Building Fast Client-side Searches

Yesterday we released a new people selector widget (which we’ve been calling Bo Selecta internally). This widget downloads a list of all of your contacts, in JavaScript, in under 200ms (this is true even for members with 10,000+ contacts). In order to get this level of performance, we had to completely rethink how we send data from the server to the client.

Server Side: Cache Everything

To make this data available quickly from the server, we maintain and update a per-member cache in our database, where we store each member’s contact list in a text blob — this way it’s a single quick DB query to retrieve it. We can format this blob in any way we want: XML, JSON, etc. Whenever a member updates their information, we update the cache for all of their contacts. Since a single member who changes their contact information can require updating the contacts cache for hundreds or even thousands of other members, we rely upon prioritized tasks in our offline queue system.

Testing the Performance of Different Data Formats

Despite the fact that our backend system can deliver the contact list data very quickly, we still don’t want to unnecessarily fetch it for each page load. This means that we need to defer loading until it’s needed, and that we have to be able to request, download, and parse the contact list in the amount of time it takes a member to go from hovering over a text field to typing a name.

With this goal in mind, we started testing various data formats, and recording the average amount of time it took to download and parse each one. We started with Ajax and XML; this proved to be the slowest by far, so much so that the larger test cases wouldn’t even run to completion (the tags used to create the XML structure also added a lot of weight to the filesize). It appeared that using XML was out of the question.

BoSelectaJsonGoodFunTimes: eval() is Slow

DJ Bo Selecta on the decks

Next we tried using Ajax to fetch the list in the JSON format (and having eval() parse it). This was a major improvement, both in terms of filesize across the wire and parse time.

While all of our tests ran to completion (even the 10,000 contacts case), parse time per contact was not the same for each case; it geometrically increased as we increased the number of contacts, up to the point where the 10,000 contact case took over 80 seconds to parse — 400 times slower than our goal of 200ms. It seemed that JavaScript had a problem manipulating and eval()ing very large strings, so this approach wasn’t going to work either.

Contacts File Size (KB) Parse Time (ms) File Size per Contact (KB) Parse Time per Contact (ms)
10,617 1536 81312 0.14 7.66
4,878 681 18842 0.14 3.86
2,979 393 6987 0.13 2.35
1,914 263 3381 0.14 1.77
1,363 177 1837 0.13 1.35
798 109 852 0.14 1.07
644 86 611 0.13 0.95
325 44 252 0.14 0.78
260 36 205 0.14 0.79
165 24 111 0.15 0.67

JSON and Dynamic Script Tags: Fast but Insecure

Working with the theory that large string manipulation was the problem with the last approach, we switched from using Ajax to instead fetching the data using a dynamically generated script tag. This means that the contact data was never treated as string, and was instead executed as soon as it was downloaded, just like any other JavaScript file. The difference in performance was shocking: 89ms to parse 10,000 contacts (a reduction of 3 orders of magnitude), while the smallest case of 172 contacts only took 6ms. The parse time per contact actually decreased the larger the list became. This approach looked perfect, except for one thing: in order for this JSON to be executed, we had to wrap it in a callback method. Since it’s executable code, any website in the world could use the same approach to download a Flickr member’s contact list. This was a deal breaker.

Contacts File Size (KB) Parse Time (ms) File Size per Contact (KB) Parse Time per Contact (ms)
10,709 1105 89 0.10 0.01
4,877 508 41 0.10 0.01
2,979 308 26 0.10 0.01
1,915 197 19 0.10 0.01
1,363 140 15 0.10 0.01
800 83 11 0.10 0.01
644 67 9 0.10 0.01
325 35 8 0.11 0.02
260 27 7 0.10 0.03
172 18 6 0.10 0.03

Going Custom

Custom Ride

Having set the performance bar pretty high with the last approach, we dove into custom data formats. The challenge would be to create a format that we could parse ourselves, using JavaScript’s String and RegExp methods, that would also match the speed of JSON executed natively. This would allow us to use Ajax again, but keep the data restricted to our domain.

Since we had already discovered that some methods of string manipulation didn’t perform well on large strings, we restricted ourselves to a method that we knew to be fast: split(). We used control characters to delimit each contact, and a different control character to delimit the fields within each contact. This allowed us to parse the string into contact objects with one split, then loop through that array and split again on each string.

that.contacts = o.responseText.split("\c");

for (var n = 0, len = that.contacts.length, contactSplit; n < len; n++) {

	contactSplit = that.contacts[n].split("\a");

	that.contacts[n] = {};
	that.contacts[n].n = contactSplit[0];
	that.contacts[n].e = contactSplit[1];
	that.contacts[n].u = contactSplit[2];
	that.contacts[n].r = contactSplit[3];
	that.contacts[n].s = contactSplit[4];
	that.contacts[n].f = contactSplit[5];
	that.contacts[n].a = contactSplit[6];
	that.contacts[n].d = contactSplit[7];
	that.contacts[n].y = contactSplit[8];

Though this technique sounds like it would be slow, it actually performed on par with native JSON parsing (it was a little faster for cases containing less than 1000 contacts, and a little slower for those over 1000). It also had the smallest filesize: 80% the size of the JSON data for the same number of contacts. This is the format that we ended up using.

Contacts File Size (KB) Parse Time (ms) File Size per Contact (KB) Parse Time per Contact (ms)
10,741 818 173 0.08 0.02
4,877 375 50 0.08 0.01
2,979 208 34 0.07 0.01
1,916 144 21 0.08 0.01
1,363 93 16 0.07 0.01
800 58 10 0.07 0.01
644 46 8 0.07 0.01
325 24 4 0.07 0.01
260 14 3 0.05 0.01
160 13 3 0.08 0.02


Ben to the Rescue

Now that we have a giant array of contacts in JavaScript, we needed a way to search through them and select one. For this, we used YUI’s excellent AutoComplete widget. To get the data into the widget, we created a DataSource object that would execute a function to get results. This function simply looped through our contact array and matched the given query against four different properties of each contact, using a regular expression (RegExp objects turned out to be extremely well-suited for this, with the average search time for the 10,000 contacts case coming in under 38ms). After the results were collected, the AutoComplete widget took care of everything else, including caching the results.

There was one optimization we made to our AutoComplete configuration that was particularly effective. Regardless of how much we optimized our search method, we could never get results to return in less than 200ms (even for trivially small numbers of contacts). After a lot of profiling and hair pulling, we found the queryDelay setting. This is set to 200ms by default, and artificially delays every search in order to reduce UI flicker for quick typists. After setting that to 0, we found our search times improved dramatically.

The End Result

Head over to your Contact List page and give it a whirl. We are also using the Bo Selecta with FlickrMail and the Share This widget on each photo page.

Lessons Learned while Building an iPhone Site

The Explore Page in the iPhone site

A few weeks ago we released a version of the Flickr site tailored specifically for the iPhone. Developing this site was very different from any other project I’ve worked on; there seems to be a new set of frontend rules for developing high-end mobile sites. A lot of the current best practices get thrown out the window in the quest for minimum page weight and fastest load times over slow cellular connections.

Here are a few of the lessons we learned (sometimes painfully) while developing this site.

1. Don’t Use a JavaScript Library or CSS Framework

This was one of the hardest things for me to come to terms with. I’m a huge fan of libraries, especially YUI, mostly because they allow me to spend my time creating new stuff instead of working around crazy browser quirks. But these libraries walk a fine line; by definition, they must work across a wide array of browsers and offer enough features to make them worth using. This means they potentially contain a lot of code that you don’t care about and won’t use. This code is dead weight to your site.

With such a high percentage of normal web users on broadband connections, we’ve gotten cavalier about what we can include in our pages. 250 KB of JavaScript or more isn’t uncommon for a large site these days. But for sites that are meant to be viewed over slow cellular connections like EDGE, 250 KB is an impossible amount of data. The only way to get the size of your JavaScript down is to selectively pull code out of libraries, and include only what you use. This means you can rip out code meant only for browsers that you won’t support (modular libraries like the new YUI 3.0 allow you to only include the code you use, preventing this problem somewhat).

The same goes for CSS. Frameworks make development faster and your final product more robust, but they, like the JavaScript libraries, include code for situations you won’t have to deal with. Every line in your CSS must be custom; each property must be scrutinized to ensure it’s needed.

