ericwaller

A query for a single-prefix term (i.e. "yank") is technically O(log(N)+M), where N is the # of items in the sorted set, and M is the # of items returned. But since in practice we often want a small # of items, we can keep both N and M small.

One tradeoff we made is that this approach doesn't support incremental updates easily, you can always add new items, but it would take a good bit of work to remove items that have expired, or update the scores of items that change.

As for memory usage, we only store each item once (in a redis hash table), and use the unique ids as entries in the sorted sets. Which sort of approaches a trie, since instead of:

ya -> yan -> yank -> yanke -> yankee -> yankees -> [data about yankees]

we have:

ya -> 101

yan -> 101

...

101 -> [data about yankees]

Can you expand on this point? This is one aspect of TDD, which seems to focus most intensely on unit testing, I've never understood. Why wouldn't full-stack, acceptance testing be just as effective?

You're welcome, and thank you for delayed_job!

The stadium itself is actually tiled bitmaps, we're using Raphael/svg for the section/row highlights. It would have been great to use vector graphics the whole way, for one it would have given us continuous zooming instead of the predefined levels we're doing now, but the performance just isn't there yet in IE.

Mouseover on the whole section is definitely nicer but unfortunately it caused some performance problems in IE (rendering a ton of paths is slow). Hopefully we can figure out a more performant way of doing that.

One disadvantage of closing the popup on any click is that you might want to drag the map with the popup open, though it would be nice to not have to click the 'x'. Certainly something to consider.

Collecting the venue data was definitely a bit of a challenge. We used a combination of publicly available seating charts, google earth (believe it or not) and our own transaction data. It's easy enough to get a rough idea of the section sizes/layout. Getting row counts and names was a bit tougher, and our transaction data was a big help there.

It seems like you're correcting me, but how is your point anything but tangential? If you're wondering why inexperienced drivers aren't also over-represented, it's probably because not many 16-21 year olds drive priuses.

But since everyone knows the odds, the lotto market is perfectly efficient and no one can win in the long run -- this is what makes it gambling.

DMCA is the new DOS

I noticed a similar gain (10-12 req/s to 25-30req/s) for action cached pages after switching out nginx+passenger for nginx+thin. This is on a 256mb slice, and seems totally counter to the general opinion that passenger is great for VPSes.

2. The automatic follow is mostly to avoid bumping into our follow limits (following a bunch of people who don't follow you back makes you look like a spammer).

3. That's the hard part, isn't it? We're starting by trying to reach out to the sports gambling community (who are heavy forum users), and have considered a few interesting things like running a retweet contest for football tickets.