Data Layer Helper Library
This library provides the ability to process messages passed onto a data layer queue.
- Quick Start
- Background
- Why Do We Need a Library?
- The Abstract Data Model
- Listening for Messages
- Summary
- Build and Test
- License
Quick Start
First, install the package with your favorite package manager:
npm install data-layer-helper
// or
yarn add data-layer-helper
Next, import the code into your javascript:
import 'node_modules/data-layer-helper/dist/data-layer-helper';
For the development version (bigger file size, but reports possible errors to the console):
import 'node_modules/data-layer-helper/dist/data-layer-helper-test-debug';
What is a Data Layer Queue?
A data layer queue is simply a JavaScript array that lives on a webpage.
<script>
dataLayer = [];
</script>
Page authors can append messages onto the queue in order to emit information about the page and its state.
<script>
dataLayer.push({
title: "Migratory patterns of ducks",
category: "Science",
author: "Bradley Wogulis"
});
</script>
These messages are simply JavaScript objects containing a hierarchy of key/value pairs. They can be metadata about the page content, information about the visitor, or data about events happening on the page. This system allows tools like analytics libraries and tag management systems to access this data in a standard way, so page authors can avoid using a bunch of proprietary, repetitive APIs.
- It provides a common, well defined system for exposing page data.
- It doesn't slow down page rendering.
- It doesn't pollute the global JavaScript namespace.
- It doesn't require page authors to learn a different, one-off API for every new tool.
- It doesn't require page authors to expose the same data multiple times.
- It allows page authors to add, remove or change vendors easily.
Why do we need a Library?
The data layer system makes things very easy for page authors. The syntax is simple, and there's no extra code to load. But this system does make life a little more difficult for vendors and tools that want to consume the data. That's where this library comes in.
This project provides the ability to listen for data layer messages and to read the key/value pairs that have been set by all the previous messages. It can be used by the tools/vendors mentioned above, or by page authors that need to read back the data they've emitted.
To use this library, you'll need to get it onto the page. The easiest way to do this is by following these instructions. Once it's on the page, you can create a new helper object like this:
const helper = new DataLayerHelper(dataLayer);
This helper object will listen for new messages on the given data layer. Each new message will be merged into the helper's "abstract data model". This internal model object holds the most recent value for all keys which have been set on messages processed by the helper.
You can retrieve values from the data model by using the helper's get()
method:
helper.get('category'); // Returns "Science".
As mentioned above, messages passed onto the data layer can be hierarchical. For example, a page author might push the following message, which has data multiple levels deep:
dataLayer.push({
one: {
two: {
three: 4
}
}
});
Using the helper, you can retrieve the nested value using dot-notation:
helper.get('one.two.three'); // Returns 4.
helper.get('one.two'); // Returns {three: 4}.
The Abstract Data Model
As mentioned above, the abstract data model is an internal representation, which holds the most recent value for all keys that have been set by a data layer message. This means that as each message is pushed onto the data layer, the abstract data model must be updated. The helper library does this using a well-defined process.
As each message is processed, its key/value pairs will be added to the abstract data model. If the key doesn't currently exist in the model, this operation is simple. The pair is simply added to the model object. But in the case of key conflicts, we have to specify how values will be overwritten and/or merged.
There are two possible actions to take when merging a key/value pair onto the abstract model; overwriting the existing value or recursively merging the new value onto the existing value. The action taken will depend on the type of the two values unless explicitly overriden. For this, we define three types of values:
- JavaScript Arrays
- "Plain" Objects
- Everything else
Hopefully, JavaScript Arrays are self-explanatory. "Plain" Objects are JavaScript
objects that were created via Object literal notation (e.g. {one: 2}
) or via new Object
.
Nulls
, Dates
, RegExps
, Windows
, DOM Elements
, etc. are not "Plain". Those fall into the
category of "everything else", along with strings
, numbers
, booleans
, undefined
, etc.
