lazy load of expensive / niche imports

ecoscope/analysis/ecograph.py

@@ -2,6 +2,7 @@
 
 import geopandas as gpd
 import igraph
+import sklearn
 import networkx as nx
 import numpy as np
 import pandas as pd
@@ -10,7 +11,7 @@
 from shapely.geometry import shape
 from skimage.draw import line
 
-import ecoscope
+from ecoscope.io.raster import RasterExtent, RasterProfile, RasterPy
 
 
 class Ecograph:
@@ -118,7 +119,6 @@ def to_geotiff(self, feature, output_path, individual="all", interpolation=None,
         transform : sklearn.base.TransformerMixin or None
             A feature transform method (Default : None)
         """
-        from sklearn.base import TransformerMixin
 
         if feature in self.features:
             if individual == "all":
@@ -130,22 +130,20 @@ def to_geotiff(self, feature, output_path, individual="all", interpolation=None,
         else:
             raise ValueError("This feature was not computed by EcoGraph")
 
-        if isinstance(transform, TransformerMixin):
+        if isinstance(transform, sklearn.base.TransformerMixin):
             nan_mask = ~np.isnan(feature_ndarray)
             feature_ndarray[nan_mask] = transform.fit_transform(feature_ndarray[nan_mask].reshape(-1, 1)).reshape(
                 feature_ndarray[nan_mask].shape
             )
 
-        raster_profile = ecoscope.io.raster.RasterProfile(
+        raster_profile = RasterProfile(
             pixel_size=self.resolution,
             crs=self.utm_crs,
             nodata_value=np.nan,
             band_count=1,
         )
-        raster_profile.raster_extent = ecoscope.io.raster.RasterExtent(
-            x_min=self.xmin, x_max=self.xmax, y_min=self.ymin, y_max=self.ymax
-        )
-        ecoscope.io.raster.RasterPy.write(
+        raster_profile.raster_extent = RasterExtent(x_min=self.xmin, x_max=self.xmax, y_min=self.ymin, y_max=self.ymax)
+        RasterPy.write(
             ndarray=feature_ndarray,
             fp=output_path,
             **raster_profile,

ecoscope/analysis/seasons.py

@@ -6,6 +6,7 @@
 import shapely
 from scipy.stats import norm
 from sklearn.preprocessing import LabelEncoder
+from sklearn.mixture import GaussianMixture
 
 logger = logging.getLogger(__name__)
 
@@ -49,8 +50,6 @@ def std_ndvi_vals(aoi=None, img_coll=None, nir_band=None, red_band=None, img_sca
 
 
 def val_cuts(vals, num_seasons=2):
-    from sklearn.mixture import GaussianMixture
-
     distr = GaussianMixture(n_components=num_seasons, max_iter=500)
     vals = vals["NDVI"].to_numpy().reshape(-1, 1)
     distr.fit(vals)

ecoscope/analysis/speed.py

@@ -5,6 +5,9 @@
 import shapely
 
 import ecoscope.base
+import lazy_loader as lazy
+
+mapclassify = lazy.load("mapclassify")
 
 
 class SpeedDataFrame(ecoscope.base.EcoDataFrame):
@@ -64,23 +67,21 @@ def apply_classification(x, k, cls_method="natural_breaks", multiples=[-2, -1, 1
         The multiples of the standard deviation to add/subtract from the sample mean to define the bins. defaults=
     """
 
-    from mapclassify import classifiers
-
     classification_methods = {
-        "equal_interval": classifiers.EqualInterval,
-        "natural_breaks": classifiers.NaturalBreaks,
-        "quantile": classifiers.Quantiles,
-        "std_mean": classifiers.StdMean,
-        "max_breaks": classifiers.MaximumBreaks,
-        "fisher_jenks": classifiers.FisherJenks,
+        "equal_interval": mapclassify.EqualInterval,
+        "natural_breaks": mapclassify.NaturalBreaks,
+        "quantile": mapclassify.Quantiles,
+        "std_mean": mapclassify.StdMean,
+        "max_breaks": mapclassify.MaximumBreaks,
+        "fisher_jenks": mapclassify.FisherJenks,
     }
 
     classifier = classification_methods.get(cls_method)
     if not classifier:
         return
 
     map_classifier = classifier(x, multiples) if cls_method == "std_mean" else classifier(x, k)
-    edges, _, _ = classifiers._format_intervals(map_classifier, fmt="{:.2f}")
+    edges, _, _ = mapclassify.classifiers._format_intervals(map_classifier, fmt="{:.2f}")
     return [float(i) for i in edges]
 
 

ecoscope/base/base.py

@@ -17,6 +17,10 @@
 )
 from ecoscope.base.utils import cachedproperty
 
+import lazy_loader as lazy
+
+astroplan = lazy.load("astroplan")
+
 
 class EcoDataFrame(gpd.GeoDataFrame):
     """
@@ -393,7 +397,6 @@ def get_daynight_ratio(self, n_grid_points=150) -> pd.Series:
         pd.Series:
             Daynight ratio for each unique individual subject in the grouby_col column.
         """
-        import astroplan
 
         locations = to_EarthLocation(self.geometry.to_crs(crs=self.estimate_utm_crs()).centroid)
 

ecoscope/mapping/map.py

@@ -21,6 +21,10 @@
 import ecoscope
 from ecoscope.contrib.foliumap import Map
 
+import lazy_loader as lazy
+
+ds = lazy.load("datashader")
+
 warnings.filterwarnings("ignore", "GeoSeries.isna", UserWarning)
 
 
@@ -508,7 +512,6 @@ def add_datashader_gdf(
         kwargs
             Additional kwargs passed to datashader.transfer_functions.shade
         """
-        import datashader as ds
 
         gdf = gdf.to_crs(epsg=4326)
         bounds = gdf.geometry.total_bounds

ecoscope/plotting/plot.py

@@ -3,6 +3,7 @@
 import plotly.graph_objs as go
 import shapely
 from plotly.subplots import make_subplots
+from sklearn.neighbors import KernelDensity
 
 
 class EcoPlotData:
@@ -242,8 +243,6 @@ def speed(trajectory):
 
 
 def plot_seasonal_dist(ndvi_vals, cuts, bandwidth=0.05, output_file=None):
-    from sklearn.neighbors import KernelDensity
-
     x = ndvi_vals.sort_values().to_numpy().reshape(-1, 1)
     kde = KernelDensity(kernel="gaussian", bandwidth=bandwidth).fit(x)
     dens = np.exp(kde.score_samples(x))