This is a quick example to show how you could parse the hourly observation data from the Met Office DataPoint API and turn it into a CSV file. The code isn't very good as I threw this together in about an hour. An example CSV file is included.

daily_obs_csv.py

#!/usr/bin/python
 
"""
Assemble a CSV file from the Hourly Observation Service from the Met Office 
DataPoint API.
 
This is a bit of a mess – it was hacked together quickly to give an example of how
you could go about having a CSV file of observation data. It does an API request,
parses the data as JSON, then turns that into a set of data model objects, those
are them flattened and written down as a CSV file.
 
Author: Nick Charlton <hello@nickcharlton.net>
License: MIT
Date: 14/08/2013
 
Known Issues:
 
* Some of the data isn't filled in (data date, data type, etc).
* Error handling just sets the field to blank.
* There's only real error handling on the observation data.
"""
 
import requests
import json
import csv
 
API_KEY = ""
URL = "http://datapoint.metoffice.gov.uk/public/data/val/wxobs/all/json/all?res=hourly&key=%s" % (API_KEY)
OUTPUT_FILE = "example_data.csv"
 
def remove_key(d, key):
    """
    Remove a key from a dictionary.
    Return a copy of the dict.
    """
    r = dict(d)
    del r[key]
    return r
 
def flatten_location(loc):
    """
    Flatten a Location instance in preparation for writing to a CSV.
    """
    location = loc.__dict__
    location_info = remove_key(location, 'data')
    location_data = []
 
    for data in location['data']:
        flat_data = data.__dict__
        flat = dict(location_info.items() + flat_data.items())
        location_data.append(flat)
 
    return location_data
 
class Location():
    """
    Data Model for locations returned from DataPoint.
    """
    def __init__(self):
        self.data_date = ""
        self.data_type = ""
        self.location_id = ""
        self.lat = ""
        self.lon = ""
        self.name = ""
        self.country = ""
        self.continent = ""
        self.data = []
 
    def __repr__(self):
        return "<Location: %s>" % (self.name)
 
class Data():
    """
    Data Model for collected data returned from DataPoint
    """
    def __init__(self):
        self.collection_time = ""
        self.wind_direction = ""
        self.pressure = ""
        self.wind_speed = ""
        self.temperature = ""
        self.visibility = ""
        self.weather_type = ""
        
    def __repr__(self):
        return "<Data: %s>" % (self.collection_time)
 
# fetch data, parse json
r = requests.get(URL)
j = r.json()
 
# get a collection of locations
locations = j['SiteRep']['DV']['Location']
parsed_locations = []
 
for location in locations:
    # turn the locations into Location objects
    loc = Location()
    loc.location_id = location['i']
    loc.lat = location['lat']
    loc.lon = location['lon']
    loc.name = location['name']
    loc.country = location['country']
    loc.continent = location['continent']
 
    # turn the location's data into Data objects
    for period in location['Period']:
        for rep in period['Rep']:
            data = Data()
            data.collection_time = rep['$']
            try:
                data.wind_direction = rep['D']
            except KeyError:
                data.wind_direction = ""
            try:
                data.pressure = rep['P']
            except KeyError:
                data.pressure = ""
            try:
                data.wind_speed = rep['S']
            except KeyError:
                data.wind_speed = ""
            try:
                data.temperature = rep['T']
            except KeyError:
                data.temperature = ""
            try:
                data.visibility = rep['V']
            except KeyError:
                data.visibility = ""
            try:
                data.weather_type = rep['W']
            except KeyError:
                data.weather_type = ""
 
            loc.data.append(data)
 
    # hold onto a flattened version of the location
    parsed_locations.extend(flatten_location(loc))
 
# write the csv
with open(OUTPUT_FILE, "wb") as f:
    writer = csv.DictWriter(f, parsed_locations[0].keys())
    writer.writeheader()
    for d in parsed_locations:
        writer.writerow(d)

example_data.csv