SQL Server JOINS - sql

Question

Can someone help explain to me how when I have 12 rows in table A and 10 in B and I do an inner join , I would get more rows than
in both A and B ?
Same with left and right joins...
This is just a simplified example. Let me share one of my issues with you
I have 2 views ; which was originally SQL on 2 base tables Culture and Trials.
And then when attempting to add another table Culture Steps, one of the team members separated the SQL into 2 views
Since this produces an error when updating(modification cannot be done as it affects multiple base tables), I would like to get
back to changing the SQL such that I no longer use the views but achieve the same results.
One of the views has
SELECT some columns
FROM dbo.Culture RIGHT JOIN
dbo.Trial ON dbo.Culture.cultureID = dbo.Trial.CultureID LEFT OUTER JOIN
dbo.TrialCultureSteps_view_part1 ON dbo.Culture.cultureID = dbo.TrialCultureSteps_view_part1.cultureID
The other TrialCultureSteps_view_part1 view
SELECT DISTINCT dbo.Culture.cultureID,
(SELECT TOP (1) WeekNr
FROM dbo.CultureStep
WHERE (CultureID = dbo.Culture.cultureID)
ORDER BY CultureStepID) AS normalstartweek
FROM dbo.Culture INNER JOIN
dbo.CultureStep AS CultureStep_1 ON dbo.Culture.cultureID = CultureStep_1.CultureID
So how can I combine the joins the achieve the same results using SQL only on tables without the need for views?

Answer 1

Welcome to StackOverflow! This link might be a good place to start in your understanding of JOINs. Essentially, the 'problem' you describe boils down to the fact that one or more of your sources (Trial, Culture, or the TrialCultureSteps view) has more than one record per CultureID - in other words, the same CultureID (#1) shows up on multiple rows.
Based solely on that ID, I'd execute the following three queries. Anything that is returned by them is the 'cause' of your duplications - the culture ID shows up more than once, so you'll have to JOIN on more than just CultureID. If, as I half-suspect, your view is the one that has multiple Culture IDs, you'll need to modify it to only return one record, or change the way that you JOIN to it.
SELECT *
FROM Trial
WHERE CultureID IN
(
SELECT CultureID
FROM Trial
GROUP BY CultureID
HAVING COUNT(*) > 1
)
ORDER BY CultureID
SELECT *
FROM Culture
WHERE CultureID IN
(
SELECT CultureID
FROM Culture
GROUP BY CultureID
HAVING COUNT(*) > 1
)
ORDER BY CultureID
SELECT *
FROM TrialCultureSteps_view_part1
WHERE CultureID IN
(
SELECT CultureID
FROM TrialCultureSteps_view_part1
GROUP BY CultureID
HAVING COUNT(*) > 1
)
ORDER BY CultureID
Let me know if any of these return values!

Answer 2

The comments explain the JOIN issues. As for rewriting, any views could be replaced with CTEs.
One other way to rewrite the query, would be : (Though having sample data and expected result would make this easier to confirm that it's correct)
;with TrialCultureSteps_view_part1 AS
(
Select Row_number() OVER (Partition BY CultureID ORDER BY CultureStepID) RowNumber
, WeekNr
, CultureID
)
SELECT some columns
dbo.trial LEFT OUTER JOIN
dbo.Culture ON dbo.Culture.cultureID = dbo.Trial.CultureID LEFT OUTER JOIN
TrialCultureSteps_view_part1 ON dbo.Culture.cultureID = dbo.TrialCultureSteps_view_part1.cultureID and RowNumber=1
Access code, I'm less familiar with the syntax, but I know that Row_Number() isn't available and I don't believe it has CTE syntax either. So, we'd need to put in some more nested derived tables.
SELECT some columns
dbo.trial LEFT OUTER JOIN
dbo.Culture ON dbo.Culture.cultureID = dbo.Trial.CultureID LEFT OUTER JOIN
( Select cs.CultureID, cs.WeekNr FROM
( SELECT CultureID, MIN(CultureStepID) CultureStepID
FROM dbo.CultureStep
GROUP BY CultureID
) Fcs INNER JOIN
CultureStep cs ON fcs.cultureStepID=cs.CultureStepID
) TrialCultureSteps_view_part1 ON dbo.Culture.cultureID = TrialCultureSteps_view_part1.cultureID
Assumptions here, is that CultureStepID is a PK for CultureStep. No assumption that a step must exist for each Culture entry.