CouchImport

Introduction

When populating CouchDB databases, often the source of the data is initially a CSV or TSV file. couchimport is designed to assist you with importing flat data into CouchDB efficiently. It can be used either as command-line utilities couchimport and couchexport or the underlying functions can be used programmatically:

• simply pipe the data file to couchimport on the command line.
• handles tab or comma-separated data.
• uses Node.js's streams for memory efficiency.
• plug in a custom function to add your own changes before the data is written.
• writes the data in bulk for speed.
• can also read huge JSON files using a streaming JSON parser.
• allows multiple HTTP writes to happen at once using the --parallelism option.

Installation

Requirements

• node.js = npm

  sudo npm install -g couchimport

Configuration

couchimport's configuration parameters can be stored in environment variables or supplied as command line arguments.

The location of CouchDB

Simply set the COUCH_URL environment variable e.g. for a hosted Cloudant database

  export COUCH_URL="https://myusername:[email protected]"

or a local CouchDB installation:

  export COUCH_URL="http://localhost:5984"

IAM Authentication

Alternatively, if you are using IAM authentication with IBM Cloudant, then supply two environment variables:

• COUCH_URL - the URL of your Cloudant host e.g. https://myhost.cloudant.com (note absence of username and password in URL).
• IAM_API_KEY - the IAM API KEY e.g. ABC123515-151215.

The name of the database - default "test"

Define the name of the CouchDB database to write to by setting the COUCH_DATABASE environment variable e.g.

  export COUCH_DATABASE="mydatabase"

Transformation function - default nothing

Define the path of a file containing a transformation function e.g.

  export COUCH_TRANSFORM="/home/myuser/transform.js"

The file should:

• be a JavaScript file
• export one function that takes a single doc and returns a single object or an array of objects if you need to split a row into multiple docs.

(see examples directory).

Delimiter - default "\t"

The define the column delimiter in the input data e.g.

  export COUCH_DELIMITER=","

Running

Simply pipe the text data into "couchimport":

  cat ~/test.tsv | couchimport

This example downloads public crime data, unzips and imports it:

  curl 'http://data.octo.dc.gov/feeds/crime_incidents/archive/crime_incidents_2013_CSV.zip' > crime.zip
  unzip crime.zip
  export COUCH_DATABASE="crime_2013"
  export COUCH_DELIMITER=","
  ccurl -X PUT /crime_2013
  cat crime_incidents_2013_CSV.csv | couchimport

In the above example we use (ccurl)[https://github.com/glynnbird/ccurl], a command-line utility that uses the same environment variables as couchimport.

Output

The following output is visible on the console when "couchimport" runs:

couchimport
-----------
 url         : "https://****:****@myhost.cloudant.com"
 database    : "test"
 delimiter   : "\t"
 buffer      : 500
 parallelism : 1
 type        : "text"
-----------
  couchimport Written ok:500 - failed: 0 -  (500) +0ms
  couchimport { documents: 500, failed: 0, total: 500, totalfailed: 0 } +0ms
  couchimport Written ok:499 - failed: 0 -  (999) +368ms
  couchimport { documents: 499, failed: 0, total: 999, totalfailed: 0 } +368ms
  couchimport writecomplete { total: 999, totalfailed: 0 } +0ms
  couchimport Import complete +81ms

The configuration, whether default or overriden by environment variables or command line arguments, is shown. This is followed by a line of output for each block of 500 documents written, plus a cumulative total.

Preview mode

If you want to see a preview of the JSON that would be created from your csv/tsv files then add --preview true to your command-line:

    > cat text.txt | couchimport --preview true
    Detected a TAB column delimiter
    { product_id: '1',
      brand: 'Gibson',
      type: 'Electric',
      range: 'ES 330',
      sold: 'FALSE' }

As well as showing a JSON preview, preview mode also attempts to detect the column delimiter character for you.

Importing large JSON documents

If your source document is a GeoJSON text file, couchimport can be used. Let's say your JSON looks like this:

{ "features": [ { "a":1}, {"a":2}] }

and we need to import each feature object into CouchDB as separate documents, then this can be imported using the type="json" argument and specifying the JSON path using the jsonpath argument:

  cat myfile.json | couchimport --database mydb --type json --jsonpath "features.*"

Importing JSON Lines file

If your source document is a JSON Lines text file, couchimport can be used. Let's say your JSON Lines looks like this:

{"a":1}
{"a":2}
{"a":3}
{"a":4}
{"a":5}
{"a":6}
{"a":7}
{"a":8}
{"a":9}

and we need to import each line as a JSON object into CouchDB as separate documents, then this can be imported using the type="jsonl" argument:

  cat myfile.json | couchimport --database mydb --type jsonl

Importing a stream of JSONs

If your source data is a lot of JSON objects meshed or appended together, couchimport can be used. Let's say your file:

{"a":1}{"a":2}  {"a":3}{"a":4}
{"a":5}          {"a":6}
{"a":7}{"a":8}



{"a":9}

and we need to import each JSON objet to CouchDB as separate documents, then this can be imported using the type="jsonl" argument:

  cat myfile.json.blob | couchimport --database mydb --type jsonl

Overwriting existing data

If you are importing data into a CouchDB database that already contains data, and you are supplying a document _id in your source data, then and values of _id will fail to write because CouchDB will report a 409 Document Conflict. If you want your supplied data to supercede existing data then supply --overwrite true/-o true as a command-line option. This will instruct couchimport to fetch the existing documents' current _rev values and inject them into the imported data stream.

Note: Using overwrite mode is slower because an additional API call is required per batch of data imported. USe caution when importing data into a data set that is being changed by another actor at the same time.

Environment variables

• COUCH_URL - the url of the CouchDB instance (required, or to be supplied on the command line)
• COUCH_DATABASE - the database to deal with (required, or to be supplied on the command line)
• COUCH_DELIMITER - the delimiter to use (default '\t', not required)
• COUCH_TRANSFORM - the path of a transformation function (not required)
• COUCHIMPORT_META - a json object which will be passed to the transform function (not required)
• COUCH_BUFFER_SIZE - the number of records written to CouchDB per bulk write (defaults to 500, not required)
• COUCH_FILETYPE - the type of file being imported, either "text", "json" or "jsonl" (defaults to "text", not required)
• COUCH_JSON_PATH - the path into the incoming JSON document (only required for COUCH_FILETYPE=json imports)
• COUCH_PREVIEW - run in preview mode
• COUCH_IGNORE_FIELDS - a comma-separated list of field names to ignore on import or export e.g. price,url,image
• COUCH_OVERWRITE - overwrite existing document revisions with supplied data
• COUCH_PARALLELISM - the maximum number of HTTP requests to have in flight at any one time (default: 1)
• COUCH_MAX_WPS - the maximum number of write API calls to make per second (rate limiting) (default: 0 - no rate limiting)
• COUCH_RETRY - whether to retry requests which yield a 429 response (default: false)

Command-line parameters

You can also configure couchimport and couchexport using command-line parameters:

• --help - show help
• --version - simply prints the version and exits
• --url/-u - the url of the CouchDB instance (required, or to be supplied in the environment)
• --database/--db/-d - the database to deal with (required, or to be supplied in the environment)
• --delimiter - the delimiter to use (default '\t', not required)
• --transform - the path of a transformation function (not required)
• --meta/-m - a json object which will be passed to the transform function (not required)
• --buffer/-b - the number of records written to CouchDB per bulk write (defaults to 500, not required)
• --type/-t - the type of file being imported, either "text", "json" or "jsonl" (defaults to "text", not required)
• --jsonpath/-j - the path into the incoming JSON document (only required for type=json imports)
• --preview/-p - if 'true', runs in preview mode (default false)
• --ignorefields/-i - a comma-separated list of fields to ignore input or output (default none)
• --parallelism - the number of HTTP request to have in flight at any one time (default 1)
• --maxwps - the maximum number of write API calls to make per second (default 0 - no rate limiting)
• --overwrite/-o - overwrite existing document revisions with supplied data (default: false)
• --retry/-r - whether to retry requests which yield a 429 response (default: false)

e.g.

    cat test.csv | couchimport --database  bob --delimiter ","

couchexport

If you have structured data in a CouchDB or Cloudant that has fixed keys and values e.g.

