Improved step down SQL query without many IF/ELSE blocks - php

Question

Consider a sample table with these rows:
+----+----------+-------+
| id | postcode | value |
+----+----------+-------+
| 1 | A1A3A3 | one |
| 2 | A1A3A4 | two |
| 3 | A1A3B | three |
| 4 | A1A3C | four |
| 5 | A1A3D | five |
| 6 | A1A3 | six |
| 7 | A1A | seven |
| 8 | A1 | eight |
+----+----------+-------+
My goal is to perform a query, whereby it steps down through the postcode column until an exact match is found.
Let's say my starting query parameter is A1A3E9. The expected return value, based on the sample table, would be six. It is important to note, that each step down, I remove one character from the end of the starting query parameter.
So first I would try and find a match for A1A3E9, and then A1A3E, and then A1A3 and so forth.
Currently, I achieve this simply with a series of IF/ELSE blocks, like this:
IF
EXISTS (
SELECT value FROM table
WHERE postcode=:userPost6_1
)
BEGIN
SELECT value FROM table
WHERE postcode=:userPost6_2
END
ELSE IF
EXISTS (
SELECT value FROM table
WHERE postcode=:userPost5_1
)
BEGIN
SELECT value FROM table
WHERE postcode=:userPost5_2
END
ELSE IF
EXISTS (
SELECT value FROM table
WHERE postcode=:userPost4_1
)
BEGIN
SELECT value FROM table
WHERE postcode=:userPost4_2
END
ELSE IF
EXISTS (
SELECT value FROM table
WHERE postcode=:userPost3_1
)
BEGIN
SELECT value FROM table
WHERE postcode=:userPost3_2
END
ELSE IF
EXISTS (
SELECT value FROM table
WHERE postcode=:userPost2_1
)
BEGIN
SELECT value FROM table
WHERE postcode=:userPost2_2
END
Note that I am using parameter binding in PHP, so just for context, my parameter bindings ultimately look like this:
$stmt->bindValue(':userPost6_1', "A1A3E9", PDO::PARAM_STR);
$stmt->bindValue(':userPost6_2', "A1A3E9", PDO::PARAM_STR);
$stmt->bindValue(':userPost5_1', "A1A3E", PDO::PARAM_STR);
$stmt->bindValue(':userPost5_2', "A1A3E", PDO::PARAM_STR);
$stmt->bindValue(':userPost4_1', "A1A3", PDO::PARAM_STR);
$stmt->bindValue(':userPost4_2', "A1A3", PDO::PARAM_STR);
$stmt->bindValue(':userPost3_1', "A1A", PDO::PARAM_STR);
$stmt->bindValue(':userPost3_2', "A1A", PDO::PARAM_STR);
$stmt->bindValue(':userPost2_1', "A1", PDO::PARAM_STR);
$stmt->bindValue(':userPost2_2', "A1", PDO::PARAM_STR);
I do not have have any concerns so far as performance is concerned, as I have an index on the postcode column (which contains 40,000+ rows). My concern is purely that this is visually, an unpleasant query to look at.
My question: Is there a cleaner way to write this query?

Answer 1

Here is one method:
select top (1) t.*
from t
where 'A1A3E9' like t.postcode + '%'
order by t.postcode desc;
The only issue is that your multiple if statements are probably faster. Getting performance is a real challenge with this type of problem. One method uses multiple joins:
select v.pc, coalesce(t0.value, t1.value, t2.value, . . . )
from (values ('A1A3E9')) v(pc) left join
t t0
on t0.postcode = v.pc left join
t t1
on t1.postcode = t0.postcode is null and
(case when len(v.pc) > 1 then left(v.pc, len(v.pc) - 1) end) left join
t t2
on t1.postcode is null and
t2.postcode = (case when len(v.pc) > 2 then left(v.pc, len(v.pc) - 2) end) left join
. . .

Answer 2

I would first spool all the potentially matching rows, eg:
select *
into #matches
from t
where postcode like 'AI%'
This can use an index on postcode and so should be cheap. Then whatever query you run against the matches will just operate over this subset. Even writing a UDF that compares postcode to a literal, eg:
select top 1 *
from #matches
order by dbo.NumberOfMatchingCharacters(postcode,'A1A3E9') desc