From f89e4968edd4d864039c4d10ebac284a6dde7e54 Mon Sep 17 00:00:00 2001
From: emanuel-schmid <schmide@ethz.ch>
Date: Fri, 27 Sep 2024 16:21:41 +0200
Subject: [PATCH] suppress datetime conversion warning

---
 climada_petals/hazard/tc_tracks_forecast.py | 58 ++++++++++++---------
 1 file changed, 32 insertions(+), 26 deletions(-)

diff --git a/climada_petals/hazard/tc_tracks_forecast.py b/climada_petals/hazard/tc_tracks_forecast.py
index ef40f9bf..786b3c1d 100644
--- a/climada_petals/hazard/tc_tracks_forecast.py
+++ b/climada_petals/hazard/tc_tracks_forecast.py
@@ -30,6 +30,7 @@
 import os
 import tempfile
 from pathlib import Path
+import warnings
 
 # additional libraries
 import eccodes as ec
@@ -506,33 +507,38 @@ def _subset_to_track(msg, index, provider, timestamp_origin, name, id_no):
         # 0 means the deterministic analysis, which we want to flag.
         # See documentation for link to ensemble types.
         ens_bool = msg['ens_type'][index] != 0
-
         try:
-            track = xr.Dataset(
-                data_vars={
-                    'max_sustained_wind': ('time', np.squeeze(wnd)),
-                    'central_pressure': ('time', np.squeeze(pre)/100),
-                    'radius_max_wind': ('time', np.squeeze(rad)),
-                    'ts_int': ('time', timestep_int),
-                },
-                coords={
-                    'time': timestamp,
-                    'lat': ('time', lat),
-                    'lon': ('time', lon),
-                },
-                attrs={
-                    'max_sustained_wind_unit': 'm/s',
-                    'central_pressure_unit': 'mb',
-                    'name': name,
-                    'sid': sid,
-                    'orig_event_flag': False,
-                    'data_provider': provider,
-                    'id_no': (int(id_no) + index / 100),
-                    'ensemble_number': msg['ens_number'][index],
-                    'is_ensemble': ens_bool,
-                    'run_datetime': timestamp_origin,
-                }
-            )
+            with warnings.catch_warnings():
+                # prevent issueing a million warnings about conversion of non-nanosecond precision
+                # datetime to nanosecond precision, e.g. in fetch_ecmwf
+                # TODO: fix it through converting those _before_ creating the xr.Dataset
+                warnings.simplefilter(action='error', category=UserWarning)
+
+                track = xr.Dataset(
+                    data_vars={
+                        'max_sustained_wind': ('time', np.squeeze(wnd)),
+                        'central_pressure': ('time', np.squeeze(pre)/100),
+                        'radius_max_wind': ('time', np.squeeze(rad)),
+                        'ts_int': ('time', timestep_int),
+                    },
+                    coords={
+                        'time': timestamp,
+                        'lat': ('time', lat),
+                        'lon': ('time', lon),
+                    },
+                    attrs={
+                        'max_sustained_wind_unit': 'm/s',
+                        'central_pressure_unit': 'mb',
+                        'name': name,
+                        'sid': sid,
+                        'orig_event_flag': False,
+                        'data_provider': provider,
+                        'id_no': (int(id_no) + index / 100),
+                        'ensemble_number': msg['ens_number'][index],
+                        'is_ensemble': ens_bool,
+                        'run_datetime': timestamp_origin,
+                    }
+                )
         except ValueError as err:
             LOGGER.warning(
                 'Could not process track %s subset %d, error: %s',