##Chapter 7 - REST Web Services
###Introduction
REST stands for representational state transfer and is an architectural principle rather than a specific technology or standard. It was introduced in a doctoral dissertation by Roy Fielding in 2000 (see the on-line syllabus for references). The REST-style architecture is just like the web. There are requests and responses that are the transfer of resources. Resources are any addressable object (as in something with a URI on the web). HTTP follows these principles summarized here:
- Client/server interaction
- Uniform interface
- Stateless (no context stored between requests)
- Cacheable (we cover caching in chapter 8)
- Layered (intermediaries exist that are transparent to clients)
A criticism of SOAP-based web services is that they are overly complicated and work in a really different way than the web even though they use it as infrastructure. For example, new vocabularies can be created at will using XMLSchema so there is no uniform interface.
Similarly, there is no obligation for SOAP-based web services to support loosely-coupled statelessness or be cacheable, although general SOA principles would like those to be true. Stateless server-side components are less complicated to design, write, and distribute across load-balanced servers. A stateless service not only performs better, it shifts most of the responsibility of maintaining state to the client application. There are, in fact, many situations where the complexity of WS-\* is needed, but the vast majority of scenarios do not require it.
So REST web services are a better choice in those cases. In fact, the majority of web application APIs are RESTful as we will see below.
A RESTful web service is implemented with HTTP and has the following specific requirements:
- All resources are identified by URIs.
- Only the HTTP methods are used as interface (POST, GET, PUT,
DELETE).
- Resources have an internet type (XML, JSON, etc.).
All resources are identified by URIs. This is the way that the web already works. RESTful web services extend this idea to give everything a URL-based resource identifier - not just web pages. Consider the cd examples we saw on w3schools. The basic XML file there is a catalog of music cds (see w3schools XML). We can improve that catalog by giving IDs to all items as in listing 7.1. XML requires that all IDs are unique within a document.
< ?xml version="1.0" encoding="utf-8"?>
Empire Burlesque
Bob Dylan
Columbia
10.90
1985
Hide your heart
Bonnie Tyler
CBS Records
9.90
1988
...
Listing 7.1. The cd catalog with IDs.
We can then create URLs to get the entire cd catalog or an individual cd. For example, we could use the URL:
`http://www.w3schools.com/catalog/cds`
to get the entire catalog (the entire XML document of listing 7.1) and:
`http://www.w3schools.com/catalog/cds/2`
to get the Bonnie Tyler cd (just that `id="cd2"` XML subtree in listing 7.1).
All the resources associated with the catalog can be identified with URLs. In reality, of course, some program is going to have to do things like find the cd with a specific id. If we used a CGI program or PHP to do that, we would need a URL something like:
http://www.w3schools.com/getCD.php?cd=2 (not a real url). The getCD program searches the file or database for the id=2 and returns the result to the requestor in XML format. RESTful APIs like to use the better looking URLs such as those in the boxes above. Most web frameworks such as the one we will discuss in chapter 9 have routing capabilities that allow a requestor to submit the good-looking URLs and they are mapped correctly to programs as the uglier URLs make explicit. These good- looking URLs expose an easily understood, directory structure-like path to represent a resource.
Only HTTP methods are used as a uniform interface. POST should be used for creating resources. For example, to create a new entry in the cd catalog, we would use listing 7.2. (There would usually be some way to auto-increment the id instead of explicitly including it as here.)
POST /catalog/cds HTTP/1.1
Host: www.w3schools.com
...(more headers)
< ?xml version="1.0" encoding="utf-8"?>
The Essential Weird Al Yankovic
Weird Al Yankovic
Columbia
15.00
2002
Listing 7.2. POST.
A developer could use standard web technology to do the non-RESTful implementation. For example:
`GET /addCD.php?id=cd3`
This would work just fine, but would not be RESTful because resources should only be created with POST according to the uniform interface.
The GET method is only for data retrieval only. GET should be free of side effects (called the idempotence property). It would be used to get information from our cd catalog as the following example which would return the Weird Al cd information:
`GET /catalog/cds/3`
The PUT method should update a cd in the catalog. Figure 7.1 shows this example. Note that the company tag was updated to Sony from Columbia.
Again, a developer could call a program explicitly and do the update with GET, but that would not be RESTful. The DELETE method is used to destroy resources.
![PUT](is651-images/f7-3_opt.png)
Figure 7.1. PUT.
