SQL WHERE.. IN clause multiple columns

Postgres SQL : version 9.6
Total records on tables : mjr_agent = 145, mjr_transaction_item = 91800

1.Using with `EXISTS` [Average Query Time : 1.42s]

SELECT count(txi.id)
FROM
mjr_transaction_item txi
WHERE
EXISTS ( SELECT 1 FROM mjr_agent agnt WHERE agnt.agent_group = 0 AND (txi.src_id = agnt.code OR txi.dest_id = agnt.code) )

2.Using with two lines `IN` Clause [Average Query Time : 0.37s]

SELECT count(txi.id) FROM mjr_transaction_item txi
WHERE
txi.src_id IN ( SELECT agnt.code FROM mjr_agent agnt WHERE agnt.agent_group = 0 )
OR txi.dest_id IN ( SELECT agnt.code FROM mjr_agent agnt WHERE agnt.agent_group = 0 )

3.Using with `INNNER JOIN` pattern [Average Query Time : 2.9s]

SELECT count(DISTINCT(txi.id)) FROM mjr_transaction_item txi
INNER JOIN mjr_agent agnt ON agnt.code = txi.src_id OR agnt.code = txi.dest_id
WHERE
agnt.agent_group = 0

So , I choosed second option.

#WARNING ABOUT SOLUTIONS:

###MANY EXISTING SOLUTIONS WILL GIVE THE WRONG OUTPUT IF ROWS ARE NOT UNIQUE

If you are the only person creating tables, this may not be relevant, but several solutions will give a different number of output rows from the code in question, when one of the tables may not contain unique rows.

#WARNING ABOUT PROBLEM STATEMENT:
###IN WITH MULTIPLE COLUMNS DOES NOT EXIST, THINK CAREFULLY WHAT YOU WANT

When I see an in with two columns, I can imagine it to mean two things:

1. The value of column a and column b appear in the other table independently
2. The values of column a and column b appear in the other table together on the same row

Scenario 1 is fairly trivial, simply use two IN statements.

In line with most existing answers, I hereby provide an overview of mentioned and additional approaches for Scenario 2 (and a brief judgement):

##EXISTS (Safe, recommended for SQL Server)
As provided by @mrdenny, EXISTS sounds exactly as what you are looking for, here is his example:

SELECT * FROM T1
WHERE EXISTS
(SELECT * FROM T2
WHERE T1.a=T2.a and T1.b=T2.b)

##LEFT SEMI JOIN (Safe, recommended for dialects that support it)
This is a very concise way to join, but unfortunately most SQL dialects, including SQL server do not currently suppport it.

SELECT * FROM T1
LEFT SEMI JOIN T2 ON T1.a=T2.a and T1.b=T2.b

##Multiple IN statements (Safe, but beware of code duplication)
As mentioned by @cataclysm using two IN statements can do the trick as well, perhaps it will even outperform the other solutions. However, what you should be very carefull with is code duplication. If you ever want to select from a different table, or change the where statement, it is an increased risk that you create inconsistencies in your logic.

**Basic solution**

SELECT * from T1
WHERE a IN (SELECT a FROM T2 WHERE something)
AND b IN (SELECT b FROM T2 WHERE something)

**Solution without code duplication (I believe this does not work in regular SQL Server queries)**

WITH mytmp AS (SELECT a, b FROM T2 WHERE something);
SELECT * from T1
WHERE a IN (SELECT a FROM mytmp)
AND b IN (SELECT b FROM mytmp)

##INNER JOIN (technically it can be made safe, but often this is not done)
The reason why I don't recommend using an inner join as a filter, is because in practice people often let duplicates in the right table cause duplicates in the left table. And then to make matters worse, they sometimes make the end result distinct whilst the left table may actually not need