Speed up this query for creating a materialized view? - postgresql

Question

I'm using Postgres 9.4 on OSX with 16GB of RAM. I am creating a materialized view on my Postgres tables, but it is tremendously slow to create and I'm wondering if there is any way to speed things up.
This are the underlying tables, showing spending items and organisations:
Table "public.spending_item"
Column | Type | Modifiers
-------------------+-------------------------+--------------------------------------------------------------------
id | integer | not null default nextval('spending_item_id_seq'::regclass)
presentation_code | character varying(15) | not null
presentation_name | character varying(1000) | not null
total_items | integer | not null
actual_cost | double precision | not null
processing_date | date | not null
organisation_id | character varying(9) | not null
Table "public.organisation"
Column | Type | Modifiers
-----------+------------------------+-----------
code | character varying(9) | not null
name | character varying(200) | not null
There are 100m rows in the spending_item table, and around 10,000 in the organisation table.
The materialized view I want to create is as follows - basically it's total spending and total items per month, grouped by by presentation code and organisation.
(I'm creating it because I need queries like "for organisations with codes like 0304%, give me total spending by month and by organisation", or "for the organisation with code 030405, give me total spending by month and by presentation_code". My data only changes once a month, so I think a materialized view makes sense for this.)
Materialized view "public.vw_spending_summary"
Column | Type | Modifiers
-------------------+-------------------------+-----------
processing_date | date |
organsiation_id | character varying(9) |
organisation_name | character varying(200) |
total_cost | double precision |
total_items | double precision |
presentation_name | character varying(1000) |
presentation_code | character varying(15) |
The query I'm using to create the materialized view is as follows:
SELECT spending_item.processing_date,
spending_item.organisation_id,
organisation.name,
spending_item.presentation_code,
spending_item.presentation_name,
SUM(spending_item.total_items) AS items,
SUM(spending_item.actual_cost) AS cost
FROM organisation, spending_item
WHERE spending_item.organisation_id=organisation.code
GROUP BY spending_item.processing_date, spending_item.presentation_name,
spending_item.presentation_code, spending_item.organisation_id,
organisation.name;
The query has been running for many hours and has not completed yet.
This is the result of an EXPLAIN ANALYZE on a more restrictive query (as above, but with WHERE spending_item.organisation_id like '0101%' added):
GroupAggregate (cost=2263051.70..2291072.45 rows=934025 width=86) (actual time=60890.505..61069.808 rows=180 loops=1)
-> Sort (cost=2263051.70..2265386.76 rows=934025 width=86) (actual time=60890.497..60945.558 rows=397717 loops=1)
Sort Key: spending_item.processing_date, spending_item.presentation_name, spending_item.presentation_code, spending_item.organisation_id, organisation.name
Sort Method: external sort Disk: 32240kB
-> Hash Join (cost=23877.71..2125733.64 rows=934025 width=86) (actual time=180.302..42251.826 rows=397717 loops=1)
Hash Cond: ((spending_item.organisation_id)::text = (organisation.code)::text)
-> Bitmap Heap Scan on spending_item (cost=23276.92..2109954.94 rows=934025 width=68) (actual time=178.521..41743.141 rows=397717 loops=1)
Filter: ((organisation_id)::text ~~ '0201%'::text)
-> Bitmap Index Scan on spending_item_organisation_id_varchar_pattern_ops_idx (cost=0.00..23043.41 rows=924684 width=0) (actual time=136.077..136.077 rows=397717 loops=1)
Index Cond: (((organisation_id)::text ~>=~ '0201'::text) AND ((organisation_id)::text ~<~ '0202'::text))
-> Hash (cost=558.13..558.13 rows=3413 width=27) (actual time=1.774..1.774 rows=3413 loops=1)
Buckets: 1024 Batches: 1 Memory Usage: 203kB
-> Seq Scan on organisation (cost=0.00..558.13 rows=3413 width=27) (actual time=0.004..0.991 rows=3413 loops=1)
Total runtime: 61089.664 ms
I'd be grateful for any suggestions on speeding this up. I've already set maintenance_work_mem, shared_buffers etc high, so I don't think there's much I can do on the settings front.
Or perhaps it would make more sense to have two different materialized views, for the two types of queries I want to run?
Please let me know if there is any other information I can provide.