Skip to content

Latest commit

 

History

History
56 lines (40 loc) · 4.42 KB

README.md

File metadata and controls

56 lines (40 loc) · 4.42 KB

forecaster

Pulls weather data to make a forecast for a particular location, part of a code test

Requirements

  • Accept an address as input
  • Retrieve forecast data for the given address. This should include, at minimum, the current temperature (Bonus points - Retrieve high/low and/or extended forecast)
  • Display the requested forecast details to the user
  • Cache the forecast details for 30 minutes for all subsequent requests by zip codes.
  • Display indicator if result is pulled from cache.

My Assumptions

  • There should be a front-end for all of this, and also a way to fetch by JSON.
  • It is better to do this with minimal dependencies.
    • Similarly, I limited installation of rails (rails new --minimal --without-activerecord)

Details on Data

  • The geocode data comes in the following format:

    [{"place_id"=>18065671, "licence"=>"Data © OpenStreetMap contributors, ODbL 1.0. http://osm.org/copyright", "osm_type"=>"relation", "osm_id"=>9671540, "lat"=>"42.345695750000004", "lon"=>"-71.11946084959486", "category"=>"building", "type"=>"yes", "place_rank"=>30, "importance"=>9.99999999995449e-06, "addresstype"=>"building", "name"=>"", "display_name"=>"84, Browne Street, North Brookline, Brookline, Norfolk County, Massachusetts, 02446, United States", "address"=>{"house_number"=>"84", "road"=>"Browne Street", "neighbourhood"=>"North Brookline", "town"=>"Brookline", "county"=>"Norfolk County", "state"=>"Massachusetts", "ISO3166-2-lvl4"=>"US-MA", "postcode"=>"02446", "country"=>"United States", "country_code"=>"us"}, "boundingbox"=>["42.3455657", "42.3457554", "-71.1195458", "-71.1193964"]}]

  • The weather data looks like this:

    {"latitude"=>42.34697, "longitude"=>-71.107056, "generationtime_ms"=>0.051975250244140625, "utc_offset_seconds"=>0, "timezone"=>"GMT", "timezone_abbreviation"=>"GMT", "elevation"=>24.0, "current_units"=>{"time"=>"iso8601", "interval"=>"seconds", "temperature_2m"=>"°F"}, "current"=>{"time"=>"2024-08-04T03:45", "interval"=>900, "temperature_2m"=>77.7}, "daily_units"=>{"time"=>"iso8601", "temperature_2m_max"=>"°F", "temperature_2m_min"=>"°F"}, "daily"=>{"time"=>["2024-08-04", "2024-08-05", "2024-08-06", "2024-08-07", "2024-08-08", "2024-08-09", "2024-08-10"], "temperature_2m_max"=>[87.4, 94.3, 74.5, 68.9, 74.0, 73.1, 66.4], "temperature_2m_min"=>[74.8, 69.8, 65.1, 59.9, 58.1, 66.5, 61.8]}}

Details on Classes

  • The ForecastsController handles all the normal controller logic.

    • It has one route (index) which handles everything.
    • Given an address, it reaches out to the Geocode class to get latitutde, longitude, and zipcode.
    • For weather data, it reaches out to the Weather class to get current temperature and extended forecast.
    • Caching is handled by the controller before fetching weather data.

  • The Geocode class encapsulates all logic regarding obtaining location data for an address.

    • It uses OpenStreetMap from Nominatim.org to do the geocoding.
    • The fetch method does the call out to nominatim.org
    • The latitude, longitude, and zipcode methods pull out their respetive data from the return value from OpenStreetMap.

  • The Weather class encapsulates all logic regarding obtaining weather data for a location.

    • It uses open-meteo.com for weather data.
    • The fetch method does the call out to open-meteo.com.
    • The current_temperature method pulls out the current temperature from the weather data.
    • The extended_forecast method pulls out the forecast data and reformats it to be more useful on the front-end.

Notes

  • Rails caching is off in development, so to see the caching behavior run 'rails dev:cache' to toggle caching.
  • Basic styling is provided by https://picocss.com/docs
  • For scalability, the biggest issue with this solution is its reliance on free services without api keys. These services only allow a few thousand hits per day, so to scale higher than that we would need to move to a paid service.
  • Since there is no database, it would be easy to scale by creating lots of copies of this service. One thing that would need to change is the caching scheme - rather than being in memory, we would want to switch to shared redis service or other shared resource.
  • If we were to add a lot of options on the front end for what kind of weather would get returned and in what format, I would probably make the fetching function of the Weather class into its own object that can build these queries.