Danny Gorog wrote an excellent review of Booko in APC Mag!

Well, I think I’ve found the bug that makes IE not work so well with Booko.  The section of the page which displays the prices is updated via AJAX (with a GET post), but the first chunk of text for that section (with the spinning icons) gets cached by IE. The solution is to add an Expires header to the response. The actions which respond to GET requests got the following code:

headers["Expires"] = "Mon, 26 Jul 1997 05:00:00 GMT"

I should probably put a more correct value in there – but it’s always going to be set in the past at some point.

Well, I’ve added a blurb to Booko’s front page. Should have done this a while ago – makes it feel more like a proper website. It’s probably a bit too much text, but I’ll work on it.

Fixed a bug reported by Timo and Dan B (And probably others) – clicking “Add to Cart” before the prices were fully loaded made the cart behave pretty strangely – usually you’d get several items in your cart or some such. The problem was that the cart price can’t be calculated until all the prices are available. I’ve hidden the link to add items to the cart until all the prices are correct.

Figured out how to set a timeout for getting prices from stores – and set it to 15 seconds.  May need to tweak this a bit. Some shops were taking too long to respond.  (Thanks Dan B for the suggestion)

Tweaked the HTML for the cart, adding another separator. This site is now, without doubt, very separated.

I’ve updated Booko to now check multiple sites for cover art.  All the sites I use for data will return “No Image” style images for books which they don’t have covers for.

Such as this:

There’s no way to tell from the URL of the image if it’s one of these pretend covers, so I had to be a bit more clever about it. Since the image itself is always the same, I calculate the md5 sum of the image – if it matches a known “No Image” image, I move on to the next site.

Given an image URL, I calculate the md5 sum like this:

digest = Digest::MD5.hexdigest(open(url).read)

Which downloads the image and calculates the hash in one nice easy step.  So, from now on, Books will be far more likely to have correct cover art images.

Some more small changes this arvo:

• Changed the separator from solid to dotted, moved it around a bit, and added separators between the recent searches / your searches sections.
• The “Last Updated” section now tells you how long ago the book was updated in relative terms.
• The Cart has been shrunk and now has a dotted line to clearly separate it from the other content.

I’ve added another feature to Booko – Booko now remembers books you’ve recently viewed. This should save you from having to search again, or add them to your cart to keep track of books you’re interested in. Of course, it’s only session based – lists will be different on different computers / browsers.

I’ve updated the program which runs off and finds the prices on the 15 shops Booko scans. Previously, it would search for a single book at a time, but would search all shops at the same time.  Now it will check *all* books which are waiting for prices at the same time.  

One of the problems with the old approach is that some stores seem to be down as much as up – the price grabbing code will patiently wait until all shops have answered or the connection times out. When a site is down, other users may be waiting for over a minute for their books to be updated. The new approach means everyone should be getting prices as soon as they’re available.  

So, if you’re the kind of person who does a search, then opens multiple tabs from the search result, you should find the results are populated much, much faster.