#ldjson-to-couchdb

Pipe STDIN in LDJSON format to CouchDB.

##Run

Install with npm and pipe data.json to CouchDB database called db on localhost:5984.

$ npm install ldjson-to-couchdb -g
$ cat data.json | ldjson-to-couchdb "http://localhost:5984/db"

It is your job to make sure that db exists:

$ curl -X PUT http://127.0.0.1:5984/db

If using the module programmatically, you can pipe the output from CouchDB as follows:

through = require('ldjson-to-couchdb')
process.stdin.pipe(through()).pipe(process.stdout)

OR you can pass the db, key and batch params like so:

through = require('ldjson-to-couchdb')
through('http://127.0.0.1:5984/db', '_id', 100).pipe(process.stdout)

###Insert vs Update

If you pass a key as an argument, we will update existing document rather than bulk insert (which us much faster).

$ cat data.json | ldjson-to-couchdb "http://localhost:5984/db" _id

###Output Stream

The output from CouchDB is piped to STDOUT in LDJSON format:

{"ok":true,"id":"2524e21ff3a67c1e0b416c999b014c11","rev":"1-f443804e079279645227a179f7e373cf"}
{"ok":true,"id":"2524e21ff3a67c1e0b416c999b015b0a","rev":"1-d35680189514ccd6046f23daa89b0f84"}

##Source

_       = require 'highland'
ndjson  = require 'ndjson'
nano    = require 'nano'
{ EOL } = require 'os'

module.exports = (db, key, batch=50) ->
  # Will throw err in bulk insert if uri not present/malformed.
  db = nano db or process.argv[2]
  # The document key on which to perform update.
  key = key or process.argv[3] unless 'TEST' of process.env

  # Output data in NDJSON format.
  output = (push, err, data) ->
    switch
      when data instanceof Array
        ( push err, JSON.stringify(res) + EOL for res in data )
      else
        push err, JSON.stringify(data) + EOL
    push null, _.nil

  through = (docs) ->
    # Bulk insert when no key specified.
    unless key?
      return _ (push, next) ->
        db.bulk { docs }, _.partial output, push

    update = (doc) ->
      _ (push, next) ->
        # Just insert if our doc does not have a key.
        unless (doc_name = doc[key])?
          return db.insert doc, _.partial output, push

        # Check the revision first.
        db.head doc_name, (err, res, headers) ->
          doc._rev = headers.etag[1...-1] if headers
          db.insert doc, doc_name, _.partial output, push

    # Update pipeline.
    _.pipeline _(docs), _.map(update), _.parallel batch

  # Split on newlines and pipe to a transform.
  _.pipeline.apply _, [
    do _
    do ndjson.parse
    _.batch batch
    _.map through
    do _.series
  ]