Row Stream Source

rows() is a source builtin. It changes what the receiver means: instead of querying one document value, it exposes a stream of rows.

`rows()`

Signature: Source -> Stream<Row>
Arity: zero
Demand behavior: forwards retained-row demand to the source
Supported stream stages: reverse, filter, find, distinct_by, take, first, map

On a normal JSON document, $.rows() treats the document itself as one row:

DOC:    {"id":1,"name":"Ada"}
QUERY:  $.rows().map({id: $.id, name: $.name})
OUT:    [{"id":1,"name":"Ada"}]

Top-level arrays are also one document row in normal JSON mode. Use normal array methods directly when the input document is an array.

In NDJSON mode, root $.rows() means all rows in the file or reader:

jetrocli --ndjson -i events.ndjson \
  -e '$.rows().filter($.active).take(10).map({id: $.id, name: $.name})'

Without $.rows(), the same CLI mode is row-local:

jetrocli --ndjson -i events.ndjson -e '$.id'

For file-backed NDJSON, reverse() scans from the end:

jetrocli --ndjson -i app.log \
  -e '$.rows().reverse().find($.level == "error").first()'

Reader-backed reverse streams are unsupported because readers cannot seek.

For Kafka compacted-topic dumps, scan newest-to-oldest and keep the first row seen for each key:

jetrocli --ndjson -i users.ndjson --payload-after '|' \
  -e '$.rows().reverse().distinct_by($.id).take(100).map({id: $.id, name: $.name})'

distinct_by in this stream order keeps the newest row for each key and drops older duplicates immediately.