Pages - Menu

Bitbucket Continuous Integration with Bitbucket Pipelines


We use Bitbucket as our SCM and other few Atlassian products in our development team. Happy to say that I am pleased with the tools that it made our daily works very productive and hassle free. 

Recently, I am looking into a CICD solution for our deployment process. I have already previously Setting Up Jenkins for GitHubSetting Up Octopus Deploy for Jenkins with nopCommerce Projects or Setup Continuous Integration with Visual Studio Online. I like the design of Octopus Deploy, but given we are now happily living with our Atlassian suite, I want to see what the Bitbucket Pipelines (formerly Bamboo) can offer me. 

I just signed up for the Bitbucket Pipelines Beta and already received the beta invitation within about 6 hours :)


  • Setup and integrate Bamboo with Bitbucket.
  • Explore possibility to automatically run the Demandware Grunt Build Suite when we commit to master.


I clicked the link to install the Bitbucket Pipelines add-on.
Went through a simple setup procedure and enabled it in my repository.

It adds a Pipelines link in my repository.

Going through the wiki to configure the bitbucket-pipelines yml.

It is essential to understand the Docker Image used by Bitbucket pipelines. The default image uses ubuntu and is good enough in our scenario.

In the setting tab, we are able to setup environment variables for username and password.

We created a bitbucket-pipelines.yml script in the root folder.

Check-ins to the master will trigger the Bamboo to run the build job and script.


I have not had much chance to explore further, but it looks quite promising and does what I wanted. I like the fact that it is tightly integrated with our bitbucket so I do not need to login to another service, the setup was simple and straightforward.

"Build servers build" - Bitbucket is a CI server and not a CICD server. We will leverage Bitbucket Pipelines for our continuous integration and our Demandware Business Manager Code Replication for deployment purpose. I am pleased with the Bamboo Pipelines, this should work well and fits our CICD strategy.

Demandware - Implementing 3rd party FTP Service for XML Drop


I developed a few Demandware cartridges that export xml via ftp for 3rd party integration. A few things that I have done such as:-

  • Piggy back a ftp drop task to the current job schedule that imports xml from a difference integration point
  • Using standard Demandware pipelet to generate a xml feed for ftp drop
  • Create a custom script node to generate a xml feed for ftp drop
It wasn't too difficult as a job, but I found a few hidden gotchas that I believe are noteworthy.

We are dealing with the latest Demandware version 16.7.


FTP Constraints

  • Ftp connection is established using passive transfer mode (PASV) only.
  • Demandware only support ftp and sftp for backend integration, no ftps.
  • Use FTPClient for ftp; and SFTPClient for sftp.


In our code, we use dw.util.ArrayList to filter out products for export. We ran into an error where the ArrayList cannot be bigger than 20,000 in size.

Front End Debugging

I was testing out a pipeline-startnode by direct url from a broswer, it did not turn out well because some of the pipelets can only be executed from the backend. The error message looks like this.

Limit for quota 'api.pipelet.ImportExport' exceeded. Limit is 0, actual is 1.

Job Schedule

We tried to schedule a job for our cartridge and learnt in a hard way that JS Controller is not supported for Business Manager Job Schedule. Only pipeplines can be used.

Traversing XML

If you happened in need of traversaing the xml, Demandware uses ECMA scripts for XML. Here is a quick start guide.


Besides the firewall rules that I needed to manage with our external vendor, we also need to request outbound firewall rule from Demandware.

As they pointed out in their doc, "Before you can make an outbound FTP connection, the FTP server IP address must be enabled for outbound traffic at the Demandware firewall for your POD. Please file a support request to request a new firewall rule".

From Demandware, there are steps to request a Firewall rule to allow outgoing connections to 3rd party services.

Demandware - Monogram by using Product Options


Our business wants to start selling personalized product by monogramming. In this particular promotion, customer gets a free monogrammed luggage tag when they buy a bag.


In Demandware, this can be achieved by using Product Options. This is the recommended way to deal with monogram in Demandware. As they stated in their document, "product options enable you to sell configurable products that have optional accessories, upgrades...".

We can set up product options by going into Business Manager. Merchant Tools -> Products and Catalogs -> Product Options.

Create a New Option and put in relevant details.

In our scenario, we have allowed up to 3 initials for our monogram. Due to Demandware limitations and our UI design, we decided to store the values in 3 different product options. Each option stores A-Z and a blank, it can vary base on your technical and UI requirement.

We will then create the System Object Attribute to store the Product Options.

In Administration -> Site Development -> System Object Types, we will add new attributes for Product.


Monogramming options can be set at per color level. Therefore, in our product detail page, we are now able to sell our chocolate color bag with monogramming options but not our black color bag.

And the monogram information is available in our mini cart, cart, checkout and all the way through to our order summary page, as well as the order confirmation email that we sent to our customers.


There were a few discussions, considerations and research done within our team when we were planning for the monogram features for our products. Once we read enough documents and especially the Demandware documents, and after a few try and error in our sandbox, things get straightened out and the solution became obvious. We are happy with the implementations because it uses the native Demandware features and less customizations we have to put in our cartridge.

Demandware - Get Active and Upcoming Promotions


We are trying to get a collection of promotions that are either active or upcoming. We want to pre-calculate all our promotional price in advance for our affiliate feeds, so they will get the most updated price without frequent feeds from us.

At the time of writing, we are using the latest DemandWare 16.7 API.


In the DemandWare API, there is no such thing about getting active and upcoming promotions. Ideally we would like to return a collection combining the following.


Since these 2 methods return an object PromotionPlan and there is no supporting methods such as promotionPlanA.Add(promotionPlanB), so we will need to write a bit more code to achieve what we want.

function getActiveAndUpcomingPromotions() : Collection
 var promos = PromotionMgr.getPromotions().iterator();
 var result : ArrayList = new ArrayList();
 var now : Date = new Date();
 while (promos.hasNext())
  var promo =;
  if ( 
  || (promo.startDate != null 
    && promo.startDate > now 
    && (promo.endDate == null || promo.endDate > now)))
 return result;


  • Depending on number of promotions in the system, this can be an expensive operation to loop through all the promotions, so it is recommended to use local caching for the collection.
  • Our custom code is better than the original PromotionMgr.getUpcomingPromotions(previewTime : Number) because it was not possible to retrieve ALL future promotions where the previewTime is compulsory.
  • Our custom code only return Collection of Promotion, this is not the same as a PromotionPlan.
  • Beware of the ArraySize Limit 20,000 in DemandWare. Pushing more than 20,000 active promotions to the array will throw exception.

Demandware - Pipeline or start node can't be found.


I created a new cartridge called int_fusion_factory. When I tried to run the Pipeline-Start Node pair with a Job Schedule, I am getting the Pipeline or start node can't be found error.


In Business Manager -> Administration -> Sites -> Manage Sites, there are separated effective cartridge path settings for Business Manager.

We need to add our new cartridge to the Business Manager Path.

Git - How to Handle Emergency Deploy to Live Environment

Master always have our latest changes in our development and could be dirty and not production worthy. In order for us to do an emergency stable deploy to our live environment, it can be achieved as follow.

  1. Suppose we have our Master M. On our last well known tag called mytag deployed to our live environment.
  2. Subsequently C1 and C2 changes are commited to our Master.
  3. And, we have an emergency deployment required for C3 change that needs to be pushed to live with out C1 and C2.
This can be achieved by the following.

# branch out to master-e from mytag
$ git checkout -b master-e mytag

# After committing change C3 to our new branch, we are then good for live deployment.
# Last thing to do is to apply C3 to Master Origin, so that our next deploy will have C3
# This can be achieved by a simple merge.
$ git checkout master
$ git merge master-e

The new branch master-e is then kept until the next master deploy, if we need other emergency deploys before our next master deploy, we can reuse master-e.

Demandware - Converting Multi-steps Checkout to One Page Checkout


Recently, we had a mini project to replace our multi-step checkout page into a new one-page checkout.

Our last checkout process consist of 4 steps. After items are added to the bag, the checkout process as follow.
  1. Login or Checkout as Guest
  2. Shipping
  3. Billing
  4. Confirm 

As observed, it is clunky and takes a lot of reading, typing and waiting before one can reach to the end and place an order. In our new design of one page checkout, we optimized all our intermittent calls in ajax so that the overall user experience is smoother. Using only one page to taking care of what we previously can only be achieved by using 4 checkout screens.


There were a few challenges and stress points in this project. 
  • The consequence of breaking a check out page is high.
  • Not many people have done an one page checkout in Demandware.
  • I was getting a lot of 500 internal errors thrown within Demandware and there were no logging or error messages available
  • Debugger tools doesn't stop at break points when 500 error thrown.
  • Our team is relatively new in Demandware.
  • And I am relatively new in Demandware too.

Technical Overview


We created a new pipeline to handle our new checkout. Some of the intermittent calls are either reusing existing pipelines or newly created.

Firstly and noticeably in our new pipeline, that the end journey is no longer an Interaction Continue Node nor Jump Node. It is replaced by an Interaction Node. This is because in our isml template, we use a bunch of <iscomponent> to do dynamic include of different checkout steps onto the one page, and these pages will have their own ajax post back.

While we are consolidating our new checkout process, we are also trying not to lose any existing functionalities. For example in our Start Node, we will call the old checkout pipelines like COCustomer-PrepareCheckout to make sure our form is prepared for the checkout process like we used to, then we will call our customized OPCheckout-PrepareCheckout pipeline for any extra functionalities. The pdicts are also updated as we post back our form.


There were about 120 files changed in our source code. We only created one file for our new pipeline, but still many new files due to template and css changes. We also retrospectively changed some of the exit points in some of our existing call nodes, they are now returning a proper error end node to the calling pipeline, so that all the jump nodes are now managed in one main pipeline only. No more yelling and screaming why an user redirected to an unexpected jump node again. :)


Our conversion rate slightly increased and our checkout funnel drop rate also moderately decreased. Although we have not yet received enough data sampling to conclude anything at this stage, but we are very confident and positive with the new checkout. And, we already receiving some good feed backs from people about the new UX design, which is a winning!!