How to read year from timeseries?

Hi there,

I have a script that uses "eval" to define a range of yearly environmental values over a large number of sites for a large number of years. I know "eval" is not suggested, but with the number of years, sites, and variables I am dealing with it makes things a whole lot easier.

Anyways, the following snippet of code works well but I would like to automate it even more. In the following, I first load data then create a structure (pars_yr) that defines the variable years I am looking at.

% Download .DAT data matrix, define first column as datetime timeseries

%third column is temp. "Date_all_site" is the entire datetime column for that site.
%  Data is recorded every 30min throughout the year

pars_site_yr = {'2014','2015','2016','2017','2018','2019'};

site_yr_start = min(Date_all_site):calyears(1):max(Date_all_site);
site_yrs = years(site_yr_start)

for i = 1:length(pars_site_yr)
    eval(['temp_site_',pars_site_yr{i},' = site_',pars_yr{i},'(3,:);'])
end

% more of my code

The problem is, every time I add more data I need to manually change "pars_site_yr". This isn't a huge problem considering I will only need to do it once a year, but I think it would be cool if the code just automatically "read" the new datetime years (especially considering I already defined the years for each site in site_yrs).

"site_yrs" has all the years based off the input "Date_all_site", but it is not in a useable "structure" form like pars_site_yr is. Each site has a different range of years that data was recorded.

I'm pretty new to matlab so hopefully this is easy. Thanks a ton for all your help!

-Devon

[Walter Roberson]

Use a struct with dynamic field names instead of eval()

[Devon Fisher-Chavez]

Could you elaborate or send a link to an article about this?

I quite literally have hundreds of different site, probe, variable, and year combinations that I do calculations to. I want to do the "proper" method, but its hard to justify when doing so would require many hours of code writing and several thousand lines.

Thanks for your response!

[Walter Roberson]

http://www.mathworks.com/matlabcentral/answers/304528-tutorial-why-variables-should-not-be-named-dynamically-eval

for i = 1:length(pars_site_yr)
    fn = ['y', pars_site_yr{i}];
   temp_site.(fn) = site.(fn)(3,:);
end

This would expect site.y2014 site.y2015 and so on, and would produce temp_site.y2014 temp_site.y2015 and so on.

But often better is cell arrays (if the data is different lengths) or multidimensional numeric arrays when practical.

for i = 1:length(pars_site_yr)
   fn = ['y', pars_site_yr{i}];
   temp_site{i} = site.(fn)(3,:);
end

now you can cellfun over temp_site, and if you want to locate a particular year then ismember() into pars_site_yr and take the second output.

During the time that you are doing the processing, using variable names or struct field names that contain metadata is mostly for presentation purposes. Sometimes it can make sense to use metadata in struct field names inside .mat files but that is the last processing step and can often be handled with cell2struct()

[Mohammad Sami]

Assuming a datetime variable. you can use year function

yr = year(datetime_var);
yr = unique(yr); % this will give you all the years in your data.
pars_site_yr = arrayfun(@num2str,yr,'UniformOutput',false);

[Devon Fisher-Chavez]

@Mohammad Sami, that works with my existing program, thanks!

@Walter Roberson, that also works, but will require I change quite a bit of my existing code. Because this is the "proper" way to do things, I will try to convert!

Thank you both for your responses!