Few programming tasks illuminate the differences between commonly used languages as clearly as that of retrieving a web document. When it comes to shell scripts, admins often turn to the curl utility, which transfers the data behind a URL without much ado and sends them to the standard output.

But, what if the URL points to a black hole? Or the server denies access? And what if the server returns a redirect? For example, curl http://google.com does not return the expected HTML page with the search form but just a note that the desired page may be available on www.google.com. Armed with the -L option, however, curl follows the reference and then returns the data from the source it finds there.

What happens with a huge file like a 4K movie containing many gigabytes of data? Will the process exhaust your RAM because it attempts to swallow everything in a single gulp? Does encryption work automatically for an HTTPS URL using the SSL protocol, and does the utility check the server's certificate correctly so that it does not fall victim to a man-in-the-middle attack? Similar to good old curl, popular programming languages offer all of this, although often only as an add-on package and often requiring quirky approaches.

Go, Hipster Go!

The relatively new Go comes with web support out the box; it is included in the net/http package with exemplary SSL support. Go programmers are required to handle any errors that occur immediately after function calls and cannot excuse any lapses by claiming that any exceptions that are thrown will be dealt with somewhere down the line, be it as hard-to-read stack traces when the program terminates. This is no doubt a philosophical question with a similar scope to finding your perfect partner or choosing between the vi and emacs editors.

Listing 1 [1] shows three different error checks, because various things can go wrong while fetching a document: the act of generating the request (e.g., unsupported protocol), the server returning an error status code (404 for document not found), or an error occurring when retrieving the data via the convoluted pipes of the Internet, which could lead to the data stream terminating abruptly. Functions in Go in general like to return two parameters, a result and an error structure, whose value is set to nil if everything has gone smoothly.

It is also interesting to note that the net/http package first executes the request, and then receives the status code of the server in line 14, but still allows some time before scooping off the body data. In fact, the data is grabbed later on by ReadAll() in line 20, courtesy of the io/ioutil package. The Close() method on the object body then finally indicates that request processing has finished and that Go can dispose of the data. The related command already can be seen in line 19 but is delayed till the end of the currently executing function by the nifty defer keyword.

Lean Python

The Python 3 implementation in Listing 2 using the state-of-art requests library is more compact, because Python throws exceptions that developers can test for later on in a central location. If the request specifies a nonexistent protocol (e.g., abc://), verification of the server's SSL certificate fails, an I/O error occurs, then an appropriate exception is thrown, which the code either checks separately or in one fell swoop as shown in line 11.

This approach is certainly more convenient, but the typing you saved here can come back like a boomerang later on if an exception is washed up from the depths of the code, like from inside a library developed by someone else. Also the readability suffers because it is not entirely clear which line in the try block threw the exception. The topic has started already countless, hard-to-settle, and virtually infinite discussions.

It is also funny that r.encoding states that the page is encoded in ISO-8859-1 after a request for google.de. Only manual shift to utf-8 in line 7 causes the following r.text to output the content with UTF-8 encoding. A status code other than 200 does not throw an exception by default and needs to be checked manually. Programmers who like things even more compact can use the raise_for_status() method to throw an exception if the server reported something other than 200.

Ruby Catchall

Ruby also comes with a built-in HTTP module named net/http; however, before sending requests, it needs to analyze the URL with another class, URL. This is far too much work for Ruby-on-Rails fans and prompted the developers to create some Ruby Gems that do the whole thing in a single action. Like Python, Ruby throws exceptions from the depths of the code, and developers can thus deal with errors further upstream.

Listing 3 shows the catchall block, which starts with begin in line 6, catches exceptions as of rescue, and finally terminates with end. The standard library does not follow redirects in its default setting. It interprets the Google.de website as ASCII-8BIT, which is probably equivalent to the ISO-8859-1 encoding identified by Python in this case.