October 14, 2006
searchfeedr Does ESP...
I was determined to lay off searchfeedr for a while, but couldn't resist doing this half hour hack - an attempt at ESP... by which I mean, rather than requiring the user to select the sort of input that provides the search limits (delicious feed, web feed, or page feed, for example), the interface tries to guess what sort of search limit input is provided.
Here's the demo: searchfeedr ESP interface
(I know, I know, I said no more URLs and I'd stick with labs - but this removes the search limit input selection buttons, though I suppose I could add them back in to allow the user to over-ride a mistaken guess about the limit input type.)
At the moment, the system guesses at the following:
- things that look like somethingHere/somethingThere end up as the limit source for a deliSearch;
- things that have /rss/ in the URL path, or end with rss or xml are used as the limit source for a feedSearch;
- everything else is a pageSearch (i.e. it gets scraped for HMTL links...)
There are a couple of downsides - one is that I have removed the linkSearch and relatedSearch options (but I have an idea for how to combine them with pageSearch, which I'll try out in a week or two, maybe). The other - as hinted at above - is that the user has no say in over-rding a bad guess by the system - which means that some input limits may be broken...
I also fixed the handling of delicious popular tag feeds. These aren't available as JSON, so the deliSearch optimisation didn't work with them, but now I trap for "popular" and treat it as feedSearch instead by using the RSS feed rather than trying to consume the non-existent JSON feed.
Anyway, have a tinker and let me know what you think (in particular, I'm after examples of things that don't work...)
PS another 20 minute hack - if you leave the left hand text box blank (i.e. if you omit search limits) you can use the right hand text box to make a normal search engine query. Don't forget, you get to choose which search engine you can use...
Posted by ajh59 at October 14, 2006 07:18 PM