littlstar / S3 Lambda

s3-lambda

s3-lambda enables you to run lambda functions over a context of S3 objects. It has a stateless architecture with concurrency control, allowing you to process a large number of files very quickly. This is useful for quickly prototyping complex data jobs without an infrastructure like Hadoop or Spark.

At Littlstar, we use s3-lambda for all sorts of data pipelining and analytics.

Disclaimer This module does not interact with the AWS Lambda service; the name s3-lambda is referring to lambda functions in computer science, and all s3 file processing happens locally.

Install

npm install s3-lambda --save

Quick Example

const S3Lambda = require('s3-lambda')

// example options
const lambda = new S3Lambda({
  accessKeyId: 'aws-access-key',       // Optional. (falls back on local AWS credentials)
  secretAccessKey: 'aws-secret-key',   // Optional. (falls back on local AWS credentials)
  showProgress: true,                  // Optional. Show progress bar in stdout
  verbose: true,                       // Optional. Show all S3 operations in stdout (GET, PUT, DELETE)
  signatureVersion: 'v4',              // Optional. Signature Version used in Authentication. Defaults to "v4"
  maxRetries: 10,                      // Optional. Maximum request retries on an S3 object. Defaults to 10.
  timeout: 10000                       // Optional. Amount of time for request to timeout. Defaults to 10000 (10s)
})

const context = {
  bucket: 'my-bucket',
  prefix: 'path/to/files/'
}

lambda
  .context(context)
  .forEach(object => {
    // do something with object
  })
  .then(_ => console.log('done!'))
  .catch(console.error)

Setting Context

Before initiating a lambda expression, you must tell s3-lambda what files to operate over by calling context. A context is defined with an options object with the following properties: bucket, prefix, marker, limit, and reverse.

lambda.context({
  bucket: 'my-bucket',       // The S3 bucket to use
  prefix: 'prefix/',         // The prefix of the files to use - s3-lambda will operate over every file with this prefix.
  marker: 'prefix/file1',    // Optional. Start at the first file with this prefix. If it is a full file path, starts with next file. Defaults to null.
  endPrefix: 'prefix/file3', // Optional. Process files up to (not including) this prefix. Defaults to null.
  match: /2017/i,            // Optional. Process files matching this regex / string. Defaults to null.
  limit: 1000,               // Optional. Limit the # of files operated over. Default is Infinity.
  reverse: false             // Optional. If true, operate over all files in reverse. Defaults to false.
})

You can also provide an array of context options, which will tell ls-lambda to operate over all the files in each.

const ctx1 = {
  bucket: 'my-bucket',
  prefix: 'path/to/files/',
  marker: 'path/to/logs/2017'
}
const ctx2 = {
  bucket: 'my-other-bucket',
  prefix: 'path/to/other/logs/',
  limit: 100
}

lambda.context([ctx1, ctx2])

Modifiers

After setting context, you can chain several other functions that modify the operation. Each returns a Request object, so they can be chained. All of these are optional.

.concurrency(c)

{Number} Set the request concurrency level (default is Infinity).

.exclude(e)

{Function} Sets the exclude function to use before getting objects from S3. This function will be called with the key and should return true if the object should be excluded.
Example: exclude png files

lambda
  .context(context)
  .exclude(key => /.png$/.test(key))
  .each(...)

.transform(f)

{Function} Sets the transformation function to use when getting objects. This transformer will be called with the raw object that is returned by the S3#getObject() method in the AWS SDK and the key, and should return the transformed object. When a transformer function is provided, objects are not automatically converted to strings, and the encoding parameter is ignored. Example: unzipping compressed S3 files before each operation

const zlib = require('zlib')

lambda
  .context(context)
  .transform((object) => {
    return zlib.gunzipSync(object.Body).toString('utf8')
  })
  .each(...)

.encode(e)

{String} Sets the string encoding to use when getting objects. This setting is ignored if a transformer function is used.

limit(l)

{Number} Limit the number of files operated over.

reverse(r)

{Boolean} Reverse the order of files operated over.

async()

Lets the resolver know that your function is async (returns a Promise).

Lambda Functions

Perform synchronous or asynchronous functions over each file in the set context.

