Re: Declarative partitioning

IIUC, this seems like a combination of 2 and 3 above:

So, we have the following list partitions (as read from the catalog)

Table 1: p1 {'b', 'f'}, p2 {'c', 'd'}, p3 {'a', 'e'}
Table 2: p1 {'c', 'd'}, p2 {'a', 'e'}, p3 {'b', 'f'}

By applying the method of 3:

Table 1: p3 {'a', 'e'}, p2 {'b', 'f'}, p1 {'c', 'd'}
Table 2: p2 {'a', 'e'}, p1 {'b', 'f'}, p3 {'c', 'd'}

Then applying 2:

Table 1: ['a', 'b', 'c', 'd', 'e', 'f'], [1, 2, 3, 3, 1, 2], [3, 1, 2]
Table 2: ['a', 'b', 'c', 'd', 'e', 'f'], [1, 2, 3, 3, 1, 2], [2, 3, 1]

This is user-specification independent representation wherein the
partition numbers in the 2nd array are based on canonical representation
(ordered lists).  To check pairwise join compatibility, simply compare the
first two arrays.  As to which partitions (think OIDs, RTEs whatever) pair
with each other, simply pair corresponding elements of the 3rd array which
are original partitions numbers (or OIDs).  Also when routing a tuple, we
find partition number in the array 2 and then look up the array 3 to get
the actual partition to insert the tuple.

Neither of these representations make the logic of checking pairwise-join
compatibility and pairing a subset of partitions (on either side) any
easier, although I haven't given it a good thought yet.