## Stuart’s MATLAB VideosWatch and Learn

Posted by Doug Hull,

I am blogging a little early this week because of the holiday. What holiday would that be? Square Root Day– 03/03/09. What makes this a particularly special event is that today’s video is number 144 (12^2) and I recorded it just minutes ago. Posted at 1:44 local time too! This kind of coincidence can not be ignored! When I first thought of this video, I figured it would take three videos to solve the problem. When I actually sat down to read data from the web and interactively filter it, I was happy to see it could be done so quickly and easily. Basically, the procedure is to read data from Many Eyes, then plot that Census data and filter it for city size. We will use URLREAD and TEXTSCAN to bring the data in and then interactive cell mode and plot to filter and visualize it.

Daniel Armyr replied on : 1 of 27
You are a big nerd. But the video was nice. Few lines of code for a large wow-factor.
dhull replied on : 2 of 27
Daniel, Thanks for the kind words about the video. Doug PS. I prefer the term Geek! :)
Tim Davis replied on : 3 of 27
URLREAD is great. It's been a great help. I have a MATLAB interface to the UF Sparse Matrix Collection called UFget, which downloads matrices from the collection into the MATLAB workspace. I used to have a Java component to it, but now I can do it all in M, with URLREAD, and my code is now much simpler (and more robust).
Jim replied on : 4 of 27
Can you post the code in a text file?
dhull replied on : 5 of 27
Russell replied on : 6 of 27
This will be really handy as soon as Wolfram|Alpha is up and running!
dhull replied on : 7 of 27
@Russell, I am not sure what you mean, how would they work together? Doug
dhull replied on : 8 of 27
Looks like this was answered in the newsgroup.
Magnus replied on : 9 of 27
Hi Doug, I tried to use you method to import data from the web, but it did not work. I did post a link on matlabcentral, do you have any suggestion how I should read the data from the web and put it into a matrix? -Magnus
Need Help replied on : 10 of 27
Hi Doug, I have a question. I want to know a way to convert the string you obtain from using urlread() into a matrix. In the case scenario described above, the data is in columns already but I was wondering if there is a function which converts it into a string without the requirement of the data being in columns already. Please let me know. Thanks
dhull replied on : 11 of 27
@Need Help The command does return the data as one big string. When you say: "I was wondering if there is a function which converts it into a string" What do you mean? What would a typical return from URLREAD() look like so we know what you are getting and what would the desired output look like? Doug
paul replied on : 12 of 27
Thanks Doug, great tutorial. I'm trying to do something similar, which I know is possible, but just a bit beyond my matlab ability. I am trying to extract historical wind data from a website, and while it is comma separated, it is a mixture of words, numbers, and annotations. I can get the data into matlab, but I can't figure out how to (1) save the data as a csv file for archiving, (2) extract specific parts of the block (like hour), and (3) turn the data into variables (double?) on which I can do math. I have found several possible solutions but keep getting stuck on one part or another. I think I could do it in a way that would require a ton of loops and if/then statements, but since I want to mine data from many websites, this would likely get slow. Any help would be much appreciated. THANK YOU!! Here's my start: block=urlread('http://www.wunderground.com/history/airport/KSVC/2010/4/1/DailyHistory.html?req_city=NA&req_state=NA&req_statename=NA&format=1'); data=textscan(block,'%s %s %s %s %s %s %s %s %s %s %s %s','delimiter',','); urwrite('http://www.wunderground.com/history/airport/KSVC/2010/4/1/DailyHistory.html?req_city=NA&req_state=NA&req_statename=NA&format=1',data.csv'); data = csvread('data.csv'); data2=textscan(block,'%s %s','delimiter',':'); clear block points=length(data{1}); dataTime=data{1}(2:points-1); dataTemperature=data{2}(2:points-1); dataHumidity=data{3}(2:points-1); dataWindDirection=data{7}(2:points-1); dataWindSpeedAvg=data{8}(2:points-1); dataWindSpeedGust=data{9}(2:points-1); dataHour=data2{1}(2:points-1);
paul replied on : 13 of 27
To those asking about converting the string data to matrix data, str2double worked for me. Somewhere somebody recommends to save the string as a csv and then read the csv back into Matlab, but this seems very roundabout to me.
dhull replied on : 14 of 27
@Paul, I like the following:

urlwrite('http://www.wunderground.com/history/airport/KSVC/2010/4/1/DailyHistory.html?req_city=NA&req_state=NA&req_statename=NA&format=1','data.csv');


You now have num and txt. These can be parsed pretty easily, especially if the format of all of these are always the same. I actually like pulling the data locally into a file and then reading it in. Something about having the file close by, even if in a temporary file that you toss later feels right. -Doug
Thisara replied on : 15 of 27
Hi Doug, Thanks for the nice piece of work. I am trying to download some netcdf data from web. I tried to use urlread, but couldn't success.File can not be scanned as you did. But if I can download the netcdf file, I can read it through Matlab in my computer. Is there any other way to download the netcdf file through Matlab ? Cheers Thisara
dhull replied on : 16 of 27
Thisara, Can you read the files if they were local? Can you use the FTP command to bring them local? Doug
Ritesh replied on : 17 of 27
Jussi replied on : 18 of 27
I have the exact same problem as Ritesh. Does anyone have an implementation for this?
Doug replied on : 19 of 27
I would look into MATLAB's FTP command. I use it all the time to automatically upload the video files you see on this blog.
Muneer replied on : 20 of 27
Hey Doug, I've hit the same roadblock as Jussi and Ritesh. Any idea on how to read data that's password protected? My url has SQL commands in the text so I was going to try and read just the login page and then write in commands for which data to select like what's pasted below (it's written in R). However, your method seems way simpler. I have to use a token as well for security with this way.
Doug replied on : 21 of 27
You will need to make sure you can put the authentication into the URL to make this work.
John replied on : 22 of 27
I have tried your code. The textscan only read first two cities not all cities. What might be the problem?
João replied on : 23 of 27
Hi Doug, I'm having trouble loading the data. Using the same code as you, the textscan only loads the first and the second line of the web txt. Any suggestion? Regards, João Miguel Silva
Alex replied on : 24 of 27
The many Eye file is no longer available. Any idea where its new location is? Thanks
Stuart McGarrity replied on : 25 of 27
Looks like it has shut down: http://www.computerworld.com/article/2930326/data-analytics/ibm-to-shutter-dataviz-pioneer-many-eyes.html
rodeopepper replied on : 26 of 27
Doug, I tried your example for pulling data from wunderground but it doesn't seem to work for me even when copied directly over. I get num as empty and txt is gibberish ('JMÉ' 'JM.Q(HLOµI*²a®ŒÔ¢T…Ìb…¼ü’ŒÌ¼t…’|...') Any help you might be able to provide would be greatly appreciated as I've been pulling my hair out on this for hours. This is exactly what I have used which I copied from above. urlwrite('http://www.wunderground.com/history/airport/KSVC/2010/4/1/DailyHistory.html?req_city=NA&req_state=NA&req_statename=NA&format=1','data.csv'); [num, txt] = xlsread('data.csv');
Stuart McGarrity replied on : 27 of 27
Hi rodeopepper, I think you copied it incorrectly. Its >> block=urlread('http://www.wunderground.com/history/airport/KSVC/2010/4/1/DailyHistory.html?req_city=NA&req_state=NA&req_statename=NA&format=1'); >> data=textscan(block,'%s %s %s %s %s %s %s %s %s %s %s %s','delimiter',',');

This site uses Akismet to reduce spam. Learn how your comment data is processed.