• each
• forEach
• map
• reduce
• filter

each

each(fn[, isasync])

Performs fn on each S3 object in parallel. You can set the concurrency level (defaults to Infinity). If isasync is true, fn should return a Promise.

lambda
  .context(bucket, prefix)
  .concurrency(5) // operates on 5 objects at a time
  .each(object => console.log(object))
  .then(_ => console.log('done!'))
  .catch(console.error)

forEach

forEach(fn[, isasync])

Same as each, but operates sequentially, one file at a time. Setting concurrency for this function is superfluous.

lambda
  .context(bucket, prefix)
  .forEach(object => { /* do something with object */ })
  .then(_ => console.log('done!'))
  .catch(console.error)

map

map(fn[, isasync])

Maps fn over each file in an S3 directory, replacing each file with what is returned from the mapper function. If isasync is true, fn should return a Promise.

This is a destructive action, meaning what you return from fn will change the S3 object itself. For your protection, you must specify inplace() to map over the existing files. Alternatively, you can use output() to output the results of the mapper function elsewhere (as demonstrated below). You can pass a third argument (a function) to rename the output key (bucket + prefix).

const addSmiley = object => object + ':)'

lambda
  .context(bucket, prefix)
  .inplace()
  .map(addSmiley)
  .then(console.log('done!'))
  .catch(console.error)

Make this non-destructive by specifying an output directory.

const outputBucket = 'my-bucket'
const outputPrefix = 'path/to/output/'

lambda
  .context(bucket, prefix)
  .output(outputBucket, outputPrefix, (key) => key.replace('-', '/'))
  .map(addSmiley)
  .then(console.log('done!'))
  .catch(console.error)

reduce

reduce(func[, isasync])

Reduces the objects in the working context to a single value.

// concatonates all the files
const reducer = (previousValue, currentValue, key) => {
  return previousValue + currentValue
}

lambda
  .context(bucket, prefix)
  .reduce(reducer)
  .then(result => { /* do something with result */ })
  .catch(console.error)

filter

filter(func[, isasync])

Destructive. Filters (deletes) files in S3. func should return true to keep the object, and false to delete it. If isasync is true, func returns a Promise.

This is a destructive action, meaning if fn is false, the object will be deleted from S3. For your protection, you must specify inplace() to filter the existing files. Alternatively, you can use output() to output the results of the filter function elsewhere (as demonstrated below). As with map, you can pass a function to output to rename the output key.

// filters empty files
const fn = object => object.length > 0

lambda
  .context(bucket, prefix)
  .inplace()
  .filter(fn)
  .then(_ => console.log('done!'))
  .catch(console.error)

Make this non-destructive by specifying an output directory.

lambda
  .context(bucket, prefix)
  .output(outputBucket, outputPrefix, (key) => key.replace('-', '/'))
  .filter(filter)
  .then(console.log('done!'))
  .catch(console.error()

S3 Functions

Promise-based wrapper around common S3 methods.

• list
• keys
• get
• put
• copy
• delete

list

list(bucket, prefix[, marker])

List all keys in s3://bucket/prefix. If you use a marker, s3-lambda will start listing alphabetically from there.

lambda
  .list(bucket, prefix)
  .then(list => console.log(list))
  .catch(console.error)

keys

keys(bucket, prefix[, marker])

Returns an array of keys for the given bucket and prefix.

lambda
  .keys(bucket, prefix)
  .then(keys => console.log(keys))
  .catch(console.error)

get

get(bucket, key[, encoding[, transformer]])

Gets an object in S3, calling toString(encoding on objects.

lambda
  .get(bucket, key)
  .then(object => { /* do something with object */ })
  .catch(console.error)

Optionally you can supply your own transformer function to use when retrieving objects. This transformer will be called with the raw object that is returned by the S3#getObject() method in the AWS SDK, and should return the transformed object. When a transformer function is provided, objects are not automatically converted to strings, and the encoding parameter is ignored.

const zlib = require('zlib')

const transformer = object => {
  return zlib.gunzipSync(object.Body).toString('utf8')
}

lambda
  .get(bucket, key, null, transformer)
  .then(object => { /* do something with object */ })
  .catch(console.error)

put

put(bucket, key, object[, encoding])

Puts an object in S3. Default encoding is utf8.

lambda
  .put(bucket, key, 'hello world!')
  .then(console.log('done!')).catch(console.error)

copy

copy(bucket, key, targetBucket, targetKey)

Copies an object in S3 from s3://sourceBucket/sourceKey to s3://targetBucket/targetKey.

lambda
  .copy(sourceBucket, sourceKey, targetBucket, targetKey)
  .then(console.log('done!')).catch(console.error)

delete

delete(bucket, key)

Deletes an object in S3 (s3://bucket/key).

lambda
  .delete(bucket, key)
  .then(console.log('done!')).catch(console.error)