Storage efficiency of Postgres index column

Question

In my PostgreSQL database, I have this table:

 Table "public.data"
 Column | Type | Modifiers 
-----------+--------+-----------
 group_id | bigint | 
 item | text | 
 price | real | 
Indexes:
 "group_id_index" btree (group_id)

item's are organized into groups, and the group_id column will have individual values appearing many times in the table. Queries to this table have the form:

SELECT item, price FROM data WHERE group_id = 1234;

group_id_index speeds up these queries (as is its purpose).

My question is, does the presence of the index make Postgres realize it does not have to store the group_id column behind-the-scenes? If only one copy (for the index) of each value of group_id is stored, that would save a lot of space.

This is, if it matters, PostgreSQL version 9.4.1

• Craig Ringer: 9.4.1? Upgrade NOW. Why on earth would you keep running an old point release like that? In this case there are known serious bugs in 9.4.1. Upgrade urgently to the latest in the 9.4.x series. You do not need to dump and reload, just install the new binaries.
• AntsySysHack: I appreciate the suggestion, but unfortunately it's not in my control. This database lives on a shared virtual cluster that is updated pretty infrequently
• Craig Ringer: then migrate somewhere else, because that one's going to destroy your data.There are severe bugs in 9.4.1. Read the release notes. Get regular dumps going. And plan your migration.

Answer 1

My question is, does the presence of the index make Postgres realize it does not have to store the group_id column behind-the-scenes?

No. PostgreSQL uses heaps, not index-organised tables.

• AntsySysHack: Hi Craig, thanks for your answer, but can you expand on it or provide any references?
• Craig Ringer: @AntsySysHack The source code is the primary reference here. But it's the absence of references - there's nothing that says it does do what you want.