🐏 A CLI tool for processing data with functional pipelines
Latest commit 33e0ed7 Feb 21, 2018


ramda-cli npm version Gitter

A command-line tool for processing data with functional pipelines.

$ npm install -g ramda-cli; alias R=ramda
$ curl -L http://bit.do/people-json | R \
  'filter (p) -> p.city is /Port/ or p.name is /^Dr\./' \
  'project <[ name city mac ]>' \
  'take 3' \
  -o table --compact
β”‚ name             β”‚ city            β”‚ mac               β”‚
β”‚ Dr. Araceli Lang β”‚ Yvettemouth     β”‚ 9e:ea:28:41:2a:50 β”‚
β”‚ Terrell Boyle    β”‚ Port Reaganfort β”‚ c5:32:09:5a:f7:15 β”‚
β”‚ Libby Renner     β”‚ Port Reneeside  β”‚ 9c:63:13:31:c4:ac β”‚

Unites Ramda's curried, data-last API and LiveScript's terse and powerful syntax.

With a variety of supported input/output types and the ability pull any module from npm, ramda-cli is a potent tool for many kinds of data manipulation in command-line environment.

Table of Contents


npm install -g ramda-cli
alias R='ramda'


cat data.json | ramda [function] ...

The idea is to compose functions into a pipeline of operations that when applied to given data, produces the desired output.

By default, the function is applied to a stream of JSON data read from stdin, and the output data is sent to standard out as stringified JSON.

Technically, function should be a snippet of LiveScript (or JavaScript with --js) that evaluates into a function. If multiple function arguments are supplied as positional arguments, they are composed into a pipeline in order from left to right (see R.pipe).

As a simple example, in echo 1 | R inc 'multiply 2', which prints 4, the computation would look roughly as follows:

input = 1
fn = pipe(inc, multiply(2))
result = fn(input) // multiply(2, inc(1))
result             // => 4

All Ramda's functions are available directly in the scope. See http://ramdajs.com/docs/ for a full list.


# Add 1 to each value in a list
echo [1,2,3] | R 'map add 1'
# Add 1 to each value with inline ES6 lambda and take product of all
echo [1,2,3] | R --js 'map(x => x + 1)' product

Ramda functions used: add, map, product

Get a list of people whose first name starts with "B"
cat people.json | R 'pluck \name' 'filter (name) -> name.0 is \B)' -o raw
Brando Jacobson
Betsy Bayer
Beverly Gleichner
Beryl Lindgren

Ramda functions used: pluck, filter Data: people.json

Create a markdown TODO list
curl -s http://jsonplaceholder.typicode.com/todos |\
  R --raw-output \
    'filter where-eq user-id: 10' \
    'map (t) -> "- [#{t.completed && "x" || " "}] #{t.title}"' \
    'take 5' \


  • ut cupiditate sequi aliquam fuga maiores
  • inventore saepe cumque et aut illum enim
  • omnis nulla eum aliquam distinctio
  • molestias modi perferendis perspiciatis
  • voluptates dignissimos sed doloribus animi quaerat aut

Ramda functions used: filter, where-eq, map

List versions of npm module with dates formatted with node-timeago

It looks for ecto/node-timeago installed to $HOME/node_modules.

npm view ramda --json | R --import timeago \
  'prop \time' 'to-pairs' \
  'map -> version: it.0, time: timeago(it.1)' \
  -o tsv | column -t -s $'\t'
0.12.0    2 months ago
0.13.0    2 months ago
0.14.0    12 days ago
Search twitter for people who tweeted about ramda and pretty print the result
twarc.py --search '#ramda' | R --slurp -p 'map path [\user, \screen_name]' uniq

Ramda functions used: map, path

Pull response status data from Graphite and visualize

HTTP status codes per minute for last hour:

graphite -t "summarize(stats_counts.status_codes.*, '1min', 'sum', false)" -f '-1h' -o json | \
  R --import sparkline 'map evolve datapoints: (map head) >> sparkline \
    'sort-by prop \target' -o table


