Search
  • Heidi Lauberger

Passing input data through Map/Reduce stages

In order to make multi-threading work for a “classless” programming language such as JavaScript, SuiteScript 2.0 Map/Reduce scripts employ a stringent way how data needs to be fed into map and reduce stages.


Further to NetSuite Help Centre Section Map/Reduce Entry Points (https://system.eu1.netsuite.com/app/help/helpcenter.nl?fid=section_4387799161.html#subsect_1518300127), I would like to share some key findings on our end:

  • Why not making it an object every time? Indeed, with Map/Reduce script, I have come to wrap up input parameters as objects whenever in doubt. Just wrap up all your data in an object objInput and pass it into the map stage: var objInput = {}; objInput[key] = {}; // Map will process these child objects on nested level 1 separately. Think carefully on how you want to split your data at this point. objInput[key][property1] = Property 1; objInput[key][property2] = Property 2; objInput[key][‘transdata’] = {}; objInput[key][‘transdata’][tranproperty1] = Transaction Property 1; objInput[key][‘transdata’][tranproperty2] = Transaction Property 2; return objInput;In the map stage, you can subsequently reconcile your data easily as follows: var keyMap = context.key; var objMapInput = JSON.parse(context.value); var objInput = {}; objInput[keyMap] = objMapInput; // If you feel likeLikewise, in order to push your data into the Reduce stage, you format your data in an object and push it into the Map/Reduce context object as follows: for ( var keyId in objRedInput ) { if ( !isEmptyObj(objRedInput) ) { context.write({ key: keyId, value: objRedInput[keyId] }); } }In the Reduce stage, you subsequently need to be careful for two reasons:

  1. The content of your input object objRedInput is stored in an array called context.values (not context.value).

  2. Presently, there appears to be a limitation wherein you cannot parse the entire array byvar objRedInput = JSON.parse(context.values);as this will return a value “L”, a reference to the underlying Java object.

Instead you can process your data as follows var keyRed = context.key; if ( context.values.length > 0 ) { var redInput = new Array; for ( var i = 0 ; i < context.values.length ; i++ ) { redInput.push(JSON.parse(context.values[i])); } } else { var redInput = JSON.parse(context.values[0]); } var objInput = {}; objInput[keyRed] = redInput; // If you feel like The key here is to loop through the array elements and parse them individually.

  • N/cache is said not to work well with Map/Reduce. I hope this will change so I don’t need to call on script deployment parameters several times in the map and reduce stage.

  • Generate output in Suitelets. This will allow to leverage on more functionality, including jQuery for formatting, which is not available in Map/Reduce. Currently, I am trying to introduce N/cache to park larger summarize input objects in the cache instead of passing it through the Suitelet deployment URL

5 views

Recent Posts

See All

E-mail for contacts

ABOUT US

SOLUTIONS

SERVICES

PRODUCTS

DOWNLOAD PRESENTATION

PDF, 3MB

icon-map-google-01.png

Office Location

Headquarters

cloudmaven_logo_1920x1920_transparent.pn

© 2020 Cloudmaven - Terms and Conditions - Impressum