Related
I need to accomplish something very similar to the "inspect element" functionality with Chrome developer tools (see attached screenshot).
The main thing I'm looking for is the ability to recognize and highlight an element on the given page. I can display an overlay, add a class to the element, etc. I know that using an iFrame and trying to access and manipulate the DOM w/in the iFrame won't work directly because its cross-domain. I also know about post messages if you have control over the site, but I may not have that all the time. The main issue I have is that I am only trying to temporarily highlight elements on a page as if I had clicked "inspect element" and displaying that. It doesn't need to persist or anything, just to highlight specific elements (h1 tags for example), when the user loads a website within my app. Even the ability to use the chrome developer tools where it lets you "edit as HTML" would work if there were a way to do so.
Is there any way, using an iFrame, a new tab, or any other means that I can use to highlight elements on a given website who's URL is provided? I've attached another image of the "goal".
It seems like you have three options to highlight the elements.
1. Browser Extension
2. Send events through postMessage to direct the iframe site to manipulate its DOM
3. Send events through postMessage to receive information about the positioning of the DOM and overlay elements on top of the iframe.
#1 seems to be a longer overhaul to do. #2 and 3 is only possible if you will always have access to the website in the iframe.
https://developer.mozilla.org/en-US/docs/Web/API/Window/postMessage
The window.postMessage() method safely enables cross-origin communication between Window objects; e.g., between a page and a pop-up that it spawned, or between a page and an iframe embedded within it.
I am running into an issue sending data from my background script to the script for my pageAction. My content script adds an <iframe /> and the JavaScript in the <iframe /> is receiving the data from my background script, but it does not seem to be retrieved in my pageAction.
In my background script I have something like:
chrome.tabs.sendMessage(senderTab.tab.id,
{
foo:bar
});
where senderTab.tab.id is the "sender" in onMessage Listener in my background script.
In the JavaScript loaded by the <iframe /> injected by my content script I have something like:
chrome.runtime.onMessage.addListener(
function(request, sender, sendResponse) {
console.log("received in iframe:", request);
}
});
The <iframe /> receives the message exactly as expected.
I put the same JavaScript in my page_action.js, but it does not receive any data from the background script. The pageAction is activated with chrome.pageAction.show(senderTab.tab.id); before I call chrome.tabs.sendMessage(senderTab.tab.id ...
Is the HTML page attached to my pageAction not part of the same tab? Since this tabId enabled me to activate/"show" the icon, I would think the listener in the JavaScript for the pageAction should also receive from chrome.tabs.sendMessage(senderTab.tab.id ...
In my content script I use the following to send data to the background script:
chrome.runtime.sendMessage({
foo: bar
});
When the content script sends the above message, the pageAction JavaScript is picking it up.
How do I get the background script to properly send data to my pageAction? I do not want to have pageAction request/poll, instead I want pageAction to just listen and receive. E.g., if the pageAction HTML it shown, it should be able to update in real time as the background page makes changes.
Communicating with a page in the background context
Pages which are open in the background context include:
background pages/scripts(MDN)
event pages (Firefox does not support event pages. All manifest.json background pages remain loaded at all times.)
browser action popups(MDN)
page action popups(MDN)
options pages(MDN1, MDN2) (in a popup, a tab, or window)
sidebar action pages (NDN) (not available in Chrome)
Any HTML content contained within your extension which is opened normally in a tab, window (e.g. a panel), or frame.1
Using tabs.sendMessage()(MDN) will not send a message to any of them. You would need to use runtime.sendMessage()(MDN) to send a message to them. The scope for any of them, except background pages and event pages, only exists when it is being displayed. Obviously, you can not communicate with the code when it does not exist. When the scope exists, you can communicate with any of them using:
Directly
From the background context, you can directly change variables, or call functions, in another page that is also in the background context (i.e. not content scripts), after having gotten a reference to its global scope, its Window, using extension.getViews()(MDN), extension.getBackgroundPage()(MDN), or other method(MDN).
For example, you can call a function created with function myFunction in the page of the first returned view by using something like:
winViews = chrome.extension.getViews();
winViews[0].myFunction(foo);
It should be noted that in your callback from tabs.create()(MDN) or windows.create()(MDN) the view for the newly opened tab or window will probably not yet exist. You will need to use some methodology to wait for the view to exist.2 See below for recommended ways to communicate with newly opened tabs or windows.
Directly manipulating values in the other page's scope allows you to communicate any type of data you desire.
