Merge pull request #676 from xgi-org/speedup-imports

Speedup imports

Author: kaiser-dan

HOW_TO_CONTRIBUTE.md

@@ -24,6 +24,14 @@ Please note we have a [code of conduct](/CODE_OF_CONDUCT.md), please follow it i
    2. notebooks by running `nbqa isort .` to sort any new import statements in the source code and tutorials.
 4. Format the source code and notebooks by running `black .` for consistent styling.
 
+## Profile importing XGI
+
+Install the `test` dependencis and visualize the import times using [tuna](https://github.com/nschloe/tuna):
+```python
+python3 -X importtime -c "import xgi" 2> test.log
+tuna test.log
+```
+
 ## New Version process
 
 1. Make sure that the Github Actions workflow runs without any errors.

examples/advanced/plot_colormaps_colorbars.py

@@ -24,7 +24,7 @@
     node_fc=H.nodes.degree,
     edge_fc=H.edges.size,
     edge_fc_cmap="viridis",
-    node_fc_cmap="mako_r",
+    node_fc_cmap="plasma",
 )
 
 node_col, _, edge_col = collections

examples/stats/plot_color_stat.py

@@ -20,7 +20,7 @@
     pos=pos,
     node_fc=H.nodes.degree,
     edge_fc="grey",
-    node_fc_cmap="mako_r",
+    node_fc_cmap="plasma",
 )
 
 node_col, _, edge_col = collections

pyproject.toml

@@ -48,7 +48,6 @@ dependencies = [
     "networkx>=2.7",
     "requests>=2.0",
     "matplotlib>=3.4",
-    "seaborn>=0.10",
 ]
 
 [project.urls]
@@ -91,10 +90,12 @@ release = [
 test = [
     "pytest>=7.2",
     "pytest-cov>=4.0",
+    "tuna",
 ]
 tutorial = [
     "jupyter>=1.0",
     "ipython<=8.12.0",
+    "seaborn>=0.10"
 ]
 all = ["xgi[benchmark,developer,docs,release,test,tutorial]"]
 

requirements/default.txt

@@ -6,4 +6,3 @@ pandas>=1.3
 networkx>=2.7
 requests>=2.0
 matplotlib>=3.4
-seaborn>=0.10

requirements/test.txt

@@ -2,3 +2,4 @@
 # Do not edit this file; modify pyproject.toml instead.
 pytest>=7.2
 pytest-cov>=4.0
+tuna

tests/drawing/test_draw.py

@@ -3,7 +3,6 @@
 import matplotlib.pyplot as plt
 import numpy as np
 import pytest
-import seaborn as sb
 
 import xgi
 from xgi.exception import XGIError
@@ -295,7 +294,7 @@ def test_draw_hyperedges_fc_cmap(edgelist8):
     ax, collections = xgi.draw_hyperedges(H, ax=ax)
     (dyad_collection, edge_collection) = collections
     assert dyad_collection.get_cmap() == plt.cm.Greys
-    assert edge_collection.get_cmap() == sb.color_palette("crest_r", as_cmap=True)
+    assert edge_collection.get_cmap() == xgi.crest_r()
     plt.close("all")
 
     # set cmap

tutorials/focus/Tutorial 5 - Plotting.ipynb

@@ -387,7 +387,7 @@
     "    node_fc=H.nodes.degree,\n",
     "    edge_fc=H.edges.size,\n",
     "    edge_fc_cmap=\"viridis\",\n",
-    "    node_fc_cmap=\"mako_r\",\n",
+    "    node_fc_cmap=\"plasma\",\n",
     ")\n",
     "\n",
     "node_col, _, edge_col = collections\n",

xgi/drawing/draw.py

@@ -6,7 +6,6 @@
 import matplotlib as mpl
 import matplotlib.pyplot as plt
 import numpy as np
-import seaborn as sb  # This cannot be removed because it is used for cmap "crest"
 from matplotlib import cm
 from matplotlib.colors import is_color_like
 from matplotlib.patches import FancyArrowPatch
@@ -25,7 +24,7 @@
 from ..convert import to_bipartite_edgelist
 from ..core import DiHypergraph, Hypergraph, SimplicialComplex
 from ..exception import XGIError
-from ..utils import subfaces
+from ..utils import crest_r, subfaces
 from .draw_utils import (
     _CCW_sort,
     _draw_arg_to_arr,
@@ -70,7 +69,7 @@ def draw(
     dyad_vmin=None,
     dyad_vmax=None,
     edge_fc=None,
-    edge_fc_cmap="crest_r",
+    edge_fc_cmap=crest_r(),
     edge_vmin=None,
     edge_vmax=None,
     edge_ec=None,
@@ -298,7 +297,6 @@ def draw(
         )
 
     elif isinstance(H, Hypergraph):
-
         ax, (dyad_collection, edge_collection) = draw_hyperedges(
             H=H,
             pos=pos,
@@ -510,8 +508,8 @@ def draw_nodes(
 
