March 2007 Archives

Step 1: Pick a community designator

I've been using the community name followed by a dot. The prefix goes in front of category description to make a set of unique tags for the aggregator. When someone wants to add something for the site they tag it with this "community.category" tag. In my example, when someone wants to list a new Quaker blog they use "quaker.blog", "quaker" being the community name, "blog" being the category name for the "New Blogs" page.

Step 2: Collect the community prefix and category name in Pipes

You begin by going into Pipes and pulling over two text inputs: one for the community prefix, the other for the specific category.

Step 3: Construct these into tags

Now use the "String Concatenation" module to turn this into the "community.category" model. The community input goes into the top slot, a dot is the second slot and the category input goes into the last slot.

Now, when you have a tag in Flickr with a dot in it, Flickr automatically removes it in the resultant RSS feed. So with Flickr you want your tag to be "communitycategory" without a dot. Simple enough: just pull another "String Concatenation" module onto your Pipes work space. It should look the same except that it won't have the middle slot with the dot.

Step 4: Turn these tags into RSS URLs

Pull three "URLBuilder" modules into Pipes, one for each of the services we're going to query. For the Base, use the non-tag specific part of the URL that each service uses for its RSS feeds. Here they are:

Under path elements, put the correct tag: for Del.icio.us and Google it should be the community.category tag, for Flickr the dot-less communitycategory tag.

Step 5: Fetch and Dedupe

Fetch is the Pipes module that pulls in URLs and outputs RSS feeds. It can also combine them. Send each URLBuilder output into the same Fetch routine.

Since it's possible that you'll might have duplicate posts, use the "Unique" module to deduplicate entries by URL. Through a little trial and error I've determined that in cases of duplicates, feeds lower in the Fetch list trump those higher. In the actual Pipe powering my aggregator I pull a second Del.icio.us feed: my own. I have that as the last entry in the Fetch list so that I can personally override every other input.

Step 6: Sort by Date

With experimentation it seems like Pipes orders the output entries by descending date, which is probably what you want. But I want to show how Pipes can work with "dc" data, the "Dublin Core" model that allows you to extend standard RSS feeds (see yesterday's post for more on this).

Google Blog Search and Del.icio.us feeds use the "dc:date" field to record the time when the post was made. Flickr uses "dc:date.Taken" to pass on the photograph's metadata about when it was taken. Pipes' "Rename" module lets you copy both fields into one you create (I've simply used "date"), which you can then run through its "Sort" module. Again, it's a moot point since Pipes seems to do this automatically. But it's good to know how to manipulate and rename "dc" data if only because many PHP parsers have trouble laying it out on a webpage.

Update: it's all moot: according to a ZDNet blog, "Pipes now automatically appends a pubDate tag to any RSS feed that has any of the other allowable date tags." This is nice: no need to hack the date every time you want to make a Pipe!

Step 7: Output

The final step for any Pipe is the "Pipe Output" module.

In action

You can see this published Pipe here, and copy and play with it yourself. The result lets you build an RSS feed based on the two inputs.

Step 1: Input tags

The Del.icio.us social bookmarking system runs much of my aggregator: users see a post they like and bookmark it in Del.icio.us with a special tag.

The first step in Pipes is to collect the input (right). Pull the "Text Input" module (above) onto your Pipes work space. This lets you collect user input. Give it both a name and a prompt. In most instances it's fine that these be the same as the prompt won't be visible in the end. It's good to put something down in Debug for later on in the Pipes process.

Step 2: Construct the RSS call

We take our two input tags and use them to construct an URL by using the "URLBuilder". The base URL is Del.icio.us's RSS feed (http://del.icio.us/rss/). The URL builder adds the user input then the tag input to give us a valid URL (http://del.icio.us/rss/user/tag/).

Step 4: Grab the feed

Yahoo Pipes' "Fetch" module takes that URL input and turns it into an RSS feed. Shown to the right is Fetch with the final "Pipe Output".

See it in action

You can see how this fits together by going to my Del.icio.us Sample page on Pipes. You can make a copy and play with it yourself. Add "&user=username&tag=tagname&_render=rss" to that URL and you've made it a feed.

Okay so I've turned a Del.icio.us RSS feed into... a Yahoo Pipes RSS feed with identical input and output. Well, we're only getting started. Our input tags can be reused for other searches and spliced together inside of a more elaborate Pipe. That's where the fun starts and I'll get there soon.

The other advantage of sending things through Pipes is that we can easily rename fields. Del.icio.us, Flickr and other services often extend RSS standards by including metadata in "dc" fields, an abbreviation for the Dublin Core standards extension. A recent entry from my Del.icio.us feed includes this:

     <dc:creator>martin_kelley</dc:creator>
     <dc:date>2007-03-15T05:18:53Z</dc:date>
     <dc:subject>tech tech.design</dc:subject>

Standard PHP parsers like MagpieRSS and SimplePie often have trouble pulling dc data. With Pipes you can rename the fields you like; in theory that should make them more accessible to the parsers. You can also combine fields and use Pipes' Regex module to operate on them with regular expressions.