Lightning fast! Performance tips for using Google APIs
 |
| Sven |
 |
| Anton |
By Anton Lopyrev
and Sven Mawson, Google Developer TeamOver a year ago, we
launched
support for partial response and partial update for a number of APIs based on the
Google Data Protocol. That launch was a part of our continuous effort to
make the web
faster. It was well received by our developer community as it significantly reduced
network, memory, and CPU resources needed to work with certain Google APIs.
Today, we are adding support for partial response and an improved version of partial
update, called
patch, to a number of newer APIs such as
Buzz,
URL Shortener,
Tasks and
many
others. In fact, all APIs available in the
Google APIs Discovery Service and
the
APIs Explorer now support
this feature.
To learn how to use partial response and partial update
with a Google API, you can see the “Performance Tips” page in the documentation of the
Tasks and
Buzz APIs. We’ll roll
out this page for all of
the
supported APIs over the next few months, but you can already use the algorithms with
all of them today.
The
partial response algorithm is
identical to what was provided by the Google Data Protocol. By supplying a
fields query parameter to any API call that returns data, you can
request specific fields. Here is an example request that returns only titles and timestamps of
a user’s public Buzz activities:
https://www.googleapis.com/buzz/v1/activities/antonlopyrev@gmail.com/@public?alt=json&pp=1&fields=items(title,updated)
Given
that
the
full response is around 53KB and
the
partial response is only 3KB, the data sent to the client is reduced by almost
95%!
While the
partial response algorithm is
unchanged, the
partial update algorithm has changed significantly compared
to what was provided by Google Data Protocol. We’ve received feedback that the old algorithm
was too complicated and hard to use, which prompted us to design something much simpler. The
basics remain the same: you can use the HTTP
PATCH verb in supported
API methods to send partial updates to Google servers. However, the mechanics are different.
Adding and modifying data uses the same 'merge' semantics as before. But deleting is
simplified; just set a field to 'null'. Of course, the devil is in the details, so please
check out the
documentation
for the nitty gritty.
You can try out both
partial
response and
patch algorithms in the
APIs Explorer. For partial
responses, the fields parameter is available for most methods. In addition, the partial update
methods are denoted by
.patch in the method name. You can try both
the fields parameter and the patch method on the “tasklist” resource in the APIs explorer.
If you are using
Java or
Python client libraries
to access Google APIs, you can already ask for partial responses and send patch requests in
the code. We are adding partial support to the rest of the
Google APIs client
libraries over time.
As our APIs get more and more use from
devices with limited resources, taking advantage of performance optimizations such as partial
response and patch is crucial for making your applications faster and more efficient. By using
these features in your applications, you are joining us in our effort to make the web faster.
For this, we thank you! Let us know of any issues and feature requests by posting to the
developer forums of your favorite APIs or by leaving a comment on this post. Happy
hacking!
Anton Lopyrev is an Associate
Product Manager for Google APIs Infrastructure. He is a computer graphics enthusiast who is
also passionate about product design.
Sven Mawson is a Software
Engineer working on Google’s API Infrastructure. He believes well-designed, beautiful APIs
need not sacrifice performance.Posted by Scott Knaster,
Editor