{
    "_id": "badger",
    "_rev": "5-a9283409e3253a0f3e07713f42cd4d40",
    "wiki_page": "http://en.wikipedia.org/wiki/Badger",
    "min_weight": 7,
    "max_weight": 30,
    "min_length": 0.6,
    "max_length": 0.9,
    "latin_name": "Meles meles",
    "class": "mammal",
    "diet": "omnivore",
    "a": true
}

then it can be exported to a CSV like so (note how we set the delimiter):

    couchexport --url http://localhost:5984 --database animaldb --delimiter "," > test.csv

or to a TSV like so (we don't need to specify the delimiter since tab \t is the default):

    couchexport --url http://localhost:5984 --database animaldb > test.tsv

or to a stream of JSON:

    couchexport --url http://localhost:5984 --database animaldb --type jsonl

N.B.

• design documents are ignored
• the first non-design document is used to define the headings
• if subsequent documents have different keys, then unexpected things may happen
• COUCH_DELIMITER or --delimiter can be used to provide a custom column delimiter (not required when tab-delimited)
• if your document values contain carriage returns or the column delimiter, then this may not be the tool for you
• you may supply a JavaScript --transform function to modify the data on its way out

Using programmatically

In your project, add couchimport into the dependencies of your package.json or run npm install couchimport. In your code, require the library with

    var couchimport = require('couchimport');

and your options are set in an object whose keys are the same as the COUCH_* environment variables:

e.g.

   var opts = { delimiter: ",", url: "http://localhost:5984", database: "mydb" };

To import data from a readable stream (rs):

    var rs = process.stdin;
    couchimport.importStream(rs, opts, function(err,data) {
       console.log("done");
    });

To import data from a named file:

    couchimport.importFile("input.txt", opts, function(err,data) {
       console.log("done",err,data);
    });

To export data to a writable stream (ws):

   var ws = process.stdout;
   couchimport.exportStream(ws, opts, function(err, data) {
     console.log("done",err,data);
   });

To export data to a named file:

   couchimport.exportFile("output.txt", opts, function(err, data) {
      console.log("done",err,data);
   });

To preview a file:

    couchimport.previewCSVFile('./hp.csv', opts, function(err, data, delimiter) {
      console.log("done", err, data, delimiter);
    });

To preview a CSV/TSV on a URL:

    couchimport.previewURL('https://myhosting.com/hp.csv', opts, function(err, data) {
      console.log("done", err, data, delimiter);  
    });

Monitoring an import

Both importStream and importFile return an EventEmitter which emits

• written event on a successful write
• writeerror event when a complete write operation fails
• writecomplete event after the last write has finished
• writefail event when an individual line in the CSV fails to be saved as a doc

e.g.

couchimport.importFile("input.txt", opts, function(err,data) {
  console.log("done",err,data);
}).on("written", function(data) {
  // data = { documents: 500, failed:6, total: 63000, totalfailed: 42}
});

The emitted data is an object containing:

• documents - the number of documents written in the last batch
• total - the total number of documents written so far
• failed - the number of documents failed to write in the last batch
• totalfailed - the number of documents that failed to write in total

Parallelism & Rate limiting

Using the COUCH_PARALLELISM environment variable or the --parallelism command-line option, couchimport can be configured to write data in multiple parallel operations. If you have the networkbandwidth, this can significantly speed up large data imports e.g.

  cat bigdata.csv | couchimport --database mydb --parallelism 10 --delimiter ","

This can be combined with the COUCH_MAX_WPS/--maxwps parameter to limit the number write API calls dispatched per second to make sure you don't exceed the number writes on a rate-limited service.