Saturday, 14 July 2018

Lighting Component Attributes - Expect the Unexpected

Lighting Component Attributes - Expect the Unexpected



A couple of weeks ago my BrightGen colleague, Firoz Mirza, described some unexpected (by us at least) behaviour when dealing with Lightning Component attributes. I was convinced that there must be something else going on so spent some time creating a couple of simple tests, only to realise that he was absolutely correct and attributes work a little differently to how I thought, although if I’m honest, I’ve never really thought about it.

Tales of the Unexpected 1

The basic scenario was assigning a Javascript variable to the result of extracting the attribute via component.get(), then calling another method that also extracted the value of the attribute, but crucially also changed it and set it back into the component. A sample component helper that works like this is:

    propertyTest : function(cmp) {
	var testVar=cmp.get('v.test');
        alert('Local TestVar = ' + JSON.stringify(testVar));
        alert('Local TestVar after leaving it alone = ' + JSON.stringify(testVar));
    changeAttribute : function(cmp) {
        var differentVar=cmp.get('v.test');
        cmp.set('v.test', differentVar);

The propertyTest function gets the attribute from the component and assigns it to testVar. After showing the value, it then invokes changeAttribute, which gets the attribute and assigns it to a new variable, updates it and then sets it back into the component.

Executing this shows the value of the property is empty, as expected to begin with:

Screen Shot 2018 07 14 at 13 49 09

but after the attribute has been updated elsewhere, my local variable reflects this:

Screen Shot 2018 07 14 at 13 50 27


Tales of the Unexpected 2

After a bit more digging, I came across something that I expected even less - updating a local variable that stored the result of the component.get() for the attribute, changed the value in the component. So I updated my sample component to test this:

    propertyTest2: function(cmp) {
	var test2=cmp.get('v.test');
        alert('Local test2 = ' + JSON.stringify(test2));
    checkAttribute: function(cmp) {
        var otherVar=cmp.get('v.test');
        alert('Other var = ' + JSON.stringify(otherVar));

Executing this shows my local property value being updated as expected:

Screen Shot 2018 07 14 at 14 01 51

but then retrieving this via another variable showed that it appeared to have been updated in the component:

Screen Shot 2018 07 14 at 14 02 28

“Maybe it’s just because everything is happening in the same request” I thought, but after this request completed I executed the first test, so extracting the value in a separate request (transaction) showed that the attribute had indeed been updated in a way that persisted across requests:

Screen Shot 2018 07 14 at 14 03 44

Checking the Docs

Convinced that I must have missed this in the docs, I went back to the Lightning Components Developer Guide, Working with Attribute Values in JavaScript section, but only found the following:

component.get(String key) and component.set(String key, Object value) retrieves and assigns values associated with the specified key on the component.

which didn’t add anything to what I already knew. 

Asking the Experts

I then asked the question of Salesforce, with my sample components, and received the following response:

The behavior you’re seeing is expected. 

component.set() is a signal to the framework that you’ve changed a value. The framework then fires change handlers, rerenders, and so forth. 

The underlying value is a JavaScript object, array, primitive, or otherwise. JavaScript object’s and arrays are mutable. Aura’s model requires that you send that signal when you mutate an object or array.


So there you have it - a local variable assigned the result of component.get() gives you a live hook to the component attribute, and if it’s mutable then any changes you make to the local variable update the attribute regardless of whether you call component.set() or not. Good to know!

I think the docs could certainly use with some additional information to make this clear. It’s written down here, at least, so hopefully more people will know going forward!

Related Posts



Saturday, 7 July 2018

Exporting Metadata with the Salesforce CLI

Exporting Metadata with the Salesforce CLI



Regular readers of this blog are all too aware that I’m a big fan of the Salesforce CLI. I rarely use anything else to interact with Salesforce orgs, outside of the main UI. Prior to this I used the CLI, which is another excellent tool, though didn’t give me quite enough information about operations to switch to it exclusively.

One area where the CLI still beats the default functionality of the Salesforce CLI is the export command, which pull back most of the metadata from an org (except reports, dashboards, and email templates, where you have to jump through a few hoops to get folder names first). Switching back to the CLI to take a backup of an org started to get a bit repetitive, so I decided to write something to replicate the functionality using the Salesforce CLI.

