Question

For the following example (from mysql source code), it uses Field **field. I am in trouble to think it as a 2 dimension array of Field.

typedef struct st_table_share  
{  
  .......  
  Field **field;           
  ....  
}  

for (Field **field=table->field ; *field ; field++)  
{  
    ...  

Can I think it in this way, a table contains many rows and a row contains multiple columns/fields. So *field means a row, and **field mean a table?

If that’s true, for the following code

for (Field **field=table->field ; *field ; field++)

it will exit when *field is null, so how could *field be null if *field is a row. Or can I say if a row has 5 columns, and field is the first column, then field+4 is the last column field, and field+5 is the end of the column which means null, so that the for loop will exit?

Answer 1

So *field means a row, and **field mean a table?

In declaration syntax, yes.

how could *field be null if *field is a row

If the authors of the code take care to preserve the invariant that field is terminated by a null pointer, as is done with argv. I.e., a table of n rows is an array of n+1 pointers, the last of which must always be null. This is a common C convention that obviates the need for an explicit count.