Address mypy (#276)

* [ci skip] Address most mypy issues * Remove data-science-types req, numpy generics * Fix minor pet peeves
scverse · Feb 15, 2021 · 96e88bd · 96e88bd
1 parent 84ee68e
commit 96e88bd
Show file tree

Hide file tree

Showing 18 changed files with 111 additions and 95 deletions.
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
@@ -60,7 +60,7 @@ jobs:
     test:
         needs: init
         if: ${{ github.event_name == 'schedule' || needs.init.outputs.skip == 'false' }}
-        timeout-minutes: 10
+        timeout-minutes: 20
         runs-on: ${{ matrix.os }}
         strategy:
             fail-fast: false

diff --git a/.mypy.ini b/.mypy.ini
@@ -1,6 +1,6 @@
 [mypy]
 mypy_path = squidpy
-python_version = 3.7
+python_version = 3.8
 
 warn_redundant_casts = True
 warn_unused_configs = True

diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
@@ -10,7 +10,7 @@ repos:
     rev: v0.800
     hooks:
     -   id: mypy
-        additional_dependencies: [numpy>=1.15.0, scipy>=1.5.0, pandas]
+        additional_dependencies: [numpy, scipy, pandas]
 -   repo: https://github.com/psf/black
     rev: 20.8b1
     hooks:

diff --git a/requirements.txt b/requirements.txt
@@ -1,6 +1,5 @@
 anndata>=0.7.4
 dask>=2.30
-data-science-types>=0.2.20  # array typehints
 docrep>=0.3.1
 ipywidgets>=7.5.1  # progress bar
 omnipath>=1.0.4

diff --git a/squidpy/_utils.py b/squidpy/_utils.py
@@ -28,11 +28,11 @@
 
 
 try:
-    from functools import singledispatchmethod  # type: ignore[attr-defined]
+    from functools import singledispatchmethod
 except ImportError:
     from functools import singledispatch, update_wrapper
 
-    def singledispatchmethod(func: Callable[..., Any]) -> Callable[..., Any]:
+    def singledispatchmethod(func: Callable[..., Any]) -> Callable[..., Any]:  # type: ignore[no-redef]
         """Backport of `singledispatchmethod` for < Python 3.8."""
         dispatcher = singledispatch(func)
 

diff --git a/squidpy/gr/_build.py b/squidpy/gr/_build.py
@@ -5,7 +5,8 @@
 from scanpy import logging as logg
 from anndata import AnnData
 
-from scipy.sparse import spmatrix, csr_matrix, SparseEfficiencyWarning
+from numba import njit
+from scipy.sparse import spmatrix, csr_matrix, isspmatrix_csr, SparseEfficiencyWarning
 from sklearn.neighbors import NearestNeighbors
 from sklearn.metrics.pairwise import cosine_similarity
 import numpy as np
@@ -114,11 +115,12 @@ def spatial_neighbors(
     if transform == Transform.SPECTRAL:
         Adj = _transform_a_spectral(Adj)
     elif transform == Transform.COSINE:
+        print("foo")
         Adj = _transform_a_cosine(Adj)
     elif transform == Transform.NONE:
         pass
     else:
-        raise NotImplementedError(transform)
+        raise NotImplementedError(f"Transform `{transform}` is not yet implemented.")
 
     neighs_key = Key.uns.spatial_neighs(key_added)
     conns_key = Key.obsp.spatial_conn(key_added)
@@ -181,10 +183,28 @@ def _build_connectivity(
     return (conns_m, dists_m) if return_distance else conns_m
 
 
-def _transform_a_spectral(a: Union[csr_matrix, np.ndarray]) -> np.ndarray:
+@njit
+def outer(indices: np.ndarray, indptr: np.ndarray, degrees: np.ndarray) -> np.ndarray:
+    res = np.empty_like(indices, dtype=np.float64)
+    start = 0
+    for i in range(len(indptr) - 1):
+        ixs = indices[indptr[i] : indptr[i + 1]]
+        res[start : start + len(ixs)] = degrees[i] * degrees[ixs]
+        start += len(ixs)
+
+    return res
+
+
+def _transform_a_spectral(a: spmatrix) -> spmatrix:
+    if not isspmatrix_csr(a):
+        a = a.tocsr()
     degrees = np.squeeze(np.array(np.sqrt(1.0 / a.sum(axis=0))))
-    return a.multiply(np.outer(degrees, degrees))
+
+    a = a.multiply(outer(a.indices, a.indptr, degrees))
+    a.eliminate_zeros()
+
+    return a
 