XML is commonly used by REST web services, but it is not required. Any web format is acceptable. A common format used by RESTful APIs is javascript object notation (JSON). This format is equivalent to XML,
less verbose, and it integrates transparently with the javascript language. Basically, the angle brackets of XML are replaced by more economical curly braces and commas. If a javascript program imports JSON, the objects are created automatically by the javascript interpreter and so no XML parsing is needed. Figure 7.3 shows JSON and the equivalent XML. We will also see JSON used in the end of chapter exercises.
A simple example of JSON in figure 7.2 shows how JSON works in javascript. The JSON is included directly in the program here, but could easily be the result of a REST query. It results in the simple text output to the browser:
`Sally Green 27`
with no XML parsing required. The dot notation for objects is used to access the objects from the JSON.
![JSON (Recall that arrays begin with index 0)](is651-images/f7-4_opt.png)
Figure 7.2. JSON (Recall that arrays begin with index 0).
![JSON vs. XML](is651-images/f7-5_opt.png)
Figure 7.3. JSON vs. XML.
REST APIs are very popular for public web services for web applications. Twitter, Flickr, YAHOO, Facebook, THe New York Times, NPR, and most other popular destinations on the web offer RESTful APIs for developers to interact with their applications. Let's look at a detailed example using the UK [Guardian newspaper](http://www.theguardian.com/us) and see more examples in the end of chapter exercises.
The Guardian API is called Open Platform and requires that you get a free api key as explained in the slides. We will use this api to illustrate REST API basics here and in the homework.
The Guardian REST API is documented here (read 'What is the Open Platform?' there):
- http://open-platform.theguardian.com/
That documentation describes how to construct a URL for the REST API. See the Services/Content API/Reference Guides/Search link for the [specific docs for the search endpoint](http://www.theguardian.com/open-platform/content-api-content-search-reference-guide?guni=Keyword:news-grid%20main-3%20Trailblock:Pickable%20with%20editable%20override:Position1:sublinks&guni=Keyword:news-grid%20main-3%20Trailblock:Pickable%20with%20editable%20override:Position1:sublinks). Using that Guardian documentation, we can create the URL below and see the truncated JSON result in listing 7.3. It is not real api key, but you can put this url in the address box of your browser and see it work after you get your key. The link below works because I deployed it in php (see the slides).
- [content.guardianapis.com/search?q=syria§ion=news&from-date=2013-09-01&api-key=xyz](http://userpages.umbc.edu/~canfield/guardian/guardian2.php)
Note the following about the JSON result in listing 7.3:
- The result is JSON and not a web page. So it is not intended for regular web users, but for developers who make these requests in their programs and those programs process the JSON result for display to the users.
- One could add &format=XML for xml output. This is contrary to the docs (!) that say that XML is default, but actually JSON is default as it is for almost all REST APIs.
- Try a few more requests using the docs.
One problem that comes up is that web browsers have a security limitation that requires any program running in the browser (using javascript with ajax) can only return results from the same domain that the original web page came from. This is called the cross-domain restriction or the same-origin policy. This restriction permits scripts running on pages originating from the same site to access each other's methods and properties with no restrictions, but prevents access to pages on different sites. This policy also applies to XMLHttpRequest (ajax). Our Guardian example would therefore not work for a web page we created on gl since our web page is from umbc.edu and Guardian is on theguardian.com. There are actually ways to get around this such as [jsonp](http://json-p.org/) or [cors](http://www.html5rocks.com/en/tutorials/cors/). We will show how to use jsonp to call server-side PHP program and parse its results.
We will use a server-side PHP program (which has no such cross-domain security restriction) to retrieve the JSON or XML from Guardian and then send it back to the user that requested the web page from gl. We will use PHP to issue the request using the Curl library. The Curl library offers a way to send a URL programmatically and handle the response. See listing 7.4. The structure of the PHP program is simple:
- The $request variable holds our REST request URL.
- All the curl lines are just boilerplate from the curl library. You only have to set the $request variable in the block as the third parameter (line 4) of setopt function and the remainder is the same unless you want to modify something such as the HTTP method - see: http://us2.php.net/curl.
- The $json variable gets the result which will be the response from the REST web service (JSON in this case).
- The echo command prints out the JSON to the browser. No parsing was done here and just the raw JSON is shown. In a real application, this would not happen. We could have done this by entering the guardian URL directly in the browser! We will improve this next by processing the return on the client-side.
