Caching API responses while experimenting

[dimo414]

When I'm experimenting with an API call, especially to filter or post-process the response, it can be nice to invoke the API itself via bkt so that the raw response is cached. This saves time and resources without getting in the way.

As an example, I was poking around at the NWS Weather API and wanted to use jq to display the relevant parts of the response in an ASCII table. It took several passes to get the behavior I wanted, and while the NWS API is fast and free there's no need to hit it repeatedly in a short time frame.

$ bkt --ttl=1h -- \
  curl --silent --show-error https://api.weather.gov/gridpoints/TOP/31,80/forecast \
  | jq -r '[" ","Temp","Wind","Forecast"], (
             .properties.periods[]
             | [ .name, ((.temperature|tostring) + .temperatureUnit ),
                 (.windSpeed + " " + .windDirection), .shortForecast ]
           ) | @tsv' \
  | column -ts $'\t'
                  Temp  Wind             Forecast
Tonight           47F   10 to 20 mph S   Mostly Cloudy
Wednesday         56F   5 to 15 mph W    Mostly Cloudy
Wednesday Night   25F   15 to 20 mph N   Partly Cloudy
Thanksgiving Day  40F   5 to 15 mph NW   Sunny
Thursday Night    27F   5 to 10 mph SW   Mostly Clear
Friday            59F   10 to 15 mph SW  Sunny
Friday Night      37F   5 mph W          Mostly Clear
Saturday          60F   5 to 15 mph NW   Mostly Sunny
Saturday Night    32F   5 to 10 mph NW   Mostly Clear
Sunday            55F   5 to 10 mph W    Sunny
Sunday Night      37F   5 to 10 mph SW   Mostly Clear
Monday            60F   10 mph W         Sunny
Monday Night      35F   10 mph NW        Mostly Clear
Tuesday           53F   10 mph NW        Sunny

This has several advantages over dumping the response to a file:

1. Minimal changes to the intended command line; simply prepend bkt -- to the front rather than splitting the pipeline into parts and adding redirects to/from files
2. Caches different API calls separately so you can experiment with different requests at the same time, without needing to keep track of one or more response files
3. Easily refresh at any point by passing --force