Ramda functions used: evolve, sortBy

Use --slurp to read multiple JSON objects into a single list before any operations
cat <<EOF | R --slurp identity
"foo bar"
"test lol"
"hello world"
  "foo bar",
  "test lol",
  "hello world"
Solution to the credit card JSON to CSV challenge using --output-type csv
#!/usr/bin/env bash

curl $data_url | R \
  'filter where creditcard: (!= null)' `# filter out those who don't have credit card` \
  'project [\name, \creditcard]'       `# pick name and creditcard fields from all objects` \
  -o csv > `date "+%Y%m%d"`.csv        `# print output as csv to a file named as the current date`
List a project's dependencies in a table
npm ls --json | R 'prop \dependencies' 'map-obj prop \version' -o table --compact
β”‚ JSONStream    β”‚ 1.0.4  β”‚
β”‚ treis         β”‚ 2.3.9  β”‚
β”‚ ramda         β”‚ 0.14.0 β”‚
β”‚ livescript    β”‚ 1.4.0  β”‚
β”‚ cli-table     β”‚ 0.3.1  β”‚

Ramda functions used: filter, where, project, mapObj, prop

Generate HTML with hyperscript

With hyperscript installed to ${HOME,PWD}/node_modules.

cat <<EOF > shopping.txt
cat shopping.txt | R --import h=hyperscript \
  -rR --slurp           `# read raw input into a list` \
  'map (h \li.item, _)' `# apply <li class="item"> into each item` \
  'h \ul#list, _'       `# wrap list inside <ul id="list">` \
  '.outer-HTML'         `# finally, grab the HTML`
<ul id="list">
  <li class="item">milk</li>
  <li class="item">cheese</li>
  <li class="item">peanuts</li>

Reason for underscores (e.g. h \ul, _) is that hyperscript API is not curried (and can't be because it's variadic). We need to explicitly state that this function is waiting for one more argument.

For more examples, see the Cookbook.


Usage: ramda [options] [function] ...

  -f, --file           read a function from a js/ls file instead of args; useful for
                       larger scripts
  -c, --compact        compact output for JSON and tables
  -s, --slurp          read JSON objects from stdin as one big list
  -S, --unslurp        unwraps a list before output so that each item is formatted and
                       printed separately
  -t, --transduce      use pipeline as a transducer to transform stdin
  -P, --json-path      parse stream with JSONPath expression
  -i, --input-type     read input from stdin as (#{format-enum-list INPUT_TYPES})
  -o, --output-type    format output sent to stdout (#{format-enum-list OUTPUT_TYPES})
  -p, --pretty         pretty-printed output with colors, alias to -o pretty
  -D, --pretty-depth   set how deep objects are pretty printed
  -r, --raw-input      alias for --input-type raw
  -R, --raw-output     alias for --output-type raw
  -n, --no-stdin       don't read input from stdin
      --[no-]headers   csv/tsv has a header row
      --csv-delimiter  custom csv delimiter character
      --js             use javascript instead of livescript
  -I, --import         require module as a variable
  -C, --configure      edit config in $EDITOR
  -v, --verbose        print debugging information (use -vv for even more)
      --version        print version
  -h, --help           displays help

-f, --file

Load a function pipeline from a file. Useful for scripts difficult to express in command-line.


// shout.js
var R = require('ramda');
module.exports = R.pipe(R.toUpper, R.add(R.__, '!'));
echo -n '"hello world"' | R --file shout.js

You can overwrite command-line arguments through the script by exporting a string in property opts.

module.exports = function() { /* ... */ }
module.exports.opts = '--slurp -o table'

-c, --compact

Print compact tables and JSON output without whitespace.