- We note that JSON result is actually wrapped with a function called jsonProcessFn. It is needed for jsonp. You will see the function's definition in the client-side JavaScript code.
{
"response":{
"status":"ok",
"userTier":"free",
"total":23,
"startIndex":1,
"pageSize":10,
"currentPage":1,
"pages":3,
"orderBy":"newest",
"results":[{
"id":"news/2013/oct/03/the-best-pictures-of-the-day",
"sectionId":"news",
"sectionName":"News",
"webPublicationDate":"2013-10-03T20:57:36Z",
"webTitle":"The best pictures of the day",
"webUrl":"http://www.theguardian.com/news/2013/oct/03/the-best-pictures-of-the-day",
"apiUrl":"http://content.guardianapis.com/news/2013/oct/03/the-best-pictures-of-the-day"
},{
"id":"news/2013/oct/02/the-best-pictures-of-the-day",
"sectionId":"news",
"sectionName":"News",
"webPublicationDate":"2013-10-02T21:43:38Z",
"webTitle":"The best pictures of the day",
"webUrl":"http://www.theguardian.com/news/2013/oct/02/the-best-pictures-of-the-day",
"apiUrl":"http://content.guardianapis.com/news/2013/oct/02/the-best-pictures-of-the-day"
},{
"id":"news/2013/oct/01/the-guardians-photo-team...-today",
"sectionId":"news",
"sectionName":"News",
"webPublicationDate":"2013-10-01T22:51:04Z",
"webTitle":"The best pictures of the day",
"webUrl":"http://www.theguardian.com/news/2013/oct/01/the-guardians-photo...-today",
"apiUrl":"http://content.guardianapis.com/news/2013/oct/01/the-guardians-photo...-today"
},{ ...
}]
}
}
Listing 7.3. The result of the Guardian request.
The curl command actually works on the command line. Try the following at the gl prompt:
`$>curl 'http://content.guardianapis.com/search?q=syria§ion=news&from-date=2013-09-01&format=XML'`
< ?php
$request = 'http://content.guardianapis.com/search?q=syria§ion=news&from-date=2013-09-01&format=XML';
$ch = curl_init(); // initialize curl handle
(http://us2.php.net/curl)
curl_setopt($ch, CURLOPT_URL, $request); // set url - must match above
curl_setopt($ch, CURLOPT_FAILONERROR, 1);
curl_setopt($ch, CURLOPT_FOLLOWLOCATION, 1);// allow redirects
curl_setopt($ch, CURLOPT_RETURNTRANSFER,1); // return into a
variable
curl_setopt($ch, CURLOPT_TIMEOUT, 3); // times out after 4s
curl_setopt($ch, CURLOPT_HTTPGET, 1); // set GET method
// run whole process and put the response from the service into the $xml variable
$xml = curl_exec($ch);
curl_close($ch);
header("Content-type: application/json"); //send http header
echo "jsonProcessFn(".$json.");"; //wraps result in a function call
?>
Listing 7.4. Curl in PHP.
Now, let's see how we can use this with JavaScript and JSONP to create a simple application that uses the Guardian REST web service. Listing 7.5 shows the javascript code.
Note the following about the code:
- The code has two script tags. The first one includes two functions to process JSON results from php. The second one includes the php file for Listing 7.4 so the jsonProcessFn function in the first script will called with JSON result.
- The functions in the first script tag are programmed using object-oriented JavaScript. Each of the function appends some information to the div tag with #guardian (# means id).
- For example, the callback outputs all the JSON in a variable called data, so we can access any part of it with that variable like data.response.userTier which would be `free`.
- This example is available at: guardian.html.
Listing 7.5. Guardian call using JavaScript with JSONP.
This completes our treatment of REST although we will look at another API in the end of chapter exercises. We have seen:
- What REST web services are and how they differ from SOAP- based web services. REST assumes a point-to-point communication model which is not usable for a distributed computing environment where messages may go through one or more intermediaries. Those are more complex situations where SOAP-based web services are a better fit.
- What a REST API is for a web service.
- How to use the documentation of a REST web service to construct a URL using the API.
- How to use JSON and XML as the return format.
- How to create a web browser-based application that uses a REST web service.
###Chapter 7 Exercises
Do the end-of-chapter exercises for each chapter of the book by following the link in the on-line syllabus.