3 comments

  • landgenoot 19 hours ago
    I think you can parse the date and time format as well to shrink your problem space.

    https://en.m.wikipedia.org/wiki/List_of_date_formats_by_coun...

    • Elliott-Diy 15 hours ago
      That's a great idea. I'll see how I can add this in today!
  • antman 16 hours ago
    I would assume you need to add your time and location at the time of the screenshot also so that you can correct the remote displayed time to Microsoft Weather time?
    • Elliott-Diy 15 hours ago
      Yep! The tool currently only works with screenshots taken within the past hour, since that's the only weather data I have access to right now. It also relies on having a consistent reference point for time. When you run the script, it uses your current timezone to help align the time shown in the screenshot with Microsoft Weather's data.
  • progbits 1 day ago
    Do you plan to scrape the data and maintain a historical database so old screenshots can be used too?

    Would be really cool but I imagine it could be difficult to get past rate limits.

    • Elliott-Diy 1 day ago
      That’s actually an end goal! I’m thinking of setting up a system to crowdsource this: if people opt in, the tool could cache the weather data it pulls and share it to help build a distributed historical dataset.
      • tgv 1 day ago
        There are meteorological services with open data access: https://www.ncei.noaa.gov/cdo-web/, https://english.knmidata.nl/open-dat, https://climatedataportal.metoffice.gov.uk/

        Mapping that would be quite a bit of work, I imagine.

        • Elliott-Diy 1 day ago
          Yeah, mapping it would definitely be a project. I looked into using open data from those sources, but ran into issues where the values didn’t match what MSN reports (which is what the Windows weather widget uses). Even small differences in temperature or how conditions are worded ("partly sunny" vs. "mostly sunny") can throw off the match. I’m planning to try using open data for historical screenshots eventually, but I expect it’ll reduce accuracy greatly.
    • rconti 1 day ago
      Using a time series database would be great here.