When used with --output-type raw, no line breaks are added to output.


seq 10 | R --input-type raw --output-type raw --compact identity # or -rRc

-s, --slurp

Read all input from stdin and wrap the data in a list before operations.


cat <<EOF | R --slurp 'map to-upper'

-S, --unslurp

After the pipeline is applied to an item and if the result is an array, its items are printed separately.


echo '[1,2,3]' | R --unslurp 'map inc'

-t, --transduce

Transform the input stream using the pipeline as a transducer. Requires all functions in the pipeline to be able to act as transducers.

This option essentially allows performing operations like R.map or R.filter on items as they come without waiting for the stream to complete or wrapping the input stream in a collection with --slurp.


echo '1 2 2 3 3 4' | R --transduce drop-repeats

-P, --json-path

Parse the input stream with given JSONPath expression.

See also: JSONStream documentation


Process a huge JSON array one by one without reading the whole thing first.

* as JSON path unwraps the array and objects are passed to identity one by one.

curl -Ls http://bit.do/countries-json | R --json-path '*' --compact identity
{"name":"Γ…land Islands","code":"AX"}

-i, --input-type

Parse stdin as one of these formats: raw, csv, tsv.


echo foo | R --input-type raw to-upper
$ cat <<EOF | R --input-type csv identity
  { "id": "1", "name": "Bob" },
  { "id": "2", "name": "Alice" }

-o, --output-type

Instead of JSON, format output as one of: pretty, raw, csv, tsv, table.

-o pretty

Print pretty output.

-o raw

With raw output type when a string value is produced, the result will be written to stdout as is without any formatting.

-o csv and -o tsv

CSV or TSV output type can be used when pipeline evaluates to an array of objects, an array of arrays or when stdin consists of a stream of bare objects. First object's keys will determine the headers.

-o table

Print nearly any type of data as a table. If used with a list of objects, uses the first object's keys as headers.


curl -Ls http://bit.do/countries-json | R 'take 3' -o table --compact
β”‚ name          β”‚ code β”‚
β”‚ Afghanistan   β”‚ AF   β”‚
β”‚ Γ…land Islands β”‚ AX   β”‚
β”‚ Albania       β”‚ AL   β”‚

-p, --pretty

Alias of --output-type pretty.

-D, --pretty-depth

When using pretty-printed output, set how deep structures are verbosely printed.

Useful when output is huge and you want to see the general structure of an object or list.

See documentation of util.inspect(object[, options])

-n, --no-stdin

Don't read stdin for input. Useful when starting a pipeline with a constant function.


R --no-stdin 'always "hello world"' 'add __, \!'
"hello world!"


Set if input csv/tsv contains a header row.

By default, csv/tsv input is assumed to contain headers.


Use a custom csv delimiter. Delimiter is comma by default.

Example: --csv-delimiter=';'


Interpret positional arguments as JavaScript instead of LiveScript.


echo '[1,2,3]' | R --js 'map(x => Math.pow(x, 2))'

-I, --import

Make modules installed with npm install accessible in positional code arguments.

Modules are looked for in $HOME/node_modules and current directory's node_modules.

Symbol = can be used to declare the variable module should appear as. Otherwise, module's name is used as the variable name.


echo test | R -rR --import c=chalk 'c.bold'

-C, --configure

Edit ramda-cli config file in $EDITOR.

See Configuration.

evaluation context


All of Ramda's functions are available, and also:

function signature description
id a β†’ a Alias to R.identity
flat * β†’ Object Flatten a deep structure into a shallow object
readFile filePath β†’ String Read a file as string
lines String β†’ [String] Split a string into lines
words String β†’ [String] Split a string into words
unlines [String] β†’ String Join a list of lines into a string
unwords [String] β†’ String Join a list of words into a string
then Function β†’ Promise Map a value inside Promise
pickDotPaths [k] β†’ {k: v} β†’ {k: v} Like R.pick but deep using dot delimited paths
renameKeysBy Function β†’ {k: v} β†’ {k: v} Like R.map but for keys instead of values


object description
process https://nodejs.org/api/process.html
console https://nodejs.org/api/console.html

process.exit() can be used to short-circuit pipeline in case of an error, for example:

  curl api | ramda 'tap (res) -> if res.error then console.error(res); process.exit(1)'

An alternative is to use Maybe type.


config file

Path: $HOME/.config/ramda-cli.{js,ls}

The purpose of a global config file is to carry functions you might find useful to have around. The functions it exports in an object are made available.