     # plot
     node_collection = ax.scatter(
-        x=xy[:, 0],
-        y=xy[:, 1],
+        xy[:, 0],
+        xy[:, 1],
         s=node_size,
         marker=node_shape,
         c=node_fc,
@@ -552,7 +550,7 @@ def draw_hyperedges(
     dyad_vmin=None,
     dyad_vmax=None,
     edge_fc=None,
-    edge_fc_cmap="crest_r",
+    edge_fc_cmap=crest_r(),
     edge_vmin=None,
     edge_vmax=None,
     edge_ec=None,
@@ -614,7 +612,7 @@ def draw_hyperedges(
 
         If None (default), color by edge size.
     edge_fc_cmap: matplotlib colormap
-        Colormap used to map the edge colors. By default, "crest_r".
+        Colormap used to map the edge colors. By default, crest_r().
     edge_vmin, edge_vmax : float, optional
         Minimum and maximum for edge colormap scaling. By default, None.
     edge_ec : color or list of colors or array-like or dict or EdgeStat, optional
@@ -775,7 +773,6 @@ def draw_hyperedges(
     edge_ec = edge_ec[ids_sorted] if len(edge_ec) > 1 else edge_ec  # reorder
 
     if edge_ec_to_map:  # edgecolors need to be manually mapped
-
         # create scalarmappable to map floats to colors
         # we use the same vmin, vmax, and cmap as for edge_fc
         norm = mpl.colors.Normalize(vmin=edge_vmin, vmax=edge_vmax)
@@ -849,7 +846,7 @@ def draw_simplices(
     dyad_vmin=None,
     dyad_vmax=None,
     edge_fc=None,
-    edge_fc_cmap="crest_r",
+    edge_fc_cmap=crest_r(),
     edge_vmin=None,
     edge_vmax=None,
     alpha=0.4,
@@ -908,7 +905,7 @@ def draw_simplices(
 
         If None (default), color by edge size.
     edge_fc_cmap: matplotlib colormap
-        Colormap used to map the edge colors. By default, "crest_r".
+        Colormap used to map the edge colors. By default, crest_r().
     edge_vmin, edge_vmax : float, optional
         Minimum and maximum for edge colormap scaling. By default, None.
     alpha : float, optional
@@ -1259,12 +1256,12 @@ def draw_multilayer(
     dyad_lw=1.5,
     dyad_style="solid",
     edge_fc=None,
-    edge_fc_cmap="crest_r",
+    edge_fc_cmap=crest_r(),
     edge_vmin=None,
     edge_vmax=None,
     alpha=0.4,
     layer_color="grey",
-    layer_cmap="crest_r",
+    layer_cmap=crest_r(),
     max_order=None,
     conn_lines=True,
     conn_lines_style="dotted",
@@ -1347,7 +1344,7 @@ def draw_multilayer(
 
         If None (default), color by edge size.
     edge_fc_cmap: matplotlib colormap, optional
-        Colormap used to map the edge colors. By default, "crest_r".
+        Colormap used to map the edge colors. By default, crest_r().
     edge_vmin, edge_vmax : float, optional
         Minimum and maximum for edge colormap scaling. By default, None.
     alpha : float, optional
@@ -1356,7 +1353,7 @@ def draw_multilayer(
         Color of layers. By default, "grey".
     layer_cmap : colormap, optional
         Colormap to use in case of numerical values in layer_color. Ignored if layer_color
-        does not contain numerical values to be mapped. By default, "crest_r", but ignored.
+        does not contain numerical values to be mapped. By default, crest_r(), but ignored.
     max_order : int, optional
         Maximum of hyperedges to plot. If None (default), plots all orders.
     conn_lines : bool, optional
@@ -1485,7 +1482,6 @@ def draw_multilayer(
 
