How to use DISTINCT in VIEWS correctly - postgresql

Question

I have two tables from which I want to make a join with some columns to provide a view for my java/hibernate application. It looks like this:
CREATE VIEW customer_contacts AS cc
SELECT DISTINCT ON (cust.id) cust.id
cust.company
cust.zip
...
con.name
con.forename
...
FROM contacts con
LEFT JOIN customer cust ON con.customer = cust.id
ORDER BY cust.id
So far so good. Very simple.
If I make a SELECT on the view like:
SELECT *
FROM cc
WHERE name ilike '%schult%'
I get 13 results.
If I make the same query directly with the view statement
SELECT DISTINCT ON (cust.id) cust.id
cust.company
cust.zip
...
con.name
con.forename
...
FROM contacts con
LEFT JOIN customer cust ON con.customer = cust.id
WHERE name ilike '%schult%'
ORDER BY cust.id
I got 75 results!
I figured out that it is the DISTINCT that corrupts the result. But why?
And how can I use it correctly?

Answer 1

Your queries (view based and direct) have different order of applying condition:
direct query searches for %shult% and then applies distinct on
view applies distinct on and then searches for %shult%
Are you aware how distinct on works?
It selects first row (it may be undeterministic if proper sort is not defined) for given attributes and leaves other.
For instance:
Let's say we have customer with id=1 and two connected contacts one with name='Schultz' and one with name='Schmidt'.
Now view based select will apply distinct on and select customer with some contact (first one, undeterministic in this case), then name ilike '%schult%' will be applied - it may happen that Schultz will be removed by distinct on.
Recommended reading:
https://www.postgresql.org/docs/9.0/static/sql-select.html#SQL-DISTINCT