Left join statement returning unexpected null values for rows/column that have data - sql

Question

I have two tables that I'm writing a query against. Some of the columns can be found in one of the tables, while some of the columns are calucalted.
For clarity, I will copy my query below:
select field_a,
cast(field_b as int),
field_c,
field_d,
Year,
coalesce( cast(field_e as float),0) America_spend,
sum( cast(field_e as float), 0) as America_spend,
coalesce( cast(field_e as float)/ sum( cast( field_e as float)) over(partition by Year) as total_spend
from table_a
left join table_b on
table_a.flield_a = table_b.field_a1 and
table_a.flield_b = table_b.field_b1 and
table_a.Year = table_b.Year
group by field_a,
field_b,
Year
I have tables that look like this
table a:
|field_a|field_b|field_c|field_d|Year|field_f|field_g|field_h
|data | 1 | data | data |2014| data | data | data
|data | 1 | data | data |2014| null | data | data
|0 | 1 | data | data |2014| data | data | data
|data | 1 | data | data |2014| null | data | data
|0 | 1 | data | data |2014| data | data | data
table b:
|field_a1|field_b1|Year|field_c1|field_j
|null | 1 |2014| data | data
|data | 1 |2015| data | data
|null | 0 |2014| data | data
|data | 1 |2015| data | data
|null | 0 |2014| data | data
The problem that I'm having is that some of the values in the 'total spend column' get assigned a value of null. Total spend is calculated per year and this field should never be null. Likewise, the year column doesn't contain a null value in either of the tables. But for some reason when I run the query, I get results that have some of the rows in the year column with a null value. This should never happen. Most of the results conform to what I would expect, but there are some that do not.
I'm guessing that is has something to do with the fact that some of the rows in field_b are null and get converted to 0, but why does this matter?
I updated the tables and the queries to more accurately reflect the structure of the database.
Yes the query runs and I have no naming conflicts.

Answer 1

#SeanLange's comment is what is most likely the issue with your expected results. That is that NULL does NOT Equal NULL (NULL <> NULL). Null is an "unknown" value in sql and 2 unknowns are not equal.
But you can eliminate your NULL if you want to match them together as the same case. Simply use COALESCE() or ISNULL() and provide the same default value on both sides of your ON condition and make sure your default is not represented within your dataset or you will get undesired results.
DECLARE #TableA AS TABLE (FieldA VARCHAR(5),FiledB INT,Yr INT)
DECLARE #TableB AS TABLE (FieldA VARCHAR(5),FiledC INT,Yr INT)
INSERT INTO #TableA (FieldA,FiledB,Yr)
VALUES (null,1,2014),('data',1,2015),(null,1,2014),('data',1,2015),(null,1,2014)
INSERT INTO #TableB (FieldA,FiledC,Yr)
VALUES (null,1,2014),('data',1,2015),(null,1,2014),('data',1,2015),(null,1,2014)
SELECT *
FROM
#TableA a
LEFT JOIN #TableB b
ON COALESCE(a.FieldA,'NULLVALUE') = COALESCE(b.FieldA,'NULLVALUE')
AND a.Yr = b.Yr
Your particular example dataset that you provided us repeats FieldA to Yr Combinations so the results are a little funky but it still works.