I put in 50 miles and it gives me people who are a 77 mile drive away. What are they using to judge distances? Why don't they use a mapping system that takes into account roads?
If I live on the CT coast I can see Long Island. Doesn't mean it is only a 10 mile drive!
I put in 50 miles and it gives me people who are a 77 mile drive away. What are they using to judge distances? Why don't they use a mapping system that takes into account roads?
If I live on the CT coast I can see Long Island. Doesn't mean it is only a 10 mile drive!
It's an incredibly difficult computational problem. It would have to calculate the driving distance between you and every other single OKC user and then query that distance. Then do that for every user. It would take a supercomputer just for that one feature. It's good enough just to use something like PostGIS to make a simple sql query that simply looks at straight-line distances between geographic coordinates.
Bullshit. Eat all the cheap ass BLTs you want. Just make sure there's some for her, too.
I feel you got the wrong end of the stick. I meant the eating a load of junk food and not caring rather than sharing. Like coming home and finding that they have eaten a whole bag of chips, the actual ones not crisps, or eaten half a loaf of bread as they were hungry.
Nope, I knew exactly what you meant. I don't see why living with someone has to mean you can't eat like that. I do it periodically.
It's an incredibly difficult computational problem. It would have to calculate the driving distance between you and every other single OKC user and then query that distance.
Not true, because you would only care about especially accurate driving distances within a relatively short radius (say 100 miles). Just find the straight-line distances first, and then get more accurate results only for the closer people.
Dating's easy if you live somewhere where there are people. ;^)
You also have to have something that gives you a legitimate reason to interact with people. Just having lots of people doesn't mean you'll meet people. I can walk up and down the quarter mile at RIT every day if I wanted and be surrounded by hundreds of people and meet no one. I need to have a class with people I haven't met who are open to interaction, or go to a club with people that aren't crazy (and open to interaction).
It's an incredibly difficult computational problem. It would have to calculate the driving distance between you and every other single OKC user and then query that distance.
Not true, because you would only care about especially accurate driving distances within a relatively short radius (say 100 miles). Just find the straight-line distances first, and then get more accurate results only for the closer people.
It shouldn't be that hard. All I use is google maps with a zip code to zip code travel search. While someone may be under 50 miles as the crow flies they are not 50 miles as the car drives.
It's an incredibly difficult computational problem. It would have to calculate the driving distance between you and every other single OKC user and then query that distance.
Not true, because you would only care about especially accurate driving distances within a relatively short radius (say 100 miles). Just find the straight-line distances first, and then get more accurate results only for the closer people.
How do you do that with an SQL query in a matter of milliseconds for a database of tens of thousands of people?
You also have to consider non-driving. Imagine a user in Manhattan. Sure, someone in Long Island is technically very geographically nearby, but what if they are beyond the reach of the subway, or not near a LIRR station? That can be up to an hour difference in travel time, but just a few miles difference as the crow flies.
Also, OKC doesn't have actual addresses in their database, nor would they add them. All they know is a very general location such as your town's or neighborhood's name.
It's an incredibly difficult computational problem. It would have to calculate the driving distance between you and every other single OKC user and then query that distance.
Not true, because you would only care about especially accurate driving distances within a relatively short radius (say 100 miles). Just find the straight-line distances first, and then get more accurate results only for the closer people.
It shouldn't be that hard. All I use is google maps with a zip code to zip code travel search. While someone may be under 50 miles as the crow flies they are not 50 miles as the car drives.
Not that hard. Google does it with their world-wide humongous supercomputer! Also, you aren't doing just one Google Maps calculation. You are doing hundreds of them. Imagine searching for directions from your house to the house of everyone on OKC simultaneously and in milliseconds.
It's an incredibly difficult computational problem. It would have to calculate the driving distance between you and every other single OKC user and then query that distance.
Not true, because you would only care about especially accurate driving distances within a relatively short radius (say 100 miles). Just find the straight-line distances first, and then get more accurate results only for the closer people.
How do you do that with an SQL query in a matter of milliseconds for a database of tens of thousands of people?
Precalculate.
You also have to consider non-driving. Imagine a user in Manhattan. Sure, someone in Long Island is technically very geographically nearby, but what if they are beyond the reach of the subway, or not near a LIRR station? That can be up to an hour difference in travel time, but just a few miles difference as the crow flies.
Your point? Either way, straight-line distance is clearly a poor metric.
Also, OKC doesn't have actual addresses in their database, nor would they add them. All they know is a very general location such as your town's or neighborhood's name.
I don't use OKC, but where is Steve getting figures like 77 miles from?
It's an incredibly difficult computational problem. It would have to calculate the driving distance between you and every other single OKC user and then query that distance.
Not true, because you would only care about especially accurate driving distances within a relatively short radius (say 100 miles). Just find the straight-line distances first, and then get more accurate results only for the closer people.
It shouldn't be that hard. All I use is google maps with a zip code to zip code travel search. While someone may be under 50 miles as the crow flies they are not 50 miles as the car drives.
Not that hard. Google does it with their world-wide humongous supercomputer! Also, you aren't doing just one Google Maps calculation. You are doing hundreds of them. Imagine searching for directions from your house to the house of everyone on OKC simultaneously and in milliseconds.
Playing devil's advocate here, couldn't OKC do something to calculate a list of town names within X number of miles within a new user's listed home, when that user signs up? That info can be cached, so that calculation only needs to happen once.