Once the type of the new and existing values has been categorized this way, we can use the following table to describe what action will happen for that key/value pair:
Existing Value | New Value | Default Merging Action |
---|---|---|
Array | Array | Recursively merge |
Array | Plain Object | Overwrite existing value |
Array | Other | Overwrite existing value |
Plain Object | Array | Overwrite existing value |
Plain Object | Plain Object | Recursively merge |
Plain Object | Other | Overwrite existing value |
Other | Array | Overwrite existing value |
Other | Plain Object | Overwrite existing value |
Other | Other | Overwrite existing value |
Overwriting Existing Values
When the merging action is "Overwrite exsting value", the result of the operation is very simple. The existing value will be completely discarded and the new value will take its place in the abstract data model. The following table provides some examples:
Existing Value | New Value | Result of Overwrite |
---|---|---|
[1, 2, 3] | 'hello' | 'hello' |
{ducks: 'quack'} | [1, 2, 3] | [1, 2, 3] |
{ducks: 'quack'} | 'hello' | 'hello' |
'hello' | [1, 2, 3] | [1, 2, 3] |
'hello' | {ducks: 'quack'} | {ducks: 'quack'} |
'hello' | 42 | 42 |
Recursively Merging Values
When the merging action is "Recursively Merge", the result of the operation will be the result of iterating through each property in the new value, and for each property, deciding how to copy that sub-key/value into the abstract data model by looking at the type of that sub-key in the existing value. If the key does not exist on the existing value, the new value is simply assigned to the abstract model. The following examples demonstrate this:
Existing Value | New Value | Result of Overwrite |
---|---|---|
{one: 1, three: 3} | {two: 2} | {one: 1, three: 3, two: 2} |
{one: 1, three: 3} | {three: 4} | {one: 1, three: 4} |
{one: {two: 3}} | {one: {four: 5}} | {one: {two: 3, four: 5}} |
{one: {two: 3}} | {two: 4} | {one: {two: 3}, two: 4} |
[] | ['hello'] | ['hello'] |
[1] | [undefined, 2] | [undefined, 2] |
[1] | [(empty), 2] | [1, 2] |
[1, {two: 3}] | [undefined, {two: 4, six: 8}] | [undefined, {two: 4, six: 8}] |
[1, {two: 3}] | [(empty), {two: 4, six: 8}] | [1, {two: 4, six: 8}] |
Notice that an index in a new value array that has been explicitly set to undefined will overwrite the corresponding index in the existing array, however an index that has not been set to any value (i.e. empty values in a sparse array) will not overwrite the corresponding index in the existing array, even though value at both indexes evaluates to undefined.
Preventing Default Recursive Merge
Occasionally you may want to avoid persisting values from earlier in the application state. This is especially true for single page applications where you may not want outdated information in the data model when routing between pages.
To prevent the default recursive merging behavior, a flag can be passed in adjacent to the object(s) or array(s) you wish to prevent merging. To do so, add a truthy '_clear' attribute to the message with the key/value pair(s) you are targeting. Here are some examples:
Existing Value | New Value | Result of Overwrite |
---|---|---|
{a: [1]} | {a: [], _clear: true} | {a: []} |
{a: {x: 1}} | {a: {}, _clear: 1} | {a: {}} |
{a: [undefined, 2]} | {a: [1], _clear: true} | {a: [1]} |
{a: {x: undefined, y: 2}} | {a: {x: 1}, _clear: true} | {a: {x: 1}} |
{one: {two: {three: 3}}, five: [1, 2]} | {one: {two: {four: 4}}, five: [3], _clear: true} | {one: {two: {four: 4}}, five: [3]} |
{one: {two: {three: 3}}, five: [1, 2]} | {one: {two: {four: 4}, _clear: true}, five: [3]} | {one: {two: {four: 4}}, five: [3, 2]} |
Meta Commands
Using the above methods alone, some operations on the abstract model are somewhat cumbersome. For example, appending items onto the end of an existing array requires you to know the length of the existing array and then requires you to clumsily build an array that can be merged onto the existing value. To make these cases easier, we provide a set of alternative syntaxes for updating values that are already in the abstract data model.
The first of these syntaxes allows you to call any method supported on the existing type.
For example, if the existing value in the abstract model is an Array
, you'd have a wide
variety of APIs that can be called (e.g. push
, pop
, concat
, shift
, unshift
, etc.). To invoke
this syntax, you would push a "command array" onto the data layer instead of a normal message
object.
dataLayer.push(['abc.push', 4, 5, 6]);
A command array is a normal JavaScript array, where the first element is a string. The string
contains the key of the value to update, followed by a dot (.), followed by the name of the
method to invoke on the value. In the above example, the key to update is abc
, and the
method to invoke is the push
method. The string may be followed by zero or more arguments,
which will be passed to the invoked method.
If the given method name does not exist on the existing value, or if the invocation throws an exception, the assignment will be ignored, and a warning message will be logged to the browser's developer console (if available).