Messaging
Receive messages using chrome.runtime.onMessage(MDN), 3 which were sent with chrome.runtime.sendMessage()(MDN). Each time you receive a message in a runtime.onMessage listener, there will be a sendResponse function provided as the third argument which allows you to directly respond to the message. If the original sender has not supplied a callback to receive such a response in their call to chrome.runtime.sendMessage(), then the response is lost. If using Promises (e.g. browser.runtime.sendMessage() in Firefox), the response is passed as an argument when the Promise is fulfilled. If you want to send the response asynchronously, you will need to return true; from your runtime.onMessage listener.
Ports
You can also connect ports, using chrome.runtime.connect()(MDN) and chrome.runtime.onConnect(MDN) for longer term messaging.
Use chrome.tabs.sendMessage() to send to content scripts
If you want to send from the background context (e.g. background script or popup) to a content script you would use chrome.tabs.sendMessage()/chrome.runtime.onMessage, or connect port(s) using chrome.tabs.connect()(MDN)/chrome.runtime.onConnect.
JSON-serializable data only
Using messaging, you can only pass data which is JSON-serializable.
Messages are received by all scripts in the background, except the sender
Messages sent to the background context are received by all scripts in the background context which have registered a listener, except the script which sent it.3 There is no way to specify that it is only to be received by a specific script. Thus, if you have multiple potential recipients, you will need to create a way to be sure that the message received was intended for that script. The ways to do so usually rely on specific properties existing in the message (e.g. use a destination or recipient property to indicate what script is to receive it, or define that some type of messages are always for one recipient or another), or to differentiate based on the sender(MDN) supplied to the message handler (e.g. if messages from one sender are always only for a specific recipient). There is no set way to do this, you must choose/create a way to do it for use in your extension.
For a more detailed discussion of this issue, please see: Messages intended for one script in the background context are received by all
Data in a StorageArea
Store data to a StorageArea(MDN) and be notified of the change in other scripts using chrome.storage.onChanged(MDN). The storage.onChanged event can be listened to in both the background context and content scripts.
You can only store data which is JSON-serializable into a StorageArea.
Which method is best to use in any particular situation will depends on what you are wanting to communicate (type of data, state change, etc.), and to which portion, or portions, of your extension you are wanting to communicate from and to. For instance, if you want to communicate information which is not JSON-serializable, you would need to do so directly (i.e. not messaging or using a StorageArea). You can use multiple methods in the same extension.
More on popups
None of the popups (e.g. browser action, or page action) are directly associated with the active tab. There is no concept of a shared or separate instance per tab. However, the user can open one popup in each Chrome window. If more than one popup is open (a maximum of one per Chrome window), then each is in a separate instance (separate scope; has its own Window), but are in the same context. When a popup is actually visible, it exists in the background context.
There is only ever one page action or browser action popup open at a time per Chrome window. The HTML file which will be open will be whichever one has been defined for the active tab of the current window and opened by the user by clicking on the page/browser action button. This can be assigned a different HTML document for different tabs by using chrome.browserAction.setPopup()(MDN), or chrome.pageAction.setPopup()(MDN), and specifying a tabId. The popup can/will be destroyed for multiple reasons, but definitely when another tab becomes the active tab in the window in which the popup is open.
However, any method of communication used will only communicate to the one(s) which is/are currently open, not ones which are not open. If popups are open for more than one Chrome window at a time, then they are separate instances, with their own scope (i.e. their own Window). You can think of this something like having the same web page open in more than one tab.
If you have a background script, the background script context is persistent across the entire instance of Chrome. If you do not have a background script the context may be created when needed (e.g. a popup is shown) and destroyed when no longer needed.
chrome.tabs.sendMessage() can not communicate to popups
As mentioned above, even if the popup did exist, it will exist in the background context. Calling chrome.tabs.sendMessage() sends a message to content scripts injected into a tab/frame, not to the background context. Thus, it will not send a message to a non-content script like a popup.
Action button: enable/disable (browser action) vs. show/hide (page action)
Calling chrome.pageAction.show()(MDN) just causes the page action button to be shown. It does not cause any associated popup to be shown. If the popup/options page/other page is not actually being shown (not just the button), then its scope does not exist. When it does not exist, it, obviously, can not receive any message
Instead of the page action's ability to show()(MDN) or hide()(MDN) the button, browser actions can enable()(MDN) or disable()(MDN) the button.