 
-def _transform_a_cosine(a: Union[spmatrix, np.ndarray]) -> Union[np.ndarray, spmatrix]:
+def _transform_a_cosine(a: spmatrix) -> spmatrix:
     return cosine_similarity(a, dense_output=False)
diff --git a/squidpy/gr/_ligrec.py b/squidpy/gr/_ligrec.py
@@ -659,8 +659,8 @@ def ligrec(
 @d.dedent
 def _analysis(
     data: pd.DataFrame,
-    interactions: np.ndarray[np.uint32],
-    interaction_clusters: np.ndarray[np.uint32],
+    interactions: np.ndarray,
+    interaction_clusters: np.ndarray,
     threshold: float = 0.1,
     n_perms: int = 1000,
     seed: Optional[int] = None,
@@ -703,13 +703,13 @@ def _analysis(
     def extractor(res: Sequence[TempResult]) -> TempResult:
         assert len(res) == n_jobs, f"Expected to find `{n_jobs}` results, found `{len(res)}`."
 
-        means: List[np.ndarray] = [r.means for r in res if r.means is not None]
-        assert len(means) == 1, f"Only `1` job should've calculated the means, but found `{len(means)}`."
-        means = means[0]
+        meanss: List[np.ndarray] = [r.means for r in res if r.means is not None]
+        assert len(meanss) == 1, f"Only `1` job should've calculated the means, but found `{len(meanss)}`."
+        means = meanss[0]
         if TYPE_CHECKING:
             assert isinstance(means, np.ndarray)
 
-        pvalues: np.ndarray = np.sum([r.pvalues for r in res if r.pvalues is not None], axis=0) / float(n_perms)
+        pvalues = np.sum([r.pvalues for r in res if r.pvalues is not None], axis=0) / float(n_perms)
         assert means.shape == pvalues.shape, f"Means and p-values differ in shape: `{means.shape}`, `{pvalues.shape}`."
 
         return TempResult(means=means, pvalues=pvalues)
@@ -743,13 +743,13 @@ def extractor(res: Sequence[TempResult]) -> TempResult:
 
 
 def _analysis_helper(
-    perms: np.ndarray[np.uint32],
-    data: np.ndarray[np.float64],
-    mean: np.ndarray[np.float64],
-    mask: np.ndarray[np.bool_],
-    interactions: np.ndarray[np.uint32],
-    interaction_clusters: np.ndarray[np.uint32],
-    clustering: np.ndarray[np.uint32],
+    perms: np.ndarray,
+    data: np.ndarray,
+    mean: np.ndarray,
+    mask: np.ndarray,
+    interactions: np.ndarray,
+    interaction_clusters: np.ndarray,
+    clustering: np.ndarray,
     seed: Optional[int] = None,
     numba_parallel: Optional[bool] = None,
     queue: Optional[SigQueue] = None,
@@ -811,7 +811,7 @@ def _analysis_helper(
     _test = globals()[fn_key]
 
     if return_means:
-        res_means = np.zeros((len(interactions), len(interaction_clusters)), dtype=np.float64)
+        res_means: Optional[np.ndarray] = np.zeros((len(interactions), len(interaction_clusters)), dtype=np.float64)
         test = partial(_test, res_means=res_means)
     else:
         res_means = None

diff --git a/squidpy/gr/_ppatterns.py b/squidpy/gr/_ppatterns.py
@@ -1,23 +1,7 @@
 """Functions for point patterns spatial statistics."""
-from __future__ import annotations
-
-from squidpy.gr._utils import (
-    _save_data,
-    _assert_positive,
-    _assert_spatial_basis,
-    _assert_categorical_obs,
-    _assert_connectivity_key,
-    _assert_non_empty_sequence,
-)
-
-try:
-    # [ py<3.8 ]
-    from typing import Literal  # type: ignore[attr-defined]
-except ImportError:
-    from typing_extensions import Literal
-
 from typing import Tuple, Union, Iterable, Optional, Sequence
 from itertools import chain
+from typing_extensions import Literal  # < 3.8
 import warnings
 
