Is it possible to rename columns in a table?

Hi,

I'm hoping there's a simple way to rename columns in a table. I don't need anything complicated, I'd just like to rename, for example, the 2nd, 7th, and 16th columns of a table - inside a function.

Thanks!

Answer 1

[per isakson]

Yes! Doc says:

Example

    %%Create a table
    T = table(['M';'M';'F';'F';'F'],[38;43;38;40;49],...
        [71;69;64;67;64],[176;163;131;133;119])
    T = 

        Var1    Var2    Var3    Var4
        ____    ____    ____    ____

        M       38      71      176 
        M       43      69      163 
        F       38      64      131 
        F       40      67      133 
        F       49      64      119 
    %%Modify variable names
    T.Properties.VariableNames = {'Gender' 'Age' 'Height' 'Weight'}
    T = 

        Gender    Age    Height    Weight
        ______    ___    ______    ______

        M         38     71        176   
        M         43     69        163   
        F         38     64        131   
        F         40     67        133   
        F         49     64        119

[Chelsea]

Thanks a lot!!

[Cbhihe]

Actually you are renaming variables here, not column (headers) as requested in OP. Is there a way to just rename headers and not the actual variable names ? (At first I mistakenly posted this comment below on one of the non-accepted answers.)

[Peter Perkins]

If I understand your terms correctly, there is no difference. The text that is displayed above the data are the variable names.

[Walter Roberson]

Expanding slightly: there is a VariableDescriptions property, normally empty. However, the only place it is used is for summary(). For all other displays of a table, it is the variable name that is displayed.

[Silver]

Hello @per isakson ! I have an issue in here ! actually I have these columns names to generate to the table :

      T.Properties.VariableNames = {'Date_Time' 'Latitude' 'Longitude' 'Course' 'Speed' 'Temp_SBE45' 'Cond_SBE45' 'SoundVel_SBE45' 'full' 'pH_SeaFET' 'Saturation' 'Salinity_SBE45' 'Temperature_Optode' 'flow_pCO2' 'pH_Meinsberg' 'Temp_in_SBE38' 'Temp_Meinsberg' 'Oxygen' 'pCO2' 'pressure' 'flow_in' 'flow_main' 'flow_pH' 'Turbidity' 'halffull' 'Chl_a' 'Variance' 'pH_Meinsberg' 'Variance' 'Temp_Meinsberg' 'Variance' 'pH_SeaFET' 'Variance' 'pCO2' 'Variance' 'pressure' 'Variance' 'flow_in' 'Variance' 'flow_main' 'Variance' 'flow_pH' 'Variance' 'flow_pCO2' 'Variance' 'halffull' 'Variance' 'full' 'Variance'}

But I always get an error message :

Duplicate variable name: 'Temperature_Optode'.

Any Ideas ?

[per isakson]

@Silver, this is a new question rather than a comment. Posting it as such increases the chances to get an answer.

There are many duplicates in that row. Obviously, duplicates are not allowed (although it's not clearly stated in the documentation(?)). However, 'Temperature_Optode' is not a duplicate.

Answer 2

[Peter Perkins]

The following are also possible:

Rename only some variables (note the parentheses on the left-hand side and the cell array of strings on the right-hand side) ...

>> T.Properties.VariableNames([1 3]) = {'Gender' 'Height'}
T = 
    Gender    Var2    Height    Var4
    ______    ____    ______    ____
    M         38      71        176 
    M         43      69        163 
    F         38      64        131 
    F         40      67        133 
    F         49      64        119

... or rename only one variable (note the braces on the LHS instead of parentheses, and the raw string on the RHS) ...

>> T.Properties.VariableNames{2} = 'Age'
  T = 
      Gender    Age    Height    Var4
      ______    ___    ______    ____
      M         38     71        176 
      M         43     69        163 
      F         38     64        131 
      F         40     67        133 
      F         49     64        119

... or rename a variable when you only know its (old) name.

>> T.Properties.VariableNames{'Var4'} = 'Weight'
T = 
    Gender    Age    Height    Weight
    ______    ___    ______    ______
    M         38     71        176   
    M         43     69        163   
    F         38     64        131   
    F         40     67        133   
    F         49     64        119

[Chelsea]

Thank you :)

[Cbhihe]

Actually you are renaming variables here, not column (headers) as requested in OP. Is there a way to just rename headers and not the actual variable names ?

Answer 3

[Oleg Komarov]

My tableutils() has renameVarNames() that does exactly that with added flexibility.

The description is usually updated on https://github.com/okomarov/tableutils

Answer 4

[Martin Patz]

I agree with Cbhihe's comment above, the proposed solutions do not directly answer the OP's question. I face a case where the differentiation between column header variable is indeed important. For instance

v1 = [1:3]';
v2 = [2:4]';
m = [v1, v2];
t1 = table(v1, v2)
t2 = table(m)

gives two different results:

t1 = 
  v1    v2
  __    __
  1     2 
  2     3 
  3     4 

t2 = 
    m   
  ______
  1    2
  2    3
  3    4

with previously posted answers it is not possible to rename the columns of t2 by specifying 'VariableNames', {'column_1', 'column_2'} as additional arguments to the table command. However you can split up a matrix into separate columns, which are then configureable, using the array2table command.

Example:

t3 = array2table(m, 'VariableNames', {'column_1', 'column_2'})

    column_1    column_2
    ________    ________
    1           2       
    2           3       
    3           4

[Walter Roberson]

I disagree. t2 has only one column, which happens to be a 3 x 2 array.

... Any other interpretation would require that you be able to give separate column headers for each character in variable-length strings.

[Peter Perkins]

This is the very reason why we try to refer to the vertical things in a table as "variables", not as "columns".

Martin, m in your example is an Nx2 double. Double arrays have no way to "name" their columns, and you are correct that just putting an Nx2 double into a table does not add that capability. You might do one of two things:

1) As you say, it's pretty easy to split the Nx2 into two Nx1 column vectors. But presumably, there's some reason why you'd want m as an Nx2 rather than splitting it.

2) You could make m itself a table with two variables, and out that in a table. That would be kind of pointless if the outer table only had m as its one variable, but in cases wher you have a table with several variables, one of which has multiple columns, it can be useful. Currently, the display for a "table in a table" is not very helpful, but all of the subscripting and so on works as you'd expect. On the other hand, a table is not a double, so while this strategy keeps those two columns "together" in some sense, you might want to work with m as a double.

Answer 5

[Mav Sam]

terrible!