Programmatically opening a tab or window with HTML from your extension
You can use tabs.create()(MDN) or windows.create()(MDN) to open a tab or window containing an HTML page from within your extension. However, the callback for both of those API calls is executed prior to the page's DOM existing and thus prior to any JavaScript associated with the page existing. Thus, you can not immediately access the DOM created by the contents of that page, nor interact with the JavaScript for the page. Very specifically: no runtime.onMessage() listeners will have been added, so no messages sent at that time will be received by the newly opening page.
The best ways to resolve this issue are:
Have the data available so the newly opening page can get the data when it is ready for. Do this by, prior to beginning the process of opening the page:
If the source is in the background context: store the data in a variable available to the global scope of the sending page. The opening page can then use chrome.extension.getBackgroundPage() to read the data directly.
If the source of the data is in either the background context or a content script: place the data into storage.local(MDN). The opening page can then read it when its JavaScript is run. For example, you could use a key called messageToNewExtensionPage.
If you are using runtime.sendMessage(), then initiate the transfer of the data from your newly opening page by sending a message from the that page's code to the source of the data (using runtime.sendMessage(), or tabs.sendMessage() for content script sources) requesting the data. The script with the data can then send the data back using the sendResponse(MDN) function provided by runtime.onMessage().
Wait to interact with the newly opening page until after at least the DOM is available, if not until after the JavaScript for the page has run. While it's possible to do this without the newly opening page providing specific notification that it's up and running, doing so is more complex and only useful in some specific cases (e.g. you want to do something prior to the JavaScript in the new page being run).2
Additional references
Chrome
Message Passing
Chrome extension overview
architecture
Communication between pages
Firefox
WebExtensions
Anatomy of a WebExtension
With some minor exceptions: e.g. using a content script to insert content into the page context.
There are multiple methods which you can use. Which way is best will depend on exactly what you are doing (e.g. when you need to access the view with respect to the code being executed in the view). A simple method would be just to poll waiting for the view to exist. The following code does that for opening a window:
chrome.windows.create({url: myUrl},function(win){
//Poll for the view of the window ID. Poll every 50ms for a
// maximum of 20 times (1 second). Then do a second set of polling to
// accommodate slower machines. Testing on a single moderately fast machine
// indicated the view was available after, at most, the second 50ms delay.
waitForWindowId(win.id,50,20,actOnViewFound,do2ndWaitForWinId);
});
function waitForWindowId(id,delay,maxTries,foundCallback,notFoundCallback) {
if(maxTries--<=0){
if(typeof notFoundCallback === 'function'){
notFoundCallback(id,foundCallback);
}
return;
}
let views = chrome.extension.getViews({windowId:id});
if(views.length > 0){
if(typeof foundCallback === 'function'){
foundCallback(views[0]);
}
} else {
setTimeout(waitForWindowId,delay,id,delay,maxTries,foundCallback
,notFoundCallback);
}
}
function do2ndWaitForWinId(winId,foundCallback){
//Poll for the view of the window ID. Poll every 500ms for max 40 times (20s).
waitForWindowId(winId,500,40,foundCallback,windowViewNotFound);
}
function windowViewNotFound(winId,foundCallback){
//Did not find the view for the window. Do what you want here.
// Currently fail quietly.
}
function actOnViewFound(view){
//What you desire to happen with the view, when it exists.
}
From MDN:
In Firefox versions prior to version 51, the runtime.onMessage listener will be called for messages sent from the same script (e.g. messages sent by the background script will also be received by the background script). In those versions of Firefox, if you unconditionally call runtime.sendMessage() from within a runtime.onMessage listener, you will set up an infinite loop which will max-out the CPU and lock-up Firefox. If you need to call runtime.sendMessage() from within a runtime.onMessage, you will need to check the sender.url property to verify you are not sending a message in response to a message which was sent from the same script. This bug was resolved as of Firefox 51.
I am running into an issue sending data from my background script to the script for my pageAction. My content script adds an <iframe /> and the JavaScript in the <iframe /> is receiving the data from my background script, but it does not seem to be retrieved in my pageAction.
In my background script I have something like:
chrome.tabs.sendMessage(senderTab.tab.id,
{
foo:bar
});
where senderTab.tab.id is the "sender" in onMessage Listener in my background script.
In the JavaScript loaded by the <iframe /> injected by my content script I have something like:
chrome.runtime.onMessage.addListener(
function(request, sender, sendResponse) {
console.log("received in iframe:", request);
}
});
The <iframe /> receives the message exactly as expected.