Native Methods
Browsers come with dozens, if not hundreds, of useful APIs. Any method supported by the existing value can be called using the command array syntax. Here are some additional examples:
Existing Key: | abc |
Existing Value: | [1, 2, 3] |
Command Array: | dataLayer.push(['abc.push', 4, 5, 6]) |
Result: | [1, 2, 3, 4, 5, 6] |
In the following example, no arguments are provided:
Existing Key: | abc |
Existing Value: | [1, 2, 3] |
Command Array: | dataLayer.push(['abc.pop']) |
Result: | [1, 2] |
In the following example, the value to update (bbb
) is nested inside a top level object (aaa
):
Existing Key: | aaa.bbb |
Existing Value: | [1, 2, 3] |
Command Array: | dataLayer.push(['aaa.bbb.push', 4]) |
Result: | [1, 2, 3, 4] |
And the following example demonstrates an operation on a Date object. Remember that all types are supported, not just Arrays:
Existing Key: | time |
Existing Value: | Fri Dec 20 2013 15:23:22 GMT-0800 (PST) |
Command Array: | dataLayer.push(['time.setYear', 2014]) |
Result: | Fri Dec 20 2014 15:23:22 GMT-0800 (PST) |
Notice that because command arrays are processed asynchronously, nothing can be done with the return values from these method invocations. This brings us to our second syntax for updating values in the abstract data model.
Custom Methods
So far, we've seen that objects (messages) can be pushed onto the data layer, as well as arrays
(command arrays). Pushing a function onto the data layer will also allow you to update the abstract
data model, but with custom code. This technique has the added benefit of being able to handle
return values of any native method calls made from within the function. When a function
is processed, the value of this
will be the abstract data model interface, described below.
The Abstract Data Model Interface
To safely access the abstract data model from within a custom method, an
API with a getter and setter is provided. get(key)
will get a key of the model,
and set(key, value)
will create or overwrite the given key with the new value.
The following examples demonstrate how these APIs can be used to update values in the abstract data model.
Existing Key: | time |
Existing Value: | Fri Dec 20 2013 15:23:22 GMT-0800 (PST) |
Custom function: | dataLayer.push(function() { this.get('time').setMonth(0); }) |
Result: | Fri Jan 20 2013 15:23:22 GMT-0800 (PST) |
The following example demonstrates updating a nested value:
Existing Key: | aaa.bbb.ccc |
Existing Value: | [1, 2, 3] |
Custom function: | dataLayer.push(function() { const ccc = this.get('aaa.bbb.ccc'); ccc.push(ccc.pop() * 2); }) |
Result: | [1, 2, 6] |
The following example demonstrates overwriting a value:
Existing Key: | abc |
Existing Value: | [1, 2, 3] |
Custom function: | dataLayer.push(function() { this.set('abc', {xyz: this.get('abc')}); }) |
Result: | {xyz: [1, 2, 3]} |
Listening for Messages
When creating a DataLayerHelper
object, you can also specify a callback function to be called
whenever a message is pushed onto the given dataLayer. This allows your code to be notified
immediately whenever the dataLayer has been updated, which is a key advantage of the message
queue approach.
Listener functions are attached to a data layer helper, not the data layer itself. As such, they should not modify the dataLayer (e.g. call dataLayer.push()), as this may cause other helpers listening to the data layer to see conflicting history.
With a Listener Function
To add a callback that listens to every push to the data layer,
Add a function under the listener
attribute in the second parameter of the DataLayerHelper
constructor.
function listener(model, message) {
// Message has been pushed.
// The helper has merged it onto the model.
// Now use the message and the updated model to do something.
}
const helper = new DataLayerHelper(dataLayer, {listener: listener});
Listening to the Past
Tools that are loaded onto the page asynchronously or lazily will appreciate that you can also
opt to listen to message that were pushed onto the dataLayer in the past. To do so, mark the
listenToPast
attribute as true in an options object for the second parameter of the DataLayerHelper
constructor.
function listener(model, message) {
// Message has been pushed.
// The helper has merged it onto the model.
// Now use the message and the updated model to do something.
}
const helper = new DataLayerHelper(dataLayer, {
listener: listener,
listenToPast: true,
});
Using this option means that your listener callback will be called once for every message that has ever been pushed onto the given data layer. And on each call to the callback, the model will represent the abstract model at the time of the message.
By Registering Processors
If you are using a command API to bring messages to the data layer, you may want to run different processors to respond to different commands pushed to the data layer instead of just having one global listener for all commands. You can register a function to run whenever commands with a specific name are pushed to the command API. The function can take any number of arbitrary arguments, and within the function, the value of this will be the abstract data model interface The return value of the function will be merged into the model following the rules in recursively merging values.
