[YSQL] Single row update overwrites unmodified columns with default values to null

[ndeodhar]

If we have a single row UPDATE statement that only writes few columns, we end up overwriting other columns to NULL.
This also ends up breaking constraints like not null which may be set on those columns.

CREATE TABLE sbtest1 (
    id int primary key, 
    k int not null default 0,
    c text default '',
    pad text default '');

yugabyte=# insert into sbtest1 values(1000001, 10, 'foo','bar');
INSERT 0 1
yugabyte=# select * from sbtest1 where id=1000001;
   id    | k  |                                                            c                                                             |                             pad                              
---------+----+--------------------------------------------------------------------------------------------------------------------------+--------------------------------------------------------------
 1000001 | 10 | foo                                                                                                                      | bar                                                         
(1 row)

yugabyte=# update sbtest1 set c='hello' where id=1000001;
UPDATE 1
yugabyte=# select * from sbtest1 where id=1000001;
   id    | k |                                                            c                                                             | pad 
---------+---+--------------------------------------------------------------------------------------------------------------------------+-----
 1000001 |   | hello                                                                                                                    | 
(1 row)

This does not happen while updating multiple rows:

yugabyte=# insert into sbtest1 values(1000002, 10, 'foo','bar'),(1000003, 10, 'foo','bar');
INSERT 0 2
yugabyte=# update sbtest1 set c='hello' where id in (1000002,1000003);
UPDATE 2
yugabyte=# select * from sbtest1 where id in(1000002,1000003);
   id    | k  |                                                            c                                                             |                             pad                              
---------+----+--------------------------------------------------------------------------------------------------------------------------+--------------------------------------------------------------
 1000002 | 10 | hello                                                                                                                    | bar                                                         
 1000003 | 10 | hello                                                                                                                    | bar                                                         
(2 rows)

Update:
This happens for columns with default values but not for columns without any default values.

yugabyte=# create table foo(a int primary key, b int default 0, c int default 0, d int);                                                                                  CREATE TABLE
yugabyte=# insert into foo values(1,1,1,1);
INSERT 0 1
yugabyte=# update foo set b=10 where a=1;
UPDATE 1
yugabyte=# select * from foo;
 a | b  | c | d 
---+----+---+---
 1 | 10 |   | 1
(1 row)

[m-iancu]

Fixed by d9a860f.