I put the same JavaScript in my page_action.js, but it does not receive any data from the background script. The pageAction is activated with chrome.pageAction.show(senderTab.tab.id); before I call chrome.tabs.sendMessage(senderTab.tab.id ...
Is the HTML page attached to my pageAction not part of the same tab? Since this tabId enabled me to activate/"show" the icon, I would think the listener in the JavaScript for the pageAction should also receive from chrome.tabs.sendMessage(senderTab.tab.id ...
In my content script I use the following to send data to the background script:
chrome.runtime.sendMessage({
foo: bar
});
When the content script sends the above message, the pageAction JavaScript is picking it up.
How do I get the background script to properly send data to my pageAction? I do not want to have pageAction request/poll, instead I want pageAction to just listen and receive. E.g., if the pageAction HTML it shown, it should be able to update in real time as the background page makes changes.
Communicating with a page in the background context
Pages which are open in the background context include:
background pages/scripts(MDN)
event pages (Firefox does not support event pages. All manifest.json background pages remain loaded at all times.)
browser action popups(MDN)
page action popups(MDN)
options pages(MDN1, MDN2) (in a popup, a tab, or window)
sidebar action pages (NDN) (not available in Chrome)
Any HTML content contained within your extension which is opened normally in a tab, window (e.g. a panel), or frame.1
Using tabs.sendMessage()(MDN) will not send a message to any of them. You would need to use runtime.sendMessage()(MDN) to send a message to them. The scope for any of them, except background pages and event pages, only exists when it is being displayed. Obviously, you can not communicate with the code when it does not exist. When the scope exists, you can communicate with any of them using:
Directly
From the background context, you can directly change variables, or call functions, in another page that is also in the background context (i.e. not content scripts), after having gotten a reference to its global scope, its Window, using extension.getViews()(MDN), extension.getBackgroundPage()(MDN), or other method(MDN).
For example, you can call a function created with function myFunction in the page of the first returned view by using something like:
winViews = chrome.extension.getViews();
winViews[0].myFunction(foo);
It should be noted that in your callback from tabs.create()(MDN) or windows.create()(MDN) the view for the newly opened tab or window will probably not yet exist. You will need to use some methodology to wait for the view to exist.2 See below for recommended ways to communicate with newly opened tabs or windows.
Directly manipulating values in the other page's scope allows you to communicate any type of data you desire.
Messaging
Receive messages using chrome.runtime.onMessage(MDN), 3 which were sent with chrome.runtime.sendMessage()(MDN). Each time you receive a message in a runtime.onMessage listener, there will be a sendResponse function provided as the third argument which allows you to directly respond to the message. If the original sender has not supplied a callback to receive such a response in their call to chrome.runtime.sendMessage(), then the response is lost. If using Promises (e.g. browser.runtime.sendMessage() in Firefox), the response is passed as an argument when the Promise is fulfilled. If you want to send the response asynchronously, you will need to return true; from your runtime.onMessage listener.
Ports
You can also connect ports, using chrome.runtime.connect()(MDN) and chrome.runtime.onConnect(MDN) for longer term messaging.
Use chrome.tabs.sendMessage() to send to content scripts
If you want to send from the background context (e.g. background script or popup) to a content script you would use chrome.tabs.sendMessage()/chrome.runtime.onMessage, or connect port(s) using chrome.tabs.connect()(MDN)/chrome.runtime.onConnect.
JSON-serializable data only
Using messaging, you can only pass data which is JSON-serializable.
Messages are received by all scripts in the background, except the sender
Messages sent to the background context are received by all scripts in the background context which have registered a listener, except the script which sent it.3 There is no way to specify that it is only to be received by a specific script. Thus, if you have multiple potential recipients, you will need to create a way to be sure that the message received was intended for that script. The ways to do so usually rely on specific properties existing in the message (e.g. use a destination or recipient property to indicate what script is to receive it, or define that some type of messages are always for one recipient or another), or to differentiate based on the sender(MDN) supplied to the message handler (e.g. if messages from one sender are always only for a specific recipient). There is no set way to do this, you must choose/create a way to do it for use in your extension.
For a more detailed discussion of this issue, please see: Messages intended for one script in the background context are received by all
Data in a StorageArea
Store data to a StorageArea(MDN) and be notified of the change in other scripts using chrome.storage.onChanged(MDN). The storage.onChanged event can be listened to in both the background context and content scripts.
