Re: [WIP]Vertical Clustered Index (columnar store extension) - take2

Thank You for working on this. I started going through the README and
tried running simple tests, have few concerns:

1)
I am not able to understand section 4.2 'WOS-to-ROS conversion'. When
whiteout-WOS says 'delete 4', what does that mean? 4 is CRID, TXID?
And when does delete-vector X represents? I did not  get why ColA-2,
ColA-4 and ColB-2, ColB-4 were removed in resultant data? Is the
diagram complete?

2)
We can make the definition consistent at both places as the first one
gives a feeling that rows are marked for deletion in WOS while the
second one says ROS.

Whiteout WOS = Record of WOS rows marked for deletion
Whiteout WOS -- TID records of WOS rows that are marked for deletion on ROS

3)
It is not part of README. But please help me understand the meaning
and usage of this GUC in VCI context:
vci.max_devices: Sets the maximum device number which can be attached.

4)
I was trying my hands on using VCI, I found that on enabling the VCI
path, autovacuum asserts in between. We can reproduce the scenario
using:
create extension vci;
create table tab1( i int, j int, k int);
create index ivci on tab1 using vci (i);
insert into tab1 SELECT generate_series(1,1000),
generate_series(1,1000), generate_series(1,1000);
insert into tab1 SELECT generate_series(1,1000),
generate_series(1,1000), generate_series(1,1000);
SET enable_seqscan TO off;
VACUUM ANALYZE public.tab1;

It asserts here:
TRAP: failed Assert("TransactionIdIsNormal(xmax)"), File: "heapam.c",
Line: 7373, PID: 79834
postgres: autovacuum worker postgres(ExceptionalCondition+0xbb)[0x55fdf8f3ccc6]
postgres: autovacuum worker
postgres(heap_pre_freeze_checks+0x18f)[0x55fdf877e26f]
postgres: autovacuum worker postgres(index_vacuum_cleanup+0x181)[0x55fdf87a1c06]
postgres: autovacuum worker postgres(vac_cleanup_one_index+0x27)[0x55fdf8a0503a]

-------------

Few typos in README:
a) Each VCI indexed column is stored as an internal relations.
--relations --> relation

b) Records are addresses by CRID (Columnar Record ID) instead of by TID.
--addresses->addressed

c) Extents can be found by ID using offsets in a column "meta-data"
internal relation.
-- by ID using offsets? Do you mean 'by using offsets' alone?

d) EXPLAIN ANALYSE -->EXPLAIN ANALYZE


thanks
Shveta