grouping relationships of many-to-many-to many - mysql

Question

I have a requirement to map a tax rate to a person based on the country the person was resident at that time.
tbl: person
| p_id | name_first | name_last |
=======++========================
| 1 | john | smith |
| 2 | joanne | smyth |
tbl: person_in_country
| p_id | iso | arrived |
===========================
| 1 | GB | 1980-01-01 |
| 2 | FR | 1987-03-21 |
| 1 | FR | 2003-06-17 |
| 1 | JP | 2008-07-02 |
| 2 | GB | 2008-10-01 |
| 1 | GB | 2009-01-10 |
tbl: country
| iso | ctry_name |
========================
| GB | United Kingdom |
| FR | France |
| JP | Japan |
tbl: tax_rates
| iso | tax_rate | tax_date |
===============================
| GB | 17.5 | 1970-01-01 |
| FR | 15.0 | 1977-03-21 |
| JP | 12.0 | 1977-06-17 |
| FR | 15.0 | 1994-03-21 |
| JP | 18.5 | 2008-07-02 |
| GB | 15 | 2008-04-01 |
| GB | 20 | 2010-05-01 |
So I need tuples containing the person in the country and the tax rate that they should have at a given time..
Something along the lines of:
select p.p_id, p.name_first, p.name_last,
pic.arrived,
c.iso, c.ctry_name,
t.tax_rate
from people p
left join (select * from person_in_country order by arrived desc) pic using (p_id)
left join country c on c.iso = pic.iso
left join (select * from tax_rates order by tax_date desc) t on t.iso = c.iso
where t.tax_date <= NOW()
group by p.pid, pic.arrived, t.tax_date
Hope this make sense... and many thanks in advance

Answer 1

Actually, you have to do a query basicly in three steps. First one you are going to retrieve a kind of "raw data" with all desired columns, joining the relating tables, whatever these columns will be used to join or to retrieve choosen data.
After that, you have to group data in order to turn only last dates from the matching join.
Finally, you have to query again tax table to retrieve tax in the tax date current at the moment of arrival.
It is possible that there is a more easy or ellegant way to do so, but this query is working. Check your system performance depending on the query demmands. It seems a bit hard at a first glance, but it isn't when taking a more carefull look. The SQL code:
SELECT
c02.iso,
c02.p_id,
c02.name_first,
c02.name_last,
c02.ctry_name,
c02.arrived,
c02.mtax_date,
tax_rates.tax_rate
FROM (
SELECT
c01.iso,
c01.p_id,
c01.name_first,
c01.name_last,
c01.ctry_name,
c01.arrived,
Max(c01.tax_date) AS mtax_date
FROM (
SELECT
country.iso,
person.p_id,
person.name_first,
person.name_last,
country.ctry_name,
person_in_country.arrived,
tax_rates.tax_date
FROM
tax_rates
INNER JOIN (
country
INNER JOIN (
person
INNER JOIN
person_in_country
ON
person.p_id = person_in_country.p_id
)
ON
country.iso = person_in_country.iso
)
ON
tax_rates.iso = person_in_country.iso
GROUP BY
country.iso,
person.p_id,
person.name_first,
person.name_last,
country.ctry_name,
person_in_country.arrived,
tax_rates.tax_date
HAVING (((tax_rates.tax_date)<=[arrived]))
) as c01
GROUP BY
c01.iso,
c01.p_id,
c01.name_first,
c01.name_last,
c01.ctry_name,
c01.arrived
) as c02
INNER JOIN
tax_rates ON (
c02.mtax_date = tax_rates.tax_date
)
AND
(
c02.iso = tax_rates.iso
);
Output:
iso p_id name_first name_last ctry_name arrived mtax_date tax_rate
GB 1 john smith United Kindom 01/01/1980 01/01/1970 18
FR 2 joanne smyth France 21/03/1987 21/03/1977 15
FR 1 john smith France 17/06/2003 21/03/1994 15
JP 1 john smith Japan 02/07/2008 02/07/2008 18
GB 1 john smith United Kindom 10/01/2009 01/04/2008 15
GB 2 joanne smyth United Kindom 01/10/2008 01/04/2008 15