You can only store data which is JSON-serializable into a StorageArea.
Which method is best to use in any particular situation will depends on what you are wanting to communicate (type of data, state change, etc.), and to which portion, or portions, of your extension you are wanting to communicate from and to. For instance, if you want to communicate information which is not JSON-serializable, you would need to do so directly (i.e. not messaging or using a StorageArea). You can use multiple methods in the same extension.
More on popups
None of the popups (e.g. browser action, or page action) are directly associated with the active tab. There is no concept of a shared or separate instance per tab. However, the user can open one popup in each Chrome window. If more than one popup is open (a maximum of one per Chrome window), then each is in a separate instance (separate scope; has its own Window), but are in the same context. When a popup is actually visible, it exists in the background context.
There is only ever one page action or browser action popup open at a time per Chrome window. The HTML file which will be open will be whichever one has been defined for the active tab of the current window and opened by the user by clicking on the page/browser action button. This can be assigned a different HTML document for different tabs by using chrome.browserAction.setPopup()(MDN), or chrome.pageAction.setPopup()(MDN), and specifying a tabId. The popup can/will be destroyed for multiple reasons, but definitely when another tab becomes the active tab in the window in which the popup is open.
However, any method of communication used will only communicate to the one(s) which is/are currently open, not ones which are not open. If popups are open for more than one Chrome window at a time, then they are separate instances, with their own scope (i.e. their own Window). You can think of this something like having the same web page open in more than one tab.
If you have a background script, the background script context is persistent across the entire instance of Chrome. If you do not have a background script the context may be created when needed (e.g. a popup is shown) and destroyed when no longer needed.
chrome.tabs.sendMessage() can not communicate to popups
As mentioned above, even if the popup did exist, it will exist in the background context. Calling chrome.tabs.sendMessage() sends a message to content scripts injected into a tab/frame, not to the background context. Thus, it will not send a message to a non-content script like a popup.
Action button: enable/disable (browser action) vs. show/hide (page action)
Calling chrome.pageAction.show()(MDN) just causes the page action button to be shown. It does not cause any associated popup to be shown. If the popup/options page/other page is not actually being shown (not just the button), then its scope does not exist. When it does not exist, it, obviously, can not receive any message
Instead of the page action's ability to show()(MDN) or hide()(MDN) the button, browser actions can enable()(MDN) or disable()(MDN) the button.
Programmatically opening a tab or window with HTML from your extension
You can use tabs.create()(MDN) or windows.create()(MDN) to open a tab or window containing an HTML page from within your extension. However, the callback for both of those API calls is executed prior to the page's DOM existing and thus prior to any JavaScript associated with the page existing. Thus, you can not immediately access the DOM created by the contents of that page, nor interact with the JavaScript for the page. Very specifically: no runtime.onMessage() listeners will have been added, so no messages sent at that time will be received by the newly opening page.
The best ways to resolve this issue are:
Have the data available so the newly opening page can get the data when it is ready for. Do this by, prior to beginning the process of opening the page:
If the source is in the background context: store the data in a variable available to the global scope of the sending page. The opening page can then use chrome.extension.getBackgroundPage() to read the data directly.
If the source of the data is in either the background context or a content script: place the data into storage.local(MDN). The opening page can then read it when its JavaScript is run. For example, you could use a key called messageToNewExtensionPage.
If you are using runtime.sendMessage(), then initiate the transfer of the data from your newly opening page by sending a message from the that page's code to the source of the data (using runtime.sendMessage(), or tabs.sendMessage() for content script sources) requesting the data. The script with the data can then send the data back using the sendResponse(MDN) function provided by runtime.onMessage().
Wait to interact with the newly opening page until after at least the DOM is available, if not until after the JavaScript for the page has run. While it's possible to do this without the newly opening page providing specific notification that it's up and running, doing so is more complex and only useful in some specific cases (e.g. you want to do something prior to the JavaScript in the new page being run).2
Additional references
Chrome
Message Passing
Chrome extension overview
architecture
Communication between pages
Firefox
WebExtensions
Anatomy of a WebExtension
With some minor exceptions: e.g. using a content script to insert content into the page context.