User signs up, says they're in town A and they're okay dating people within 20 miles. OKC determines that towns X, Y, and Z are within 20 miles of A, and caches those towns. Now, when User looks for matches, OKC finds results from A, X, Y, and Z - no need to calculate, it's just a WHERE statement.
As the crow flies is basic a^2 + b^2 = c^2. And I think they only know city and zip code, so they can likely index all of the relevant distances between those points and not even do the computation.
Just geocoding the specific addresses requires way more information, especially in rural nowhere.
Once you've got that information, you've got to first find the optimal route (or a near optimal route). Realistically you also want to estimate by time as opposed to raw driving distance.
Even though you could expect to do all these calculations, you're already a couple orders of magnitude more complex than the shitty original look up table.
And yes, you can obviously always build reference tables for approximate information. Certainly conceivable, but you're trading disk space and ram for processing (which you should do, but the cost is still so much higher than the current crap method).
Straight line distance is a fine metric for this use case. If you set it to ~50 miles, you're already willing to date someone pretty far away. If a distance of ~10 miles doesn't find anyone, you're already casting a pretty wide net.
It's an incredibly difficult computational problem. It would have to calculate the driving distance between you and every other single OKC user and then query that distance.
Not true, because you would only care about especially accurate driving distances within a relatively short radius (say 100 miles). Just find the straight-line distances first, and then get more accurate results only for the closer people.
It shouldn't be that hard. All I use is google maps with a zip code to zip code travel search. While someone may be under 50 miles as the crow flies they are not 50 miles as the car drives.
Not that hard. Google does it with their world-wide humongous supercomputer! Also, you aren't doing just one Google Maps calculation. You are doing hundreds of them. Imagine searching for directions from your house to the house of everyone on OKC simultaneously and in milliseconds.
Do you think OKC does a new query every time or does it cache the results?
When I do send coding work a few years back for a social site they used a database that assigned a log/lat coordinate for every zip code in the area and used that to judge distance. While might require more storage space a db could be created with an a end, z end and drive distance.
Dating's easy if you live somewhere where there are people. ;^)
You also have to have something that gives you a legitimate reason to interact with people. Just having lots of people doesn't mean you'll meet people. I can walk up and down the quarter mile at RIT every day if I wanted and be surrounded by hundreds of people and meet no one. I need to have a class with people I haven't met who are open to interaction, or go to a club with people that aren't crazy (and open to interaction).
Can of worms indeed.
Rym wasn't a podcaster in College at RIT, what he and we all did was get involved. Wanna meet people get freaking involved. Jesus. You've been on these forums long enough to know that. You can't meet people unless you find reasons to interact with people....
I met my first RIT girlfriend on the second day of classes freshman year by going to the gaming club (RWAG). We were flirty by the next week (the first D&D session involving the FRC ever), and the rest was history.
I may have had... relations... with a girl prior to that at the orientation weekend by, gasp, playing ping pong and hanging out.
I went to more club meetings and events my first weeks of RIT than I did classes. That's the secret to meeting people.
Straight line distance is a fine metric for this use case.
I pretty much agree. I just don't think Scott's evaluation of the problem is accurate.
I think it is pretty spot on if you take into account the cost/time to implement anything better, coupled with the fact that it would have no impact on their overall usability.
You can also say hi to people outside of official get-togethers. Someone sitting in the library, hello! Someone sitting next to you in a computer lab, hello! Someone also buying a beverage at the corner store, hello!
I also went to clubs. Mostly only met crazy people. Also met my first RIT girlfriend. That ended...How it ended. Continuing to go to clubs did not result in me meeting more people.
Comments
If I live on the CT coast I can see Long Island. Doesn't mean it is only a 10 mile drive!
Can of worms indeed.
You also have to consider non-driving. Imagine a user in Manhattan. Sure, someone in Long Island is technically very geographically nearby, but what if they are beyond the reach of the subway, or not near a LIRR station? That can be up to an hour difference in travel time, but just a few miles difference as the crow flies.
Also, OKC doesn't have actual addresses in their database, nor would they add them. All they know is a very general location such as your town's or neighborhood's name.
User signs up, says they're in town A and they're okay dating people within 20 miles. OKC determines that towns X, Y, and Z are within 20 miles of A, and caches those towns. Now, when User looks for matches, OKC finds results from A, X, Y, and Z - no need to calculate, it's just a WHERE statement.
Just geocoding the specific addresses requires way more information, especially in rural nowhere.
Once you've got that information, you've got to first find the optimal route (or a near optimal route). Realistically you also want to estimate by time as opposed to raw driving distance.
Even though you could expect to do all these calculations, you're already a couple orders of magnitude more complex than the shitty original look up table.
And yes, you can obviously always build reference tables for approximate information. Certainly conceivable, but you're trading disk space and ram for processing (which you should do, but the cost is still so much higher than the current crap method).
Not to say they shouldn't do "something."
When I do send coding work a few years back for a social site they used a database that assigned a log/lat coordinate for every zip code in the area and used that to judge distance. While might require more storage space a db could be created with an a end, z end and drive distance.
/game on
I may have had... relations... with a girl prior to that at the orientation weekend by, gasp, playing ping pong and hanging out.
I went to more club meetings and events my first weeks of RIT than I did classes. That's the secret to meeting people.
Continuing to go to clubs did not result in me meeting more people.
Maybe next year when I'm back in classes. But probably not because I am not even close to fit.