Imagine you have the table create table my_table (id serial primary key); and these two migrations on separate branches:
alter table my_table add column column_1 int;
alter table my_table add column column_a text;
At first you might think these don't conflict, but actually they do, the end result is dependent on order:
create table my_table (
id serial primary key,
column_1 int,
column_a text
);
versus:
create table my_table (
id serial primary key,
column_a text,
column_1 int
);
Now if you try and represent this table value as a tuple - (1, 2, 'three')::my_table - then this will work against one DB but fail against the other. Order of columns is significant.
So it's best to ensure that your migrations are completely linear.
Imagine you have the table
create table my_table (id serial primary key);
and these two migrations on separate branches:At first you might think these don't conflict, but actually they do, the end result is dependent on order:
versus:
Now if you try and represent this table value as a tuple -
(1, 2, 'three')::my_table
- then this will work against one DB but fail against the other. Order of columns is significant.So it's best to ensure that your migrations are completely linear.