 from scanpy import logging as logg
@@ -33,6 +17,14 @@
 
 from squidpy._docs import d, inject_docs
 from squidpy._utils import Signal, SigQueue, parallelize, _get_n_cores
+from squidpy.gr._utils import (
+    _save_data,
+    _assert_positive,
+    _assert_spatial_basis,
+    _assert_categorical_obs,
+    _assert_connectivity_key,
+    _assert_non_empty_sequence,
+)
 from squidpy._constants._pkg_constants import Key
 
 try:
@@ -145,7 +137,7 @@ def moran(
     adata: AnnData,
     connectivity_key: str = Key.obsp.spatial_conn(),
     genes: Optional[Union[str, Sequence[str]]] = None,
-    transformation: Literal["r", "B", "D", "U", "V"] = "r",  # type: ignore[name-defined]
+    transformation: Literal["r", "B", "D", "U", "V"] = "r",
     n_perms: int = 1000,
     corr_method: Optional[str] = "fdr_bh",
     layer: Optional[str] = None,
@@ -236,7 +228,7 @@ def _moran_helper(
     gen: Iterable[str],
     adata: AnnData,
     weights: W,
-    transformation: Literal["r", "B", "D", "U", "V"] = "B",  # type: ignore[name-defined]
+    transformation: Literal["r", "B", "D", "U", "V"] = "r",
     permutations: int = 1000,
     layer: Optional[str] = None,
     seed: Optional[int] = None,
@@ -281,11 +273,11 @@ def _set_weight_class(adata: AnnData, key: str) -> W:
     fastmath=True,
 )
 def _occur_count(
-    clust: Tuple[np.ndarray[np.int32], np.ndarray[np.int32]],
-    pw_dist: np.ndarray[np.float32],
-    labs_unique: np.ndarray[np.int32],
-    interval: np.ndarray[np.float32],
-) -> np.ndarray[np.float32]:
+    clust: Tuple[np.ndarray, np.ndarray],
+    pw_dist: np.ndarray,
+    labs_unique: np.ndarray,
+    interval: np.ndarray,
+) -> np.ndarray:
     num = labs_unique.shape[0]
     out = np.zeros((num, num, interval.shape[0] - 1), dtype=ft)
 

diff --git a/squidpy/gr/_utils.py b/squidpy/gr/_utils.py
@@ -1,5 +1,15 @@
 """Graph utilities."""
-from typing import Any, List, Tuple, Union, Hashable, Iterable, Optional, Sequence
+from typing import (
+    Any,
+    List,
+    Tuple,
+    Union,
+    Hashable,
+    Iterable,
+    Optional,
+    Sequence,
+    TYPE_CHECKING,
+)
 
 from scanpy import logging as logg
 from anndata import AnnData
@@ -71,7 +81,6 @@ def _create_sparse_df(
     -------
     Each column of the DataFrame is stored as a :class:`arrays.SparseArray`.
     """
-    from pandas import DataFrame
     from pandas._libs.sparse import IntIndex
     from pandas.core.arrays.sparse.accessor import (
         SparseArray,
@@ -83,6 +92,8 @@ def _create_sparse_df(
         data = csc_matrix(data)
         sort_indices = False
     else:
+        if TYPE_CHECKING:
+            assert isinstance(data, spmatrix)
         data = data.tocsc()
         sort_indices = True
 
@@ -108,7 +119,7 @@ def _create_sparse_df(
         arr = SparseArray._simple_new(array_data[sl], idx, dtype)
         arrays.append(arr)
 
-    return DataFrame._from_arrays(arrays, columns=columns, index=index, verify_integrity=False)
+    return pd.DataFrame._from_arrays(arrays, columns=columns, index=index, verify_integrity=False)
 
 
 def _assert_categorical_obs(adata: AnnData, key: str) -> None:

diff --git a/squidpy/im/_container.py b/squidpy/im/_container.py
@@ -16,15 +16,25 @@
     Sequence,
     TYPE_CHECKING,
 )
+from pathlib import Path
 from itertools import chain
+from typing_extensions import Literal
 import re
 
