Add shape check to Dataset initialization (#106)

* Check input array shapes when Dataset is initialized

* Add test for nimare.utils._check_inputs_shape

* Cover two missing lines by the test

* Add test when n or v is None

* Remove extra clause

.zenodo.json

@@ -10,6 +10,11 @@
       "affiliation": "Florida International University",
       "orcid": "0000-0001-9813-3167"
     },
+    {
+      "name": "Peraza, Julio A.",
+      "affiliation": "Florida International University",
+      "orcid": "0000-0003-3816-5903"
+    },
     {
       "name": "Nichols, Thomas E.",
       "affiliation": "Big Data Institute, University of Oxford",

pymare/core.py

@@ -5,7 +5,7 @@
 import numpy as np
 import pandas as pd
 
-from pymare.utils import _listify
+from pymare.utils import _check_inputs_shape, _listify
 
 from .estimators import (
     DerSimonianLaird,
@@ -94,6 +94,10 @@ def __init__(
         self.X = X
         self.X_names = names
 
+        _check_inputs_shape(self.y, self.X, "y", "X", row=True)
+        _check_inputs_shape(self.y, self.v, "y", "v", row=True, column=True)
+        _check_inputs_shape(self.y, self.n, "y", "n", row=True, column=True)
+
     def _get_predictors(self, X, names, add_intercept):
         if X is None and not add_intercept:
             raise ValueError(

pymare/tests/test_utils.py

@@ -1,10 +1,40 @@
 """Tests for pymare.utils."""
 import os.path as op
 
+import numpy as np
+import pytest
+
 from pymare import utils
 
 
 def test_get_resource_path():
     """Test nimare.utils.get_resource_path."""
     print(utils.get_resource_path())
     assert op.isdir(utils.get_resource_path())
+
+
+def test_check_inputs_shape():
+    """Test nimare.utils._check_inputs_shape."""
+    n_rows = 5
+    n_columns = 4
+    n_pred = 3
+    y = np.random.randint(1, 100, size=(n_rows, n_columns))
+    v = np.random.randint(1, 100, size=(n_rows + 1, n_columns))
+    n = np.random.randint(1, 100, size=(n_rows, n_columns))
+    X = np.random.randint(1, 100, size=(n_rows, n_pred))
+    X_names = [f"X{x}" for x in range(n_pred)]
+
+    utils._check_inputs_shape(y, X, "y", "X", row=True)
+    utils._check_inputs_shape(y, n, "y", "n", row=True, column=True)
+    utils._check_inputs_shape(X, np.array(X_names)[None, :], "X", "X_names", column=True)
+
+    # Raise error if the number of rows and columns of v don't match y
+    with pytest.raises(ValueError):
+        utils._check_inputs_shape(y, v, "y", "v", row=True, column=True)
+
+    # Raise error if neither row or column is True
+    with pytest.raises(ValueError):
+        utils._check_inputs_shape(y, n, "y", "n")
+
+    # Dataset may be initialized with n or v as None
+    utils._check_inputs_shape(y, None, "y", "n", row=True, column=True)

pymare/utils.py

@@ -19,3 +19,39 @@ def _listify(obj):
     This provides a simple way to accept flexible arguments.
     """
     return obj if isinstance(obj, (list, tuple, type(None), np.ndarray)) else [obj]
+
+
+def _check_inputs_shape(param1, param2, param1_name, param2_name, row=False, column=False):
+    """Check whether 'param1' and 'param2' have the same shape.
+
+    Parameters
+    ----------
+    param1 : array
+    param2 : array
+    param1_name : str
+    param2_name : str
+    row : bool, default to False.
+    column : bool, default to False.
+    """
+    if (param1 is not None) and (param2 is not None):
+        if row and not column:
+            shape1 = param1.shape[0]
+            shape2 = param2.shape[0]
+            message = "rows"
+        elif column and not row:
+            shape1 = param1.shape[1]
+            shape2 = param2.shape[1]
+            message = "columns"
+        elif row and column:
+            shape1 = param1.shape
+            shape2 = param2.shape
+            message = "rows and columns"
+        else:
+            raise ValueError("At least one of the two parameters (row or column) should be True.")
+
+        if shape1 != shape2:
+            raise ValueError(
+                f"{param1_name} and {param2_name} should have the same number of {message}. "
+                f"You provided {param1_name} with shape {param1.shape} and {param2_name} "
+                f"with shape {param2.shape}."
+            )