Timezone issue

[Garamotte]

Hello,

I recently started using this very useful lib, and discovered a timezone issue in the format_data method.

Enedis data are located in Europe/Paris timezone. This means that during the hour change day, we can have 2 values more (in October) or 2 values less (in March) for a day.

The format_data method uses a simple relativedelta addition on a timezone naive datetime object, which introduces a shift of all values by one hour when hitting an hour change.

Example : There was an hour change on 28th October 2018. Asking for data with start=2018-10-28 and end=2018-10-29 will give 50 values, because the hour between 2AM and 3AM is doubled.

The only way I know to fix this is to switch to UTC and return the values in UTC as well because switching back to Europe/Paris for returned values would break data again in October (2 values will be lost during the doubled hour). But changing the behaviour of the method can break existing programs that use the lib...

I am using the lib in a custom script to write data into InfluxDB, displayed in Grafana, and everything works well with this patch:

--- a/client.py 2019-05-05 11:06:42.958021515 +0200
+++ b/client.py 2019-05-05 11:05:59.677447347 +0200
@@ -2,6 +2,7 @@
 import simplejson
 import base64
 import datetime
+import pytz
 from dateutil.relativedelta import relativedelta
 import requests
 from fake_useragent import UserAgent
@@ -142,7 +143,9 @@
             periode = data.get("periode")
             if not periode:
                 return []
-            start_date = datetime.datetime.strptime(periode.get("dateDebut"), "%d/%m/%Y").date()
+            start_date = datetime.datetime.strptime(periode.get("dateDebut"), "%d/%m/%Y")
+            start_date = pytz.timezone('Europe/Paris').localize(start_date)
+            start_date = pytz.utc.normalize(start_date)
 
         # Calculate final start date using the "offset" attribute returned by the API
         inc = 1