json-wrangling

Working with JSON

jq is available in the shell. Use it rather than reading a JSON file and reasoning about it by eye — it is cheaper and it does not make arithmetic mistakes.

Understand the shape first

Never assume a structure. Ask:

jq 'keys' /workspace/data.json
jq 'type' /workspace/data.json
jq '.[0] | keys' /workspace/list.json
jq -r 'paths(scalars) | join(".")' /workspace/data.json | sort -u

That last one prints every leaf path in the document and is the fastest way to learn an unfamiliar payload.

Everyday queries

jq '.version' package.json
jq -r '.dependencies | to_entries[] | "\(.key)@\(.value)"' package.json
jq '[.items[] | select(.active)] | length' data.json
jq -r '.records[] | [.id, .name, .total] | @csv' data.json
jq 'map(.amount) | add' orders.json
jq -s 'group_by(.status) | map({status: .[0].status, n: length})' events.json

Validation

jq exits non-zero on malformed input, which makes it a validator:

jq empty /workspace/config.json && echo valid

Editing JSON

To change a value, prefer jq to hand-editing — it cannot produce invalid JSON:

jq '.version = "2.0.0"' config.json > /workspace/.tmp.json && mv /workspace/.tmp.json config.json

Note jq cannot write in place, so route through a temporary file as above.

Not available

There is no sqlite3, no python, and no node in this Workspace — the packaged builds ship without their compiled runtimes. For anything beyond jq, use awk, sed, sort, cut and join, which are all present. If a task genuinely needs a database or a real script interpreter, say so rather than improvising something fragile.