ENH: Make parsing Wyoming post-data values more robust (Fixes #749)

The main goal is to allow for missing station information, which results
in '******' for latitude, no longitude, and -9999.0 for elevation. As a
result, avoid the fixed index parsing of this information, and instead
rely on the text to the left of the ':' to identify appropriate
sections.

src/siphon/simplewebservice/wyoming.py

@@ -15,6 +15,11 @@
 from ..http_util import HTTPEndPoint
 
 
+def _safe_float(s):
+    """Convert to float, handling ****** as a string for missing."""
+    return pd.NA if all(c == '*' for c in s) or s == '-9999.0' else float(s)
+
+
 class WyomingUpperAir(HTTPEndPoint):
     """Download and parse data from the University of Wyoming's upper air archive."""
 
@@ -83,18 +88,16 @@ def _get_data(self, time, site_id):
         meta_data = soup.find_all('pre')[1].contents[0]
         lines = meta_data.splitlines()
 
-        # If the station doesn't have a name identified we need to insert a
-        # record showing this for parsing to proceed.
-        if 'Station number' in lines[1]:
-            lines.insert(1, 'Station identifier: ')
-
-        station = lines[1].split(':')[1].strip()
-        station_number = int(lines[2].split(':')[1].strip())
-        sounding_time = datetime.strptime(lines[3].split(':')[1].strip(), '%y%m%d/%H%M')
-        latitude = float(lines[4].split(':')[1].strip())
-        longitude = float(lines[5].split(':')[1].strip())
-        elevation = float(lines[6].split(':')[1].strip())
-        pw = float(lines[-1].split(':')[1].strip())
+        # Convert values after table into key, value pairs using the name to the left of the :
+        post_values = dict(tuple(map(str.strip, l.split(': '))) for l in lines[1:])
+
+        station = post_values.get('Station identifier', '')
+        station_number = int(post_values['Station number'])
+        sounding_time = datetime.strptime(post_values['Observation time'], '%y%m%d/%H%M')
+        latitude = _safe_float(post_values.get('Station latitude', '******'))
+        longitude = _safe_float(post_values.get('Station longitude', '******'))
+        elevation = _safe_float(post_values.get('Station elevation', '******'))
+        pw = float(post_values['Precipitable water [mm] for entire sounding'])
 
         df['station'] = station
         df['station_number'] = station_number

tests/test_wyoming.py

@@ -6,6 +6,7 @@
 from datetime import datetime
 
 from numpy.testing import assert_almost_equal
+import pandas as pd
 import pytest
 
 from siphon.simplewebservice.wyoming import WyomingUpperAir
@@ -140,3 +141,14 @@ def test_wyoming_heights():
 
     assert_almost_equal(df['height'][140], 10336.0, 2)
     assert_almost_equal(df['direction'][1], 145.0, 1)
+
+
+# GH #749
+@recorder.use_cassette('wyoming_missing_station_info')
+def test_missing_station():
+    """Test that we can still return data for stations missing from the Wyoming archive."""
+    df = WyomingUpperAir.request_data(datetime(2012, 1, 1, 0),  '82244')
+    assert df['station'][0] == ''
+    assert pd.isna(df['latitude'][0])
+    assert pd.isna(df['longitude'][1])
+    assert pd.isna(df['elevation'][0])