[Fix] Update plotting imports and thread handling in Pepper class (#382)

- plt.subplot does not work reliably with async/ threads.
- Bug in thread run that would fail with env set (string to number)

Reviewed-on: enviPath/enviPy#382
Co-authored-by: Tobias O <tobias.olenyi@envipath.com>
Co-committed-by: Tobias O <tobias.olenyi@envipath.com>

pepper/__init__.py

@@ -46,7 +46,7 @@ class PepperPrediction(PropertyPrediction):
 
         import matplotlib.patches as mpatches
         import numpy as np
-        from matplotlib import pyplot as plt
+        from matplotlib.figure import Figure
         from scipy import stats
 
         """
@@ -101,7 +101,8 @@ class PepperPrediction(PropertyPrediction):
         mask_red = x > vp
 
         # Plot
-        fig, ax = plt.subplots(figsize=(9, 5.5))
+        fig = Figure(figsize=(9, 5.5))
+        ax = fig.subplots()
         ax.plot(x, y, color="#1f4e79", lw=2, label="Lognormal PDF")
 
         if np.any(mask_green):
@@ -146,13 +147,12 @@ class PepperPrediction(PropertyPrediction):
         ]
         ax.legend(handles=patches, frameon=True)
 
-        plt.tight_layout()
+        fig.tight_layout()
 
         # --- Export to SVG string ---
         buf = io.StringIO()
         fig.savefig(buf, format="svg", bbox_inches="tight")
         svg = buf.getvalue()
-        plt.close(fig)
         buf.close()
 
         return svg

pepper/impl/pepper.py

@@ -187,8 +187,9 @@ class Pepper:
         groups = [group for group in dataset.group_by("structure_id")]
 
         # Unless explicitly set compute everything serial
-        if os.environ.get("N_PEPPER_THREADS", 1) > 1:
-            results = Parallel(n_jobs=os.environ["N_PEPPER_THREADS"])(
+        n_threads = int(os.environ.get("N_PEPPER_THREADS", 1))
+        if n_threads > 1:
+            results = Parallel(n_jobs=n_threads)(
                 delayed(compute_bayes_per_group)(group[1])
                 for group in dataset.group_by("structure_id")
             )