4.1.13 Where statement

A where() combines the definition and application of a mask all in one go and can lead to succinct code. The full syntax of a where() statement is as follows:

     // Single assign (the 'elsewhere' block is optional)
     where(mask)
        var1=expr1;
     elsewhere
        var1=expr2;
     
     // Multiple assigns
     where(mask){
         var1=expr1;
         var2=expr2;
         ...
     }elsewhere{
         var1=expr3
         var2=expr4
         var3=expr5;
         ...
     }

• The only expression allowed in the predicate of a where is assign, i.e., 'var=expr'. This assign differs from a regular ncap2 assign. The LHS var must already exist in Input or Output. The RHS expression must evaluate to a scalar or a variable/attribute of the same size as the LHS variable.
• Consider when both the LHS and RHS are variables: For every element where mask condition is True, the corresponding LHS variable element is re-assigned to its partner element on the RHS. In the elsewhere part the mask is logically inverted and the assign process proceeds as before.
• If the mask dimensions are a subset of the LHS variable's dimensions, then it is made to conform; if it cannot be made to conform then script execution halts.
• Missing values in the mask evaluate to False in the where code/block statement and to True in the elsewhere block/statement. LHS variable elements set to missing value are not re-assigned. For these reasons, do not explicitly reference missing values in the masking condition, e.g., where(foo=foo.get_missing()) foo=1; will not work as expected.

Example: Consider the variables float lon_2D_rct(lat,lon); and float var_msk(lat,lon);. Suppose we wish to multiply by two the elements for which var_msk equals 1:

     where(var_msk==1) lon_2D_rct=2*lon_2D_rct;

Suppose that we have the variable int RDM(time) and that we want to set its values less than 8 or greater than 80 to 0:

     where(RDM < 8 || RDM > 80) RDM=0;

Consider irregularly gridded data, described using rank 2 coordinates: double lat(south_north,east_west), double lon(south_north,east_west), double temperature(south_north,east_west). To find the average temperature in a region [lat_min,lat_max] and [lon_min,lon_max]:

     temperature_msk[$south_north,$east_west]=0.0;
     where(lat >= lat_min && lat <= lat_max) && (lon >= lon_min && lon <= lon_max)
       temperature_msk=temperature;
     elsewhere
       temperature_msk=temperature@_FillValue;
     
     temp_avg=temperature_msk.avg();
     temp_max=temperature.max();