Improve and refactor finding extrema

visualizer/plotting.py

@@ -2,6 +2,7 @@
 
 import math
 from io import BytesIO
+import operator
 from types import SimpleNamespace
 
 from flask import Response
@@ -84,8 +85,7 @@ class TimeSeries(Plotter):
         axes.hlines(y=avg, xmin=data[0, 0], xmax=data[-1, 0])
         axes.set_ylabel(f'{meas.title.title()} ({meas.units})')
         axes.grid(True)
-        maximum = max(data, key=lambda row: row[meas.field])
-        minimum = min(data, key=lambda row: row[meas.field])
+        maximum, minimum = data_max_min(data, field=meas.field)
         axes.set_title(
             (f'Max {meas.title}: {maximum[meas.field]} {meas.units} '
              f'on {maximum[0]:%F at %R} / '
@@ -168,15 +168,11 @@ class Overlay(Plotter):
                         alpha=0.875 if i else 1.0)
         axes.set_ylabel(f'{meas.title.title()} ({meas.units})')
 
+        row_meas = operator.itemgetter(meas.field)
         for dset in datasets:
-            comparable_data = [row for row in dset.data
-                               if not math.isnan(row[meas.field])]
-            dset.max = max(comparable_data, key=lambda row: row[meas.field],
-                           default=dset.data[0])
-            dset.min = min(comparable_data, key=lambda row: row[meas.field],
-                           default=dset.data[0])
-        max_dset = max(datasets, key=lambda dset: dset.max[meas.field])
-        min_dset = min(datasets, key=lambda dset: dset.min[meas.field])
+            dset.max, dset.min = data_max_min(dset.data, key=row_meas)
+        max_dset = max(datasets, key=lambda dset: row_meas(dset.max))
+        min_dset = min(datasets, key=lambda dset: row_meas(dset.min))
 
         axes.grid(True)
         axes.set_title(
@@ -239,23 +235,17 @@ class Boxplot(Plotter):
         start_year, end_year = sorted([year1, year2])
         years = range(start_year, end_year + 1)
 
-        maximum = minimum = None
-
-        def at_index(row):
-            """Get the selected field of `row`."""
-            return row[meas.field]
-
+        row_meas = operator.itemgetter(meas.field)
+        years_maxima = []
+        years_minima = []
         for year in years:
             data = read_data(station, year)
             plot_data.append(data[:, meas.field])
-
-            year_max = max(data, key=at_index)
-            year_min = min(data, key=at_index)
-            if maximum is None:
-                maximum, minimum = year_max, year_min
-            else:
-                maximum = max(maximum, year_max, key=at_index)
-                minimum = min(minimum, year_min, key=at_index)
+            year_max, year_min = data_max_min(data, key=row_meas)
+            years_maxima.append(year_max)
+            years_minima.append(year_min)
+        maximum = max(years_maxima, key=row_meas)
+        minimum = min(years_minima, key=row_meas)
 
         fig, axes = plt.subplots()
         fig.set_figheight(6)
@@ -289,3 +279,23 @@ def savefig_response(fig, filename=None):
     if filename is not None:
         res.headers['Content-Disposition'] = f'inline; filename="{filename}"'
     return res
+
+
+def data_max_min(data, field=None, key=None):
+    """Smarter, NaN-aware `max_min()`, for plot functions."""
+    if key is None:
+        key = operator.itemgetter(field)
+    comparable = non_nan(data, key=key)
+    return max_min(comparable, key=key, default=data[0])
+
+
+def max_min(*args, **kwargs):
+    """Call `max()` and `min()` with the same arguments."""
+    return max(*args, **kwargs), min(*args, **kwargs)
+
+
+def non_nan(items, key=None):
+    """Filter out items that are NaN."""
+    if key is None:
+        key = lambda x: x
+    return [i for i in items if not math.isnan(key(i))]