importccl: IMPORT PGDUMP could fail when the targeted columns are ordered differently than when they are created

[Anzoteh96]

When running an IMPORT from a PGDUMP file, it seems like it expects the ordering of the targeted columns to match the declaration of the columns. In particular, it also requires all columns to be targeted.

To illustrate this, let's consider the following pgdump file called myfile:

A,1
B,2

Take an example of the following command:

CREATE TABLE t (a INT, b STRING)
IMPORT INTO t (b, a) PGDUMP DATA ("myfile")

It gives an error row 1: cannot parse "A" as int when we expect it to return the table

a   b
---+---
1 | A
2 | B

Take another example independent of the example above, but still using "myfile":

CREATE TABLE t (a STRING, b INT, c INT)
IMPORT INTO t (a, b) PGDUMP DATA ("myfile")

This gives error expected 3 values, got 2 ["A", "1"], when we expect the following:

a   b   c
---+---+---
A | 1 | NULL
B | 2 | NULL

This will therefore cause problems when we perform import on a table from PGDUMP file when there's non-targeted columns with default expressions.