This is a little command line tool meant to eat newline-delimited JSON (CouchDB documents) on stdin and POST
ing them to the _bulk_docs
endpoint of a CouchDB server.
This is a fork from couchdb-bulk, with the following modifications:
- posts 1000 docs per bulk request (instead of only 1) when the input is newline-delimited JSON
- drops supports for anything else than newline-delimited JSON (1 doc per line), assuming that getting to that data format is the job of another tool, such as jq
- drops supports for the module interface, only the CLI mode remains
- modernized code and dependencies
npm install -g couchdb-bulk2
couchdb-bulk2 url [file]
The [file]
argument is optional, if its missing (or if its '-'), input is expected to be piped via stdin
Example:
cat ./test/fixtures/docs.ndjson | couchdb-bulk2 http://username:password@localhost:5984/testdb
// OR
couchdb-bulk2 http://username:password@localhost:5984/testdb ./test/fixtures/docs.ndjson
couchdb-bulk2
expects the input to be newline-delimited JSON.
See http://jsonlines.org for more info on this format.
Each line should be a single doc:
{ "_id": "one" }
{ "_id": "two" }
{ "_id": "three" }
This newline-delimited JSON format can easily be obtained from a JSON document containing an array of docs using a tool such as jq
cat view_reponse.json | jq -c '.docs[]' | couchdb-bulk2 http://username:password@localhost:5984/testdb
By default, couchdb-bulk2
generates files with the bulk operations results, unless the stream, stdout
or stder
, is already being redirected to a file. To override that behavior, you can redirect those streams yourself:
cat view_reponse.json | jq -c '.docs[]' | couchdb-bulk2 http://username:password@localhost:5984/testdb > ./stdout 2> ./stderr
-l, --batch-length <number>
: Set the number of documents to be sent in bulk to CouchDB per batch (Default: 1000)-s, --sleep <milliseconds>
: Defines the amount of time (in milliseconds) to wait once a batch was sent before sending a new one (Default: 0)-o, --output <path>
: Customize output directory for the stdout or stderr streams that are not already redirected to a file-q, --quiet
: Do not log output files and operations statistics (Default: false')
couchdb-bulk2
works great in combinaison withndjson-apply
: see data transformation workflow in the Inventaire project