Your question is very difficult to answer, certainly to give a single conclusive answer. In fact, I am tempted to flag this question as "too broad", which is supported by the other answers and comments.
Take, for example, just your element(by.id("id"));
. Looking through the Selenium source, most drivers just take whatever id you give, and pass it off to the wire protocol:
public WebElement findElementById(String using) {
if (getW3CStandardComplianceLevel() == 0) {
return findElement("id", using);
} else {
return findElementByCssSelector("#" + cssEscape(using));
}
}
As you know, each browser vendor implements their own wire protocol in a separate binary. Feel free to go further into the code, to dig a deeper hole for your self.
For other browsers that do not support the wire protocol, for example HtmlUnit, you just have something like:
public List<WebElement> findElementsById(String id) {
return findElementsByXPath("//*[@id='" + id + "']");
}
and then they parse the available DOM.
As for your performance question, anything that anyone gives you will be 1) just a feeling, or 2) pure BS! Which you can already see from the other answers and comments you are getting.
To get a real answer (supported by actual data), there are just too many variables to consider:
- Wire protocol as implemented by different browser vendors, plus various optimizations in different versions.
- DOM engines as implemented by different browser vendors, plus various optimizations in different versions.
- JavaScript engines as implemented by different browser vendors, plus various optimizations in different versions.
Also, whatever results you get for your web app / web page will most like not apply to a different web app / web page, due to differences in the framework used to build that site.
Bottom line is: If you are concerned about performance testing, then Selenium is the wrong answer. Selenium is a functional test library, optimized to give you the best end-user representation. Performance is a distant afterthought.
If your goal is to get your tests to run faster, your time will be better spent looking at your test structure:
- How frequently do you open/close the browser. This is often the most time consuming activity in a test.
- How often do you refresh your element cache, how often do you need to? Consider moving your elements to Page Factory model, which lazy-loads all elements for you.
- And of course the biggest speedup factor: running your tests in parallel on multiple machines.
But I think this is getting off topic (some might suggest "ranty") from your initial question.