Anyone here know much about Java?

[andrewducker]

I have user input for a URL field. I want them to be able to enter anything from "http://andrewducker.wordpress.com/xmlrpc.php" to "andrewducker.wordpress.com" and be able to end up at the same end point.

I've wasted a couple of hours messing around with the various constructors for URL and not got to anywhere satisfactory, should I just do string checking and construct it myself?

I should make it clear - I always want the /xmlrpc.php bit to be what's on the end of the URL, that's a Wordpress standard, so I don't need to do any complex discovery. I just need to append that if it's not there.

I was hoping that someone would have written a class that could append bits of URLs together, but the basic stuff in the built in URL class doesn't quite cut it.

Comments

[pseudomonas]

Can you do all the manipulations as Strings and then just convert it to URL at the end?

[andrewducker]

If you wander over to the LJ conversation you'll see that's the option I've pretty much decided on!

[pseudomonas]

Ah, right. I only follow you here.

[andrewducker]

Sadly, most people are LJ-people, and the comments tend to be almost exclusively over there. Tis a shame.

[bugshaw.livejournal.com]

Would "Enter your Wordpress domain" work?

[andrewducker]

I reckon I'd get a 40%/40% split between "andrewducker.wordpress.com" and "http://andrewducker.wordpress.com", with some "http://andrewducker.wordpress.com/" thrown in for good luck.

[drplokta]

Regardless of programming language, that sounds to me like a job for a single regular expression -- something like (untested): s/(htttp\:)*(\/)*(.*)(/(xmlrpc(\.)*(php)*)*)*/\3/

[andrewducker]

That might recognise if what they've input is one of those URLs, but my actual need is to convert it into the final form...

[drplokta]

The regular expression should extract the actual domain name with the optional extraneous bits removed, so you can then prepend "http://" and append "/xmlrpc.php" and you're done.

[andrewducker]

Aaaah, gotcha.

Of course that's not going to work if their wordpress setup is behind https, but I don't know if that's at all likely. Or if it sits in a subdirectory, so that it's
"https://AShareddomain.co.uk/andy/blog/xmlrpc.php"

[skington.livejournal.com]

Sounds like a series of regular expressions is your best bet, then. In Perl, that would be e.g.

if ($url !~ m{^ https?:// }x) {
$url = 'http://' . $url;
}
if ($url !~ m{ /xmlrpc\.php $ }x) {
$url .= '/xmlrpc.php';
}

And then use your standard libraries to check whether that URL is valid or not. (Validation step might just be to call the resulting "url" and see if it works.)

[pozorvlak.livejournal.com]

The https issue is easy to fix: add s? to tge regex after http. The other issue is handled already - the .* is greedy, and consumes as many characters as it can.

Of course, we're now up to two edge cases with no guarantee we've thought of them all, which is the usual weakness of regexp-based approaches.

[andrewducker]

Yup. I'm thinking I might be as well off with something like:
if first four characters are not "http" prepend "http://"
if last 10 characters are not "xmlrpc.php" then (check if the last character is a / and append one if not) append "xmlrpc.php"

As that basically satisfies my problem without involving regexes at all, which might be simpler :->

[pozorvlak.livejournal.com]

Typically character-index based solutions have all of the brittleness and none of the readability of rexexp-based solutions, but in this case I think your algorithm makes sense.

[pozorvlak.livejournal.com]

*regexp

[andrewducker]

Actually, your regex looks like it will cope with that. I'll have a play tonight and see if I can make it work. Cheers!

[drplokta]

I think you'll need to add an (s)* to cope with the https possibility.

[pozorvlak.livejournal.com]

I for one find your use of * in place of ? surprising. A hangover from pre-PCRE regexes?

[drplokta]

I just simplify matters for myself by not bothering to remember what ? does unless I actually need it. In this case, it doesn't really matter if something like "http:http://" gets stripped off the beginning or "xmlrpc.phphp" gets stripped off the end, so I didn't bother. If it was important to match only a single occurrence, I'd look it up again. Yes, most of the *s can probably be ?s, except in the "(\/)*" bit where * is needed.

[pozorvlak.livejournal.com]

I spot one bug: "htttp" instead of "http".

[strawberryfrog.livejournal.com]

If that was c#, I would not use a regex - I'd
1) construct a Uri from the string
2) read off some of the properties of the uri. Something like uri.Scheme + uri.Host + "/xmlrpc.php" should be what you want (unless port numbers are involved).

[andrewducker]

Only works if the install of wordpress is in the root.

[strawberryfrog.livejournal.com]

Good point. That is in the "AbsolutePath" property but you might have to massage it - remove everything after the last slash and substitute "xmlrpc.php"

[ami-bender.livejournal.com]

Would something like this be useful?

http://hc.apache.org/httpcomponents-client-ga/httpclient/apidocs/org/apache/http/client/utils/URLEncodedUtils.html

[andrewducker]

That doesn't seem to do what I want. Any pointers?