There are multiple methods which you can use. Which way is best will depend on exactly what you are doing (e.g. when you need to access the view with respect to the code being executed in the view). A simple method would be just to poll waiting for the view to exist. The following code does that for opening a window:
chrome.windows.create({url: myUrl},function(win){
//Poll for the view of the window ID. Poll every 50ms for a
// maximum of 20 times (1 second). Then do a second set of polling to
// accommodate slower machines. Testing on a single moderately fast machine
// indicated the view was available after, at most, the second 50ms delay.
waitForWindowId(win.id,50,20,actOnViewFound,do2ndWaitForWinId);
});
function waitForWindowId(id,delay,maxTries,foundCallback,notFoundCallback) {
if(maxTries--<=0){
if(typeof notFoundCallback === 'function'){
notFoundCallback(id,foundCallback);
}
return;
}
let views = chrome.extension.getViews({windowId:id});
if(views.length > 0){
if(typeof foundCallback === 'function'){
foundCallback(views[0]);
}
} else {
setTimeout(waitForWindowId,delay,id,delay,maxTries,foundCallback
,notFoundCallback);
}
}
function do2ndWaitForWinId(winId,foundCallback){
//Poll for the view of the window ID. Poll every 500ms for max 40 times (20s).
waitForWindowId(winId,500,40,foundCallback,windowViewNotFound);
}
function windowViewNotFound(winId,foundCallback){
//Did not find the view for the window. Do what you want here.
// Currently fail quietly.
}
function actOnViewFound(view){
//What you desire to happen with the view, when it exists.
}
From MDN:
In Firefox versions prior to version 51, the runtime.onMessage listener will be called for messages sent from the same script (e.g. messages sent by the background script will also be received by the background script). In those versions of Firefox, if you unconditionally call runtime.sendMessage() from within a runtime.onMessage listener, you will set up an infinite loop which will max-out the CPU and lock-up Firefox. If you need to call runtime.sendMessage() from within a runtime.onMessage, you will need to check the sender.url property to verify you are not sending a message in response to a message which was sent from the same script. This bug was resolved as of Firefox 51.
What is the difference between JQuery mobile pagecontainer's "load" versus "change" functions and what is intended usage of each? I'm not interested in how the events are for each but rather what does each to do the DOM.
I've been using change to switch pages; but I'm getting some "unexpected" behavior.
Here's my scenario. There is a light page that is presented to the user for login. After logging in, I call "change" to get a page that loads a heavy page (a dashboard). In that dashboard page, I've got a button hooked up that has JQ "change" to a dialog page. When the user confirms the dialog, JQ "changes" to the dashboard page.
The issue is that upon changing back to the dashboard page the dashboard UI is all new. For example, say a user flipped a flipswitch in the dashboard, upon returning to the dashboard the user finds the switch isn't flipped.
After inspecting the DOM I see that when switching to the dialog page, JQ removes the dashboard page from the DOM and inserts the dialog. When switching back to the dashboard, JQ removes the dialog from the DOM and inserts the dashboard. Thus the fresh UI; but if the behavior is such, what is the intent of the "reload" property of each method if "change" chucks the page out of the DOM?
I tried changing things to using "load" and load inserts the DOM in but doesn't make it visible. I can't figure out how to use change to make it visible...
Ok, JavaScript/jQuery I got that, I can work with that to do what I want in general. However what I am trying to do currently is work with the DOM of the open/current tab, and I am wondering if that is possible or is all the work I do with an extension limited to the html I provide it with a "background.html" or equivalent.
For todays attempt at taking a strike at a google extension I want to take images on a page and store them in an array to create a slide show like effect from it via a bar I want to add to the bottom of the page appending it to the existing DOM of the open/current tab. I then want "hide" elements in the DOM til the bar is closed.
So my first question is, is something like this possible, can I manipulate the DOM in such a way and read from it.
Content Scripts are scripts which run in an environment between a page and the Chrome extension. These scripts are loaded on every page load, and have full access to the page's DOM. DOM methods, such as document.getElementById() behave as if they were a part of the page.
The global window objects (including frames and HTMLIFrameElement.contentWindow) are the only objects which are not directly readable by Content scripts.
Content scripts run on every page as defined in the manifest file. Specify a match pattern at the "content_scripts" section, to define on which pages the content script has to run. Also add this pattern to the "permissions" section, to unlock the ability to modify the page's DOM.
Note: Only a few of the chrome.* APIs, can be used by these scripts (such as chrome.extension.sendRequest to communicate with the background page).
In a background page, the DOM of a page is not directly accessible. You can inject scripts in an arbitrary tab using chrome.extension.executeScript, but you won't be able to get a direct reference to an element. For this purpose, Content scripts are required.