the stratification for the data.
-
-    Args:
-        strats: Stratification input
-
-    Returns:
-        Set of all classes and a dictionary mapping the entity names to their class
-    """
-    # parse the stratification
-    if isinstance(strats, Path) and strats.is_file():
-        if strats.suffix[1:].lower() == "csv":
-            stratification = dict(read_csv(strats, ","))
-        elif strats.suffix[1:].lower() == "tsv":
-            stratification = dict(read_csv(strats, "\t"))
-        else:
-            raise ValueError()
-    elif isinstance(strats, dict):
-        stratification = strats
-    elif isinstance(strats, Callable):
-        stratification = strats()
-    elif isinstance(strats, Generator):
-        stratification = dict(strats)
-    else:
-        return {0: 0}, None
-
-    classes = {s: i for i, s in enumerate(set(stratification.values()))}
-    return classes, stratification
-
-
 def read_folder(folder_path: Path, file_extension: Optional[str] = None) -> Generator[Tuple[str, str], None, None]:
     """
     Read in all PDB file from a folder and ignore non-PDB files.

datasail/version.py

@@ -1 +1 @@
-__version__ = "1.0.3"
+__version__ = "1.0.4"

recipe/meta.yaml

@@ -1,6 +1,6 @@
 package:
   name: "datasail"
-  version: '1.0.3'
+  version: '1.0.4'
 
 source:
   path: ..

recipe_lite/meta.yaml

@@ -1,6 +1,6 @@
 package:
   name: "datasail-lite"
-  version: '1.0.3'
+  version: '1.0.4'
 
 source:
   path: ..

tests/test_caching.py

@@ -1,5 +1,4 @@
 import copy
-import os
 from pathlib import Path
 
 import pytest