+from scanpy import logging as logg
+from anndata import AnnData
+
+import numpy as np
+import xarray as xr
+
 import matplotlib as mpl
 import matplotlib.pyplot as plt
 
+from imageio import imread
 from skimage.util import img_as_float
 from skimage.transform import rescale
 
+from squidpy._docs import d
 from squidpy._utils import singledispatchmethod
 from squidpy.gr._utils import (
     _assert_in_range,
@@ -33,23 +43,6 @@
     _assert_spatial_basis,
     _assert_non_empty_sequence,
 )
-
-try:
-    from typing import Literal  # type: ignore[attr-defined]
-except ImportError:
-    from typing_extensions import Literal
-
-from pathlib import Path
-
-from scanpy import logging as logg
-from anndata import AnnData
-
-import numpy as np
-import xarray as xr
-
-from imageio import imread
-
-from squidpy._docs import d
 from squidpy.im._utils import (
     _num_pages,
     CropCoords,
@@ -830,10 +823,10 @@ def apply(
             res = res[..., np.newaxis]
 
         if copy:
-            res = ImageContainer(res, layer=layer, channel_dim=channel_dim)
-            res.data.attrs = self.data.attrs.copy()
+            cont = ImageContainer(res, layer=layer, channel_dim=channel_dim)
+            cont.data.attrs = self.data.attrs.copy()
 
-            return res  # type: ignore[no-any-return]
+            return cont
 
         self.add_img(
             res,

diff --git a/squidpy/im/_feature_mixin.py b/squidpy/im/_feature_mixin.py
@@ -337,7 +337,7 @@ def convert_to_full_image_coordinates(x: np.ndarray, y: np.ndarray) -> np.ndarra
             y = coord.slice[0].start + (y_slc.stop - y_slc.start) * y
             x = coord.slice[1].start + (x_slc.stop - x_slc.start) * x
 
-            return np.c_[x, y]
+            return np.c_[x, y]  # type: ignore[no-any-return]
 
         label_layer = self._get_layer(label_layer)
 
@@ -356,7 +356,7 @@ def convert_to_full_image_coordinates(x: np.ndarray, y: np.ndarray) -> np.ndarra
         else:
             channels = ()
 
-        features = {}
+        features: Dict[str, Any] = {}
         # calculate features that do not depend on the intensity image
         tmp_features = skimage.measure.regionprops_table(
             self[label_layer].values[:, :, 0], properties=no_intensity_props

diff --git a/squidpy/im/_segment.py b/squidpy/im/_segment.py
@@ -143,7 +143,7 @@ def _segment(self, arr: np.ndarray, thresh: Optional[float] = None, geq: bool =
 
         markers, _ = ndi.label(local_maxi)
 
-        return watershed(arr, markers, mask=mask)
+        return np.asarray(watershed(arr, markers, mask=mask))
 
 
 class SegmentationCustom(SegmentationModel):
@@ -163,7 +163,7 @@ def __init__(self, func: Callable[..., np.ndarray]):
         super().__init__(model=func)
 
     def _segment(self, arr: np.ndarray, **kwargs: Any) -> np.ndarray:
-        return self._model(arr, **kwargs)
+        return np.asarray(self._model(arr, **kwargs))
 
     def __repr__(self) -> str:
         return f"{self.__class__.__name__}[function={getattr(self._model, '__name__', None)}]"

diff --git a/squidpy/im/_utils.py b/squidpy/im/_utils.py
@@ -13,8 +13,7 @@
 
 def _circular_mask(arr: np.ndarray, y: int, x: int, radius: float) -> np.ndarray:
     Y, X = np.ogrid[: arr.shape[0], : arr.shape[1]]
-
-    return ((Y - y) ** 2 + (X - x) ** 2) <= radius ** 2
+    return np.asarray(((Y - y) ** 2 + (X - x) ** 2) <= radius ** 2)
 
 
 def _num_pages(fname: str) -> int: