SQL：将一个表中的 2 列与另一表中的 1 列进行比较

Question

I have 2 tables in MSSQL one is called [DATA] and the other one is called [2020]

IN [2020] I have user info for example: name , ID, adress, and so on. IN [DATA] I have multiple columns that I need to show about the users

the catch is that I have no common column between those tables, in [2020] I have ID and adress, in [DATA] I have that info in one column called ID-adress

SO I tried this

SELECT
    A.ID,
    A.Nro,
    A.name,
    A.adress,
    B.CGE AS 'GE',
    B.CGG as 'GG',
    B.CPE AS 'PE'

FROM 
    [202009] A
    LEFT JOIN [DATA] B ON str(A.ID )+'-'+cast(A.adress as varchar) = B.[ID-adress]

But I get NULL on the columns that I want to show on B ¿what I'm doing wrong?

Answer 1

From your rather sparse description, I would guess that the JOIN condition never evaluates to true.

Based on the naming of the columns, you may want:

FROM [202009] A LEFT JOIN
     [DATA] B
     ON CONCAT(A.ID, '-', A.adress) = B.[ID-adress]

This is speculation based on a reasonable interpretation of the question.

There are two potential problems with your construct:

1. str() tends to pad values with spaces.
2. varchar without a length has a default length depending on context that may not be sufficient for what you want to do.

Answer 2

以下是调试建议：在 SELECT 中添加表达式（ str(A.ID )+'-'+cast(A.adress as varchar) ）以查看结果是否符合您的预期。

Answer 3

Your immediate issue, I believe, is that your string functions str(A.ID )+'-'+cast(A.adress as varchar) have several problems.

I suggest just putting that SQL fragment into a SELECT to see what you get (eg,

SELECT A.ID, A.adress, str(A.ID )+'-'+cast(A.adress as varchar) 
FROM [202009] A

to see what results you get.

I believe

• The ID is numeric (int, probably). str(A.ID) will usually put leading spaces before it, and when comparing strings, ' 5' is different from '5' . Put an LTRIM in front of it eg, ltrim(str(A.ID))
• cast(A.adress as varchar) needs a length of the varchar eg, cast(A.adress as varchar(100)). If you don't do this, it assumes length 1.

To check, run a similar statement as above, but with fixes

SELECT A.ID, A.adress, ltrim(str(A.ID))+'-'+cast(A.adress as varchar(100)) 
FROM [202009] A

However (and I don't like to say this), but this design is ... problematic. For example, What happens when a user changes their address? All fields in the data table will be unusable.

The biggest-bang-for-buck fix (imo) is to have 2 separate fields in the Data table for ID and address . If needed, you can still keep your single field as well (ID-address).

As you say, make sure they have common columns.

Once you have the two fields, you can join them on fields that at least match to get the relevant data rows for that address eg,

SELECT
    A.ID,
    A.Nro,
    A.name,
    A.adress,
    B.CGE AS 'GE',
    B.CGG as 'GG',
    B.CPE AS 'PE'

FROM 
    [202009] A
    LEFT JOIN [DATA] B ON A.ID = B.ID AND A.adress = B.adress

And then, if you just want all the data rows for a given user, you can do the same thing but do not join on address (just A.ID = B.ID).

Note that this won't fix all the issues (eg, when an address in one table is different from the other table because of things like spaces, or 'St' vs 'Street') - but will be a good start.