Get countries from IP addresses

[Samuel Mehr]

Hi, I've got about 1000 IP addresses that I'd like to get country information for. R has a package for grabbing info from the freegeoip API. Does Stata have anything similar? I've played around a bit with the packages 'googleplaces' and 'geocodehere', but these require finding an actual place (like an address) to get the relevant info, whereas something like freegeoip outputs a bunch of really clean info (country, region, postcode, time zone etc).

Is anything comparable available in Stata? It's a bit of a pain to need to go back and forth to R for this project so I'm hoping there is a solution.

thanks!

[Sergio Correia]

You can wrap this code in a for loop:

insheet using http://ip-api.com/csv/208.80.152.201, clear nonames
list

You might also want to add a sleep 500 command in every loop, or you'll end up banned:

Our system will automatically ban any IP addresses doing over 150 requests per minute. To unban your IP click here.

You are free to use ip-api.com for non-commercial use. We do not allow commercial use without prior approval.

[Samuel Mehr]

Oh cool, that sounds like it'll work great. It also looks like other geolocator APIs have this functionality too (like freegeoip.net).

Naïve question, but how would you recommend writing the forloop? Is there a simple way to dump all observations of a single variable into a list, so that I can loop over each IP that I need to get info for? I haven't had to do anything like this before.

And I'm guessing that it wouldn't make sense to build the dataset in real time but rather to save each csv as a one-line dta and then append them all together in another loop?

[Samuel Mehr]

Whoops, two searches led me to...

levelsof ip, local(list_of_ips)
foreach ip of local `list_of_ips' {
      insheet using http://ip-api.com/csv/`ip', clear nonames
      save `ip'
}

[Samuel Mehr]

Hmm, this should work in principle, but I think the formatting of IP addresses is causing Stata to throw an r(198) error [invalid name]. Can that be avoided somehow?

[Samuel Mehr]

I think I've tricked Stata into treating IPs as strings....

loc N = _N
forval i = 1/`N' {            // make local vars for each IP
    loc ip`i' = ip[`i']
    }
forval j = 1/`N' {            // get CSV for each IP, save as dta
    insheet using "http://freegeoip.net/csv/`ip`j''", clear nonames
    save `ip`j'', replace
    }
forval k = 1/`N' {            // build dataset
    append using `ip`k''
    rm `ip`k''
    }

Thanks for the tip, Sergio. If you have a sec, could you tell me if there is a more elegant way to do this?

[Sergio Correia]

Sure. I wrapped it up in a program and put it on Github here:

https://github.com/sergiocorreia/geocode_ip

How to use it:

geocode_ip ip, clear

Let me know if it's useful and I might add it to SSC.

[Samuel Mehr]

Thanks! What I posted above is working for now and I'm traveling for the next few days, but I will take a look at the cleaner version soon. My guess is that people would use this, though, since there currently isn't a clean, one-command solution to this straightforward task, as there is in R.

[Carlos Lagorio]

Hello Sergio,

I´ve tried using the geocode_ip command but notice that the freegoip.net website is now redirected to a new ipstack.com address. I notice from the latest ado file that the command still queries freegoip. Is the command no longer working?

[Sergio Correia]

Hello Sergio,

I´ve tried using the geocode_ip command but notice that the freegoip.net website is now redirected to a new ipstack.com address. I notice from the latest ado file that the command still queries freegoip. Is the command no longer working?

I haven't used it lately, and you are right, they now require an API key, which would require some changes in the code.

[Nicholas Winter]

I recently uploaded to SSC a new program, checkipaddresses, . It queries a different API (iphub.info, which is mostly focused on detecting VPNs and proxies. But it might be useful as-is or as a basis for modification.