2. Load Page Fragments Instead of Full Pages

Loading fragments saves 92.2% of the page size

When navigating through a site, most of what changes from page to page is the actual content; the JavaScript, CSS, header and footer stay mostly the same. We can use this to our advantage by only loading the part of each page that changes. We did this by hijacking all links of the page: when a link is clicked, we intercept the event, fetch the page fragment using Ajax, and insert the HTML into a new div. This has several benefits:

  • Since you control the entire life cycle of the page fetch, you can display loading indicators or a wireframe version of the page while new pages load
  • All pages that have been fetched will exist within the DOM; clicking the back button (or clicking on a link for a page that has already been fetched) results in an instantaneous page load
  • The page fragments are extremely small; ours are about 800 bytes (gzipped) on average

Using this system complicates your code a bit. You need JavaScript to handle the hijacking, the page fragment insertion, and the address bar hash changes (which allow the back and forward buttons to work normally). You also need your backend to recognize requests made with Ajax, and to only send the page content instead of the full HTML document. And lastly, if you want normal URLs to work, and each of your pages to be bookmarkable, you will need even more JavaScript.

Despite these downsides, the benefits can’t be ignored. The extra JavaScript code is a one-time cost, but the extra page content that we would have downloaded is saved for every page load. We found it was worth the complication and additional JS in order to dramatically reduce the time it took to load each page.

3. Don’t Build for Just One Device

It’s really tempting to build the site for just the iPhone: you can use modern CSS (including things like CSS3 selectors and transformations), you don’t have to hack around annoying browser quirks, and testing is extremely easy. But any single device, even one as ubiquitous as the iPhone, has a limited share of the mobile market, especially internationally. Rarely can you justify the cost of creating a one-off site for a very small number of your users.

Luckily the current generation of high-end mobile browsers is excellent in terms of support for modern features. Many phones use a WebKit derivative, including the iPhone, and Symbian and Android phones. Other phones either come with or can use Opera Mobile or the new mobile version of Firefox (called Fennec). For the most part, very few changes are needed in order to support these browsers.

Most of the differences lie in layout. It’s important to structure your pages around a grid that can expand as a percentage of the page width. This allows your layouts to work on many different screen sizes and orientations. The iPhone, for example, allows both landscape and portrait viewing styles, which have vastly different layout requirements. By using percentages, you can have the content fill the screen regardless of orientation. Another option is to detect viewport width and height, and use JavaScript to dynamically adjust classes based on those measurements (but we found this was overkill; CSS can handle most situations on its own).

4. Optimize Everything

The browsers on mobile devices operate under much stricter constraints than their desktop cousins. Slower CPUs, smaller amounts of memory, and smaller hard drives mean that less data can be cached. On the iPhone, for instance, only files smaller than 25 KB are cached. This puts very specific limits of the size of your files. For a large site like Flickr, 25 KB worth of JavaScript and CSS barely scratches the surface. To put our files under the limit, we ran everything through the YUI Compressor using the most aggressive settings. We ran all images through compression tools as well (we like pngout and Smushit), reducing each image file by an average of 40%. We also made heavy use of sprites, where possible.

In the end, we were able to go from 90+ second load times over EDGE to less than 7 for an empty cache experience. Using page fragments, we are able to load and display new pages in under a second (though the images in those pages take longer to load). These are not trivial gains, and make the difference between a good mobile experience and a one that is so awful the user gives up halfway through the page load.

5. Tell the User What is Happening

Once we hijacked all clicks actions in order to load page fragments, it wasn’t always clear to the user if anything was happening when they clicked on a link. This is especially true on touch devices, where it is difficult to know if the device even detected your action. To combat this problem, we added loading indicators to every link. These tell the user that something is happening, and reassures them that their action was detected. It also makes the pages seem to load much faster, since something is happening right away; if the indicators weren’t there, it would seem like nothing was happening for a few seconds, and then the page would load suddenly. In our testing, these indicators were the difference between a UI that seems snappy and responsive, and one that seemed slow and inconsistent.

Loading indicators

One Easy Option

The iUI framework implements a lot of these practices for you, and might be a good place to start in developing any mobile site (though keep in mind it was developed specifically for the iPhone). We found it especially useful in the early stages of development, though eventually we pulled it out and wrote custom code to run the site.