Programatically export and/or delete Heroku dataclips

delete-dataclips.sh

#!/usr/bin/env bash

# there were a handful of clips that I didn't want to delete versus hundreds that I did,
# so I just removed the UUIDs of the clips I wanted to keep from uuids.txt by hand

cat uuids.txt | while read uuid
do
        curl \
          -H "authorization: Bearer $(heroku auth:token)" \
          -H "content-type: application/json" \
          -d '{"query":"mutation DeleteDataclip($clipId: ID!) { deleteClip(clipId: $clipId) }","variables":{"clipId":"'$uuid'"}}' \
          https://data-api.heroku.com/graphql
done

export-dataclips.sh

#!/usr/bin/env sh

# dump a JSON file with a UUID, title and the most recent SQL for every clip

curl \
  -H "authorization: Bearer $(heroku auth:token)" \
  -H "content-type: application/json" \
  -d '{ "query": "query { listClips { id title versions(limit: 1) { sql } } }" }' \
  https://data-api.heroku.com/graphql \
  > clips.json
  
  # also munge the output JSON into a text list of just UUIDs for additional manipulation
  
  less clips.json|\
    python -m json.tool|\ # get attributes on their own lines
    grep \"id\"|\ # filter for just the lines with UUIDs
    awk '{print substr($2,2,length($2)-3)}'\ # awk out just the UUID
    >uuids.txt

We decided that using Heroku's Dataclips was an anti-pattern at our company and needed a way to get rid of all of the crufty clips that were still sitting around. They don't make it easy to either export the queries or do any sort of bulk operations, so... a little reverse-engineering of their GraphQL endpoints and a shell script came to the rescue here.

It goes without saying that this is neither a sanctioned usage nor a stable API, so it may break without warning. C'est la vie.