Request for Feedback on New Feedster Feature or "Its all Andy's Fault" ...

About an hour or three ago, I got an interesting IM from Andy.  He had an opml file of related RSS urls and wanted a way to look at what was going on with them over the past week's time.  He correctly pointed out that I really do have all the pieces to do this pretty easily -- all I need is a way to both query temporally and constrain it to a subset of the database.  Here was his opml url and as you can see it's all about HCI -- human computer interaction.  Now not that I don't have other things to do but Andy's a wicked smart guy and someone I respect.  So to the "code cave"!

Here you go: View New Feature

So the feedback I'm looking for is something on the order of this:

• What do I name it ?
• Is it useful ?
• Do you like it ?
• And would you actually use it ?

It's definitely similar to a web based aggregator when used with your own opml file but I'm wondering if Andy isn't on to something here with the idea of topical groupings.  It gives a nice way to scan across multiple blogs all at once and get caught up.

Please leave comments with your feedback.  Much appreciated!

Note: This is not in any way intended to compete with true client side aggregators, particularly those that support Feedster. 

Note: This is just a prototype.  Its limited and specifically the RSS feeds for it WILL NOT WORK!

Note: The feed was put togehter for a blogging BoF at CHI 2003 as an introduction to the blogging world.

7:20:07 PM      Google It!   comment []    IM Me About This   

The Smarter You Are ... The More Subtle the Bug

I've been asked by users for quite a few days (ok weeks) now why I'm not showing paginated search results and a next button -- you know -- like every other search engine in existence.  The answer was pretty simple -- hadn't gotten around to it yet actually.  And its interesting but not all that interesting so I back burnered it until today.  Oy.  It isn't that the problem was all that hard actually; easier than I expected to be honest.  Its that that last bug was subtle.

Example of Feedster with Paginated Results 
(using Pirillo as a test query)

Now I do ok intellectually but I honestly don't think I'm all that smart as a general rule of thumb.  I've hired smart people before and they make me feel like a drooling idiot.  Still I'm definitely both skilled and experienced and I read a lot which people sometimes attribute to smartness.  And today those things reared up and smacked me in the head and then dribbled said head down the virtual basketball court and scored an extra point. 

I got the page navigation in place and passed some sample urls over to my friend James who nicely kibitzed on it and pointed out some foolish errors on my part.  In specific, I was initially testing with just the relevance sort and needed to support date and rank sorting.  No big.  All was frothy happiness, sweetness and light. 

Then I changed to date sort. Storm clouds filled the horizon and the room grew dark.

You see what happened is the number of results went from 10 to 6 -- and I was telling my routines to produce 10 results.  That's odd I thought.  Then I tried it again and got the same thing.  Then I tried sorting by Rank and I got 10 results as well.  Hm.....  Scratch head.  No luck.  Display query.  Nothing unusual.  Scratch head some more.  Nada.  Try query in actual database without my own results code and get back 10 results.  *Lightning*.  Now at least I know it is me.  Not that I thought it wasn't but you know -- any straw you can grasp when debugging.

So I groveled over the code for quite a bit.  Then it hit me like a bolt of lighting: "Duplicates"!  By jove I'm removing duplicates!  And since I'm not doing it at the database level to eliminate the performance hit of a distinct query in conjunction with a full text query, I get back 10 results and then drop it back to 6.  Finally it makes sense.  And the reason that it didn't happen in Relevance and Rank order testing ?  Simple -- there weren't any duplicates then; dupes are most common with chronological results when Post A is sent to Blogs B and C.  So if I really think this is the cause then it actually has been in Feedster since the beginning.  Sure enough -- do the right search, get back "50" results and find out that its only 36. 

So there you have it -- I was "smart" in getting rid of dupes and even (perhaps) "smarter" in trying to avoid a database performance hit by avoiding distinct.  And what did that get me?  Just a wickedly subtle bug. 

 

So here you go:
Scott's 3rd Law of Coding: The Smarter You Are ... The More Subtle the Bug

 

Note: I know that when you go to the 21st page of results, the window of #s doesn't go 21 to 40.  That is something for a future fix.  The next button works correctly and that was my bigger concern.  And, yes Virginia, the limit, offset variables aren't fixed in code and can be user altered should that be necessary.  If limit needs to be greater than 100 then contact me directly since right now there's a boundry condition. 

***Any problems that you find with this, please let me know asap.  Much appreciated.***

12:47:56 PM      Google It!   comment []    IM Me About This