Searching values in a dataset

Hi

I need to apply some filters to a dataset, or just look for some specific quantity. However, I receive an error that the eq operator is not defined for datasets. I know I can convert a particular column to, e.g., double and then use the eq operator for double, but if the dataset is big, I am concerned about performance overheads due to constant conversions.

So

• Is there any way to apply a filter to a dataset without converting it first to other type (e.g., double)?
• Is there any way to look for values in specific columns of a dataset without converting such column first to other type (e.g., double)?

Thanks in advance

Answer 1

[Peter Perkins]

Quite often, this is the kind of thing you'd do to pick out rows of a dataset array:

>> load hospital
>> hospital(hospital.Age>45 & hospital.Sex=='Male',{'LastName' 'Age' 'Smoker'})
ans = 
               LastName          Age    Smoker
    KOQ-996    'MARTIN'          48     true  
    XBA-581    'ROBINSON'        50     false 
    YYV-570    'SCOTT'           47     false 
    FLJ-908    'STEWART'         49     true  
    DAU-529    'REED'            50     true  
    FCD-425    'GONZALES'        48     false 
    MEZ-469    'GRIFFIN'         49     false 
    ZZB-405    'HAYES'           48     false

There's not really any need to explicitly "convert" Age to double, just by subscripting using dot, you get back the Age variable itself. Similarly for the Sex variable. Performance wise, this is actually (under the hood) a shared copy of something that already exists, so it should be fast.

[Mark]

Thanks, that works when you know the field name beforehand, but what about the case when you are given the field name as a string? For example:

>> load hospital;

>> strField='Sex';strValue='Male'; hospital(hospital(:, {strField})==strValue,{'LastName' 'Age' 'Smoker'})

??? Undefined function or method 'eq' for input arguments of type 'dataset'.

That gives me the above error. How can I reference a particular field if I'm only given the field name in a string?

Sorry if my initial question was not clear.

[Mark]

I found the answer myself, but with your valuable help of course:

>> strField='Sex';

>> strValue='Male';

>> hospital( hospital.(strField) ==strValue,{'LastName' 'Age' 'Smoker'})

I didn't know about the syntax .(field name)

[Peter Perkins]

Yes, just as with a structure array, you can use that "dot parentheses" syntax if the name is stored in a string variable.

Your initial guess, which involved

hospital(:, {strField})==strValue

illustrates a point that is sometimes hard to understand at first: a dataset array (much like a cell array) is a container, and subscripting using parentheses preserves that "containerness", and so hospital(:, {strField}) is dataset array that only contains one variable, but it is not the same thing as that variable. Much the same as with a scalar structure, it's the dot subscripting that gets you the "contents".

One more minor point: hospital(:, strField) (without the braces around strField) would have been fine, since you've only got one variable name.