First, set up a commandAPI
function
const dataLayer = [];
const helper = new DataLayerHelper(dataLayer);
function commandAPI() {
dataLayer.push(arguments);
}
Next, use the registerProcessor
function.
helper.registerProcessor('add', function(number1, number2) {
// The return value will be merged into the model.
return {sum: number1 + number2};
});
// Important: to access the model using this,
// registered processors must not be arrow functions.
helper.registerProcessor('copy', function() {
const sum = this.get('sum');
// We could also do this.set('ans', sum), but changing
// the model inside of a registered processor is discouraged.
return {ans: sum};
});
// If multiple functions are registered with the same key, they
// will be called in the order that they have been registered. Hence,
// this function will be called second. However, it still will not
// update finalAns on the first call, because updates from return values
// won't be merged into the model until all functions have been called.
helper.registerProcessor('copy', function() {
const ans = this.get('ans');
return {finalAns: ans};
});
The above functions won't be called until we call commandAPI
with the first parameter equal to the first parameter of registerProcessor
.
// Abstract data model is {}.
commandAPI('add', 'a', 1, 2);
// Abstract data model is {sum: 3}.
commandAPI('copy');
// Abstract data model is {sum: 3, ans: 3, finalAns: undefined}.
commandAPI('copy');
// Abstract data model is {sum: 3, ans: 3, finalAns: 3}.
To register multiple processors when constructing a helper, you can use the second parameter of the
DataLayerHelper
constructor as an options object. Pass the processors in under the commandProcessors
attribute following the data structure presented:
const helper = new DataLayerHelper(dataLayer, {
commandProcessors: {
'add': [
function(a, b) {return {sum: a + b}},
function(a, b) {return {totalSum: this.get('totalSum') + a + b}},
],
'event': function(eventName) {return {lastEvent: eventName}},
},
});
Similar to the registerProcessor
method, the add
command will invoke the two functions passed in to
the add
attribute in order.
Delaying Processing
Tools that are loaded onto the page asynchronously or lazily will appreciate that you can also opt to delay the processing the dataLayer until all command API processors are registered. Since command API processors do not reprocess past dataLayer messages, it is important to register them before processing the commands intended to invoke them.
To delay processing, mark the processNow
attribute as false in an options object for the second parameter
of the DataLayerHelper
constructor. When the helper is ready to begin processing the dataLayer, simply call
the process
method of DataLayerHelper
. Note that the helper will not respond to any messages pushed onto
the dataLayer until process
has been called.
const helper = new DataLayerHelper(dataLayer, {
processNow: false,
});
// Command processors are registered lazily as asynchronous actions occur.
// The abstract data model is empty as the helper has not yet processed the dataLayer.
// Once all async actions are completed, we are ready for our helper to process the dataLayer.
helper.process();
// All messages that have been pushed onto the dataLayer are processed and added to the data model.
// If any commands were pushed, they invoke the registered command processors.
Summary
We've seen above that the data layer provides a simple API for page authors. They simply define
an array called dataLayer
, then push messages onto it.
There are three types of messages:
- Standard Messages (Objects)
- Native Method Calls (Command Arrays)
- Custom Method Calls (Functions)
This helper library provides tools and vendors a way to consume these messages. It automatically
listens for new messages and merges them onto its abstract data model. You can query the model
using the get()
API, or you can get message notifications with a callback function.
At this point, we highly recommend that you read the code and browse the tests for examples of how the library works and how it can be used.
Build and Test
A few prerequisites:
Clone a copy of the project repo by running:
git clone --recursive git://github.com/google/data-layer-helper.git
Enter the data-layer-helper directory and install the Node dependencies, this time without specifying a global install:
cd data-layer-helper
yarn install
That should be everything. You can try running the build, which will run the linter, compile/minify the JavaScript and run the tests in chrome.
yarn start
The built version (data-layer-helper.js) will be in the dist/
subdirectory. However, this build will do it's best to fail silently
when strange inputs are passed into it. For development, you may want to use a build that will give more detailed logging. Run the following command:
yarn build-debug
to make the debug build available (dist/data-layer-helper-debug.js).
To run the tests for more than a single run (i.e. for development), you can the run the command
yarn unit
or
yarn integration
depending on if you would like to run the unit or integration tests.
License
Copyright 2012 Google Inc. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.