Any ideas on exactly how the new google instant search works? It seems to just be AJAX calls to the old search, but it's pretty hard to simplify Google that much. Anybody have speculations?
EDIT: I know there is AJAX sent with each keypress, but is it predictive? Or do you think it's just a regular ol' google search?
On
Converting the first value from key u in Daniel Vassallo's answer's Ajax response gives a URL, which then leads to some (invalid) JSON. I'm not sure what this bit is for yet.
The d gives some HTML, which gets inserted into the page. Here's a slightly decompressed version. It's mostly a script tag, which passes in 2 further chunks of HTML to a function je.pa(). These chunks seem to be the same as the non-instant search results.
On
It's predictive. The search results are from the first autocomplete suggestion (or the selected one if you hit the down arrow or mouse over a suggestion).
In other words if I type "vinyl s", the first autocomplete suggestion is "vinyl siding," and the search results clearly show a search for "vinyl siding." If I do a normal google search for "vinyl s," the search results are different.
So google instant is predictive in a way that normal google searches are not... it searches for the suggested autocompletion instead of the literal unfinished search query.
On
As others have said, Google is instantly returning the results for the top result of what Google Suggest suggests. Therefore, if you type in fac, Google Suggests thinks you are typing in facebook. Interestingly, only the partial search term, fac, is sent to Google in the JSON request, so Google Suggest is used at both ends.
Searches are not actually performed on every keystroke. If you type fast enough, some can get missed.
I suspect this doesn't increase the load on Google's servers as much as you might think. Most of the additional searches are for very popular words when you start typing stuff in — if you starting searching for something beginning with f, it will search for facebook, if you start searching for something beginning with ju, it will search for justin bieber. Thanks to caching, these searches are not very expensive to do.
On
I while ago I wrote a vaguely popular instant search GUI for the Solr search engine. Its really simple, so you should be able to look at the code and understand the general concepts http://fergiemcdowall.github.io/solrstrap/
UPDATE: Google have just published a blog article called Google Instant, behind the scenes. It's an interesting read, and obviously related to this question. You can read how they tackled the extra load (5-7X according to the article) on the server-side, for example. The answer below examines what happens on the client-side:
Examining with Firebug, Google is doing an Ajax GET request on every keypress:
I guess it's working the same way as the auto completion. However this time, it also returns the search results of the partially complete search phrase in JSON format.
Examining one of the JSON responses while typing "Stack Overflow":
We can see that the JSON response contains the content to construct the search results as we type.
The formatted JSON responses look something like this:
Further non-technical reading: