Thursday, November 22, 2007

Google Custom Search with Bloglines Search OS X mashup

This was fun, but very geeky. It's the kind of experiment that I only get to do on lazy holidays (we do a low key family thanksgiving).

It started because I wanted my Google Co-Op Mac OS X Search custom search engine to also search the Mac OS X blogs I read. (Note: Google may be sunsetting some of the services referenced here.)

I figured I'd have to add my blogs one at a time, but that's kind of tedious. It would be more fun if there was a way to add them all at once.

It turns out Google's custom search engine service (formerly Google Co-Op) now allows one to enter the URL of a web page, and Google will add each linked item as a search engine.

That reminded me of a Bloglines technique I learned exploited about three years ago. One of the key features of Bloglines is the ability to publish your blogroll (including OPML export) and, with a bit of trickery, to publish it in such a way that portions can be embedded in an IFRAME (see example on the right side of our family newspaper).

I went back to the web page I created four years ago and extracted the OS X portion into a separate file.

I then gave Google's custom search engine control panel the URL of my OS X Bloglines file. The result looks like this (bolded blogroll file)
docs.info.apple.com/*
forums.macosxhints.com/*
http://devworld.apple.com/documentation/MacOSX/
http://discussions.apple.com/*
http://forum.parallels.com/
http://googlefaughnan.blogspot.com/
http://www.macintouch.com/*
http://www.parallels.com/en/support/
www.faughnan.com/blogroll_mac.html
www.macosxhints.com/*
I don't think it's working quite yet, but I'm sure it will be live soon.

Update 11/22/07: It's broken. I tried it with both dynamic html (original) and static html, but Google is not finding test results belong to the blogs I added "by reference". If I add the individual blogs by hand Google does search them, so this looks like a bug. I'll try it again next week.

I've posted a "bug report".

Update 11/23/07: Yes, it's a bug. OMR has found:
So it seems that the makeannotations tool is not working reliably at this time.
It frequently returns the "Bad Request" error, even for a valid request.
Update 11/24/07: It's probably coincidental, but since OMR replied to my post and identified the bug Google's makeannotations tool has started working. Unfortunately it doesn't seem to support javascript generated pages, at least not the way Bloglines generates the pages. I'm going to see if bloglines wants to look into this.

No comments: