Change defaults for parse_metar_file

[dopplershift]

Right now, the signature of parse_metar_file is:

parse_metar_file(filename, year=datetime.now().year, month=datetime.now().month):

This can create some confusion:

• Generated docs show fixed default values, taken when the docs were last built (i.e. currently parse_metar_file(filename, year=2020, month=1))
• If a user has a long running process, IPython/Jupyter window open for awhile, etc., calls to parse_metar_file are not going to use the current year/month, but the year/month when MetPy was imported.

So we should change this to default to None, and then check for None and convert that to the appropriate values.

[akrherz]

My opinion is that you need more granularity here to also specify the day, so to get things right when parsing archived METARs around the first of the month.

[kgoebber]

So I couple with @akrherz as I have run into this issue quite a bit and think we may need to add in hour as well. Here is some code that would work to alleviate the problem, but with the need to have year, month, day, and hour, then we might as well make the input be a date time object.

from datetime import datetime, timedelta

import numpy as np

file_day = 31
file_hour = 23
file_minute = 50

known_date = datetime(2020, 2, 1, 0)
try:
    file_date = datetime(known_date.year, known_date.month, file_day, file_hour, file_minute)
except:
    file_date = datetime(known_date.year, known_date.month-1, file_day, file_hour, file_minute)

if np.abs(known_date - file_date) > timedelta(hours=1):
    if month == 1:
        file_date = datetime(known_date.year-1, 12, file_day, file_hour, file_minute)
    else:
        file_date = datetime(known_date.year, known_date.month-1, file_day, file_hour, file_minute)

print(known_date)

print(file_date)

Output:

2020-02-01 00:00:00
2020-01-31 23:50:00

[akrherz]

So I couple with @akrherz

Only on days ending with the letter 'y'.

Likely preaching to the choir here, but there's craziness on the noaaport feed, mostly international METARs

1. Products that come every hour with the same WMO header timestamp, for example one is coming today with a day timestamp of 31.
2. Products that come with a valid and current WMO header timestamp, but have actual METARS with various random dates that could be in the future or past or neither.
3. Products with valid WMO headers that update hourly, but the same METAR repeated every hour that is likely valid some number of years ago.