For example,

// ~/.config/ramda-cli.js
exports.date = (val) => new Date(val);
exports.timeago = require('timeago');
exports.debug = (val) => {
  console.log('debug:', val);
  return val;
echo 1442667243000 | R date debug timeago
debug: Sat Sep 19 2015 12:54:03 GMT+0000 (UTC)
"12 minutes ago"

default options

To make some options be passed by default, it is best to use a shell alias. For example:

# always interpret as javascript
alias R="ramda --js"
echo 1 | R '(x) => x + 1'
# always use identity function, instead of showing help without args
alias R="ramda identity"
echo 1 | R

using packages from npm

Packages installed to $HOME/node_modules or $PWD/node_modules can used with require().

date -v +7d +%s | R -rR --js 'require("moment").unix' 'x => x.fromNow()'
in 7 days

Additionally, there is -I, --import which can be used as a shorthand for require().

echo test | R -rR --import c=chalk 'c.bold'


Promise values are unwrapped at the end of pipeline.

then helper function can be used to map promise values.

echo 1 | R --js 'x => Promise.resolve(x)' 'then(add(5))'
echo '\ngoogle.com\nyahoo.com' | \
  R -r --js --import ping 'ping.promise.probe' 'then(omit(["output", "numeric_host"]))' | \
  R --slurp -o table --compact
β”‚ host        β”‚ alive β”‚ time    β”‚ min     β”‚ max     β”‚ avg     β”‚ stddev β”‚
β”‚ β”‚ true  β”‚ 1.325   β”‚ 1.325   β”‚ 1.325   β”‚ 1.325   β”‚ 0.000  β”‚
β”‚ google.com  β”‚ true  β”‚ 10.729  β”‚ 10.729  β”‚ 10.729  β”‚ 10.729  β”‚ 0.000  β”‚
β”‚ yahoo.com   β”‚ true  β”‚ 115.418 β”‚ 115.418 β”‚ 115.418 β”‚ 115.418 β”‚ 0.000  β”‚


You can turn on the debug output with -v, --verbose flag. Use -vv for even more verbose output.

Verbose output shows what entered LiveScript compiled to.

To debug individual functions in the pipeline, you can use something like treis.

echo 1 | R --import treis 'treis(add(1))'


LiveScript is a language which compiles to JavaScript. It has a straightforward mapping to JavaScript and allows you to write expressive code devoid of repetitive boilerplate.

comparison table

All expressions in the table evaluate to a function, and are valid in ramda-cli.

Ramda LiveScript JavaScript
not (not) x => !x
nth(0) (.0) x => x[0]
prop('name') (.name) x => x.name
add(1) (+ 1) x => x + 1
add(__, '!') (+ '!') x => x + '!'
gt(__, 2) (> 2) x => x > 2
contains(__, xs) (in xs) x => xs.includes(x)
pipe(length, gt(__, 2)) (.length > 2) x => x.length > 2
isNil (~= null) x => x == null
complement(isNil) (!~= null) x => x != null
match(/foo/) (is /foo/) x => x.match(/foo/)
replace('a', '') (- 'a') x => x.replace('a', '')
join(',') (* ',') x => x.join(',')
split(',') (/ ',') x => x.split(',')
toUpper (.to-upper-case!) x => x.toUpperCase()

See also: Essential LiveScript for ramda-cli

javascript support

With the release of node v4 came the ES6 arrow function syntax without a compilation penalty. This makes JS more attractive choice for ramda-cli, so now there is --js flag.

echo '[1,2,3]' | R --js 'map(x => x + 1)'

questions or comments?

wercker status