     # plot layers
     for jj, d in enumerate(orders):
-
         z = [sep * d] * H.num_nodes
 
         # draw surfaces corresponding to the different orders
@@ -1585,7 +1581,6 @@ def draw_multilayer(
 
     # draw nodes (last)
     for d in orders:
-
         z = [sep * d] * H.num_nodes
 
         node_collection = ax.scatter(
@@ -1629,12 +1624,12 @@ def draw_bipartite(
     edge_marker_lw=1,
     edge_marker_size=7,
     edge_marker_shape="s",
-    edge_marker_fc_cmap="crest_r",
+    edge_marker_fc_cmap=crest_r(),
     max_order=None,
     dyad_color=None,
     dyad_lw=1,
     dyad_style="solid",
-    dyad_color_cmap="crest_r",
+    dyad_color_cmap=crest_r(),
     node_labels=None,
     hyperedge_labels=None,
     arrowsize=10,
@@ -1952,7 +1947,7 @@ def draw_undirected_dyads(
     dyad_color=None,
     dyad_lw=1,
     dyad_style="solid",
-    dyad_color_cmap="crest_r",
+    dyad_color_cmap=crest_r(),
     rescale_sizes=True,
     **kwargs,
 ):
@@ -2114,7 +2109,7 @@ def draw_directed_dyads(
     dyad_color=None,
     dyad_lw=1,
     dyad_style="solid",
-    dyad_color_cmap="crest_r",
+    dyad_color_cmap=crest_r(),
     arrowsize=10,
     arrowstyle="->",
     connectionstyle="arc3",
@@ -2336,7 +2331,6 @@ def to_marker_edge(marker_size, marker):
                 ems = edge_marker_size
 
             for n in tail:  # lines going towards the center
-
                 xy_source = node_pos[n]
                 xy_target = edge_pos[e]
 

xgi/utils/tensor.py

@@ -1,13 +1,9 @@
 ## Tensor times same vector in all but one (TTSV1) and all but two (TTSV2)
+import math
 from collections import defaultdict
 from itertools import combinations
-from math import factorial
 
 import numpy as np
-from numpy import prod
-from scipy.signal import convolve
-from scipy.sparse import coo_array
-from scipy.special import binom as binomial
 
 __all__ = [
     "pairwise_incidence",
@@ -78,9 +74,10 @@ def banerjee_coeff(size, max_size):
     Sinan Aksoy, Ilya Amburg, Stephen Young,
     https://doi.org/10.1137/23M1584472
     """
+    from scipy.special import binom
+
     return sum(
-        ((-1) ** j) * binomial(size, j) * (size - j) ** max_size
-        for j in range(size + 1)
+        ((-1) ** j) * binom(size, j) * (size - j) ** max_size for j in range(size + 1)
     )
 
 
@@ -119,15 +116,15 @@ def ttsv1(node_dict, edge_dict, r, a):
     """
     n = len(node_dict)
     s = np.zeros(n)
-    r_minus_1_factorial = factorial(r - 1)
+    r_minus_1_factorial = math.factorial(r - 1)
     for node, edges in node_dict.items():
         c = 0
         for e in edges:
             l = len(edge_dict[e])
             alpha = banerjee_coeff(l, r)
             edge_without_node = [v for v in edge_dict[e] if v != node]
             if l == r:
-                gen_fun_coef = prod(a[edge_without_node])
+                gen_fun_coef = np.prod(a[edge_without_node])
             elif 2 ** (l - 1) < r * (l - 1):
                 gen_fun_coef = _get_gen_coef_subset_expansion(
                     a[edge_without_node], a[node], r - 1
@@ -175,8 +172,11 @@ def ttsv2(pair_dict, edge_dict, r, a, n):
     Sinan Aksoy, Ilya Amburg, Stephen Young,
     https://doi.org/10.1137/23M1584472
     """
+    from scipy.signal import convolve
+    from scipy.sparse import coo_array
+
     s = {}
-    r_minus_2_factorial = factorial(r - 2)
+    r_minus_2_factorial = math.factorial(r - 2)
     for (v1, v2), edges in pair_dict.items():
         c = 0
         for e in edges:
@@ -282,7 +282,7 @@ def _get_gen_coef_subset_expansion(edge_values, node_value, r):
             for i in range(len(subset_lengths))
         ]
     )
-    return total / factorial(r)
+    return total / math.factorial(r)
 
 
 def _get_gen_coef_fft_fast_array(edge_without_node, a, node, l, r):
@@ -317,6 +317,8 @@ def _get_gen_coef_fft_fast_array(edge_without_node, a, node, l, r):
     Sinan Aksoy, Ilya Amburg, Stephen Young,
     https://doi.org/10.1137/23M1584472
     """
+    from scipy.signal import convolve
+
     coefs = [1]
     for i in range(1, r):
         coefs.append(coefs[-1] * a[node] / i)