To Plugin or Not to Plugin

After some consideration, I decided not to plugin. I need to do some work and then execute a deployment via the Salesforce CLI, and there’s currently no way to execute a command from a plugin. I could shell out of the plugin and execute the CLI as a new child process, but that seemed a bit clunky. Plus where does it end? We’d end up with plugins creating child sfdx commands that were plugins that created child sfdx commands - a house of cards made of processes and just as fragile. I’d also seen Dave Carroll call this approach an anti-pattern on twitter and I didn’t want him mad at me, so I went with a node script that wraps everything up into a single command from the users point of view.

The Node Code

As I plan to do more of this sort of thing, rather than having loads of different scripts to execute I went with one script that offers subcommands. Kind of like a CLI all of it’s own, but very lightweight.

In order to support subcommands I used the commander module - I’ve used this a few times in the past and it’s really straightforward. I just configure it with the subcommands and their options and it figures out which one to execute and parses the arguments for me:

var exportCmd=program.command('export')
		 .option("-u, --sfdx-user [which]", "Salesforce CLI username")
		 .option("-d, --directory [which]", "Output directory")
                 .action(function(options) {new ExportCommand(options).execute()});


The export command is responsible for generating a package.xml manifest file containing details of all the metadata to retrieve and then executing a metadata API retrieve.


It contains a list of the names metadata components to extract:

const allMetadata=[
... "ValidationRule", "Workflow" ];

The one wrinkle here is the standard objects. If I wildcard the CustomObject metadata type, this will just retrieve my custom objects. To include the standard objects, I have to explicitly name each of them. Luckily the Salesforce CLI provides a way for me to extract these - force:schema:sobject:list, so I can pull these back and turn them into a list of names:

var standardObjectsObj=child_process.execFileSync('sfdx', 
                        '-u', this.options.sfdxUser, 
                        '-c', 'standard']);

var standardObjectsString=standardObjectsObj.toString();

for (var idx=standardObjects.length-1; idx>=0; idx--) {
    if ( (standardObjects[idx].endsWith('__Tag')) ||
         (standardObjects[idx].endsWith('__History')) ||
         (standardObjects[idx].endsWith('__Tags')) ) {
         standardObjects.splice(idx, 1);

Then I can generate the package.xml by iterating the list of metadata names and adding a wildcard entry for each of them, and append the standard objects when the metadata name is CustomObject:

for (var idx=0; idx<allMetadata.length; idx++) {
    var mdName=allMetadata[idx];
    if (mdName=='CustomObject') {
        for (var soIdx=0; soIdx<standardObjects.length; soIdx++) {

Once the package.xml is written, I can then execute a force:mdapi:retrieve to pull back all of the available metadata:

                    '-u', this.options.sfdxUser, 
                    '-k', this.packageFilePath]);

Export all the Metadata!

(The full source code is available at the Github repo. If you want to try this out yourself, clone the repo, change into the directory containing the clone and run npm install to pull down the dependencies.)

To export the metadata I execute:

node index.js export -u -d output

Where -u specifies the username that I’ve previously authorised with the Salesforce CLI (I’m using the client dev org for my training system), and -d is the directory where the package.xml will be created and the metadata extracted to. If this directory doesn’t exist it will be created.

The output is just a couple of lines telling me it succeeded:

Exporting metadata
Metadata written to output/

The contents of the output subdirectory shows the package.xml and zip file retrieved from Salesforce:

$ ls -lrt
total 928
-rw-r--r-- 1 kbowden staff 20110 7 Jul 12:49 package.xml
-rw-r--r-- 1 kbowden staff 453236 7 Jul 12:50

unzipping the file creates an unpackaged directory containing the retrieved metadata:

$ unzip
inflating: unpackaged/labels/CustomLabels.labels

$ cd unpackaged
$ ls -l
total 48
drwxr-xr-x 16 kbowden staff 512 7 Jul 12:52 applications
drwxr-xr-x 4 kbowden staff 128 7 Jul 12:52 assignmentRules
drwxr-xr-x 7 kbowden staff 224 7 Jul 12:52 tabs
drwxr-xr-x 4 kbowden staff 128 7 Jul 12:52 triggers

and hey presto, my org is exported without having to revert back to the CLI.

Related Posts