Ask HN: Database with country names down to the city, etc. | Hacker News Reader

Ask HN: Database with country names down to the city, etc.

9 pointswenbert12 comments

Hello HN,

I remember reading these in HN a few months ago. Apparently they were free. They had coordinates, country names, cities, towns, etc.

Can anyone point me to a reliable site with a good database for such?

Thanks!

-Wenbert

View on HN

hey Wenbert,

I've got one, let me dig it up for you, just a moment.

edit:

Ok, ready for download, it's a mysql dump so you can simply read it back in to any mysql server:

http://autotagger.com/geoinfo.dump.gz

41 Megabytes, 2.67 million towns & cities, fields are country, region,name,accented name,latitude,longitude,verified,current,prefered and population.

The 'verified', 'current' and 'prefered' fields are all internal to an application I wrote so you can lose those safely.

I've also added tables with countries, country aliases and regions.

edit2: holy guacamole, easy guys, that server is only on a 100 Mbit uplink.

I figured there'd be one download...

Thanks a ton, jacquesm.

This would've been very useful for a past project, and while I don't necessarily need it at the moment, I suspect it'll come in handy sometime in the future.

Thanks! What about for updates? How do you regularly update your data?

LOL. It downloaded in about a minute. Nice server!

Again, thank you so much!

This is a hybrid of a public dataset and a lot of manual polishing to get rid of duplicates (no doubt there are a still quite a few of those, especially in Asia).

I don't actively maintain the dataset, but if you want I can create a small website around it so people can update it.

The reason I built this was a project called 'autotagger', the idea behind it is to identify people, places and subjects in texts automatically.

It's one of my pet projects but I don't have enough funds to continue to work on it full time without a way to monetize it, so it's been sitting there rusting for the last year or so.

I think I have most of the datastructures and algorithms worked out, it needs a massive rewrite to get it production ready and fast enough.

The fields most subject to change are the population and the names, cities/towns don't get created too often but merging smaller towns does happen.

They rarely move around, so I figure the longitude/latitude should be good for a while ;)

Hehe. I am from Asia (Philippines), but in Norway right now for a short training. Things are a little bit unpredictable right now (schedule, etc.) for me but I find your project very interesting. It would be good if I can find something else to do since I have a few hours at night after the training.

Would be good if you can give me some more info about it.

Autotagger is one of the more interesting things I've worked on in the last couple of years, it takes a pretty weird approach to the problem but it worked surprisingly well.

When I have some time I'll do a write-up on it and post it.

The basic idea behind it is 'google backwards', if you can figure out how it works from those two words drop me a line ;)

I have no idea what you mean by "google backwards".

Anyway, thanks! I keep an eye on this project.

Not exactly to your requirements but related: we use the MaxMind GeoLite city database to correlate IPs with lat/lon and city/country names. It's free (the GeoLite version; the pay version has a little more detail but doesn't cost very much) and simple to update and use.

http://www.maxmind.com/app/geolitecity

The one we used for our analytics product came from here:

http://www.geonames.org/export/

Thanks again! I will compare this with jacquesm's file.

http://www.geonames.org/ (download area at http://download.geonames.org/export/dump/)

Used as the placename source for http://www.dopplr.com among others.

HTH,

Brendan.

I have found the link in HN: http://news.ycombinator.com/item?id=530086

Reply on news.ycombinator.com