added historical module - #5
Conversation
| self, | ||
| zipCode: str, | ||
| *, | ||
| date: Optional[Union[date_, datetime, str]] = None, |
There was a problem hiding this comment.
I think even though the API allows a call with no date and returns an empty set of observations, it makes more sense for a historical API to make this required and do our own check for presence in the code. This seems to work a bit differently than the forecast which defaults to today if the parameter isn't sent.
| '''Airnow needs T00-0000 appended to the date parameter for historical calls''' | ||
| if date and isinstance(date, str): | ||
| y, m, d = date.split('-') | ||
| params['date'] = date_(int(y), int(m), int(d)).isoformat() + "T00-0000" | ||
| elif date and isinstance(date, datetime): | ||
| params['date'] = date.date().isoformat() + "T00-0000" | ||
| elif date and isinstance(date, date_): | ||
| params['date'] = date.isoformat() + "T00-0000" |
There was a problem hiding this comment.
I believe this is actually an odd form of extended ISO 8601 timestamp where the T00 is the time with only hours (minutes and seconds may be omitted if they are zero) and -0000 is the UTC offset. I think a better way to do it is to get the argument into a datetime, ensure it has a time zone offset set, then use strftime to put it into the correct %Y-%m-%dT%H%z format instead of relying on isoformat and hard-coded strings.
I probably should have done something similar originally in the Forecast module, although interestingly the Forecast API endpoint does not like the T00-0000 so the EPA is using different parsers on the backend.
| '''create a timezone object with no utc offset''' | ||
| tz = timezone(timedelta()) | ||
| params['date'] = datetime(int(y), int(m), int(d), tzinfo=tz).strftime("%Y-%m-%dT%H%z") | ||
| elif isinstance(date, datetime): | ||
| tz = timezone(timedelta()) | ||
| date = date.replace(tzinfo=tz) | ||
| params['date'] = date.strftime("%Y-%m-%dT%H%z") | ||
| elif isinstance(date, date_): | ||
| tz=timezone(timedelta()) | ||
| params['date'] = datetime(date.year, date.month, date.day, tzinfo=tz).strftime("%Y-%m-%dT%H%z") | ||
| else: | ||
| raise AirNowError("Bad date type. Date parameter is either date, datetime, or a string.") |
There was a problem hiding this comment.
@asymworks I switched to using strftime to get the right date formatting as you suggested. Interestingly, when a datetime object is set to utc timezone, the offset becomes +0000 and not -0000. I tested this with some simple calls to the api and it works just fine either way.
|
@Big-Gray CI checks are failing due to running into a broken dependency with Python 3.12. I fixed the workflow in the master branch - please rebase your branch to pull those updates in and CI should pass. |
|
I'd be interested in this feature. I'm curious if there are any updates? I might also have some time to help out and rebase it. |
3 participants
I'm not sure why but historical observation calls always need a "T00-0000" appended to the date string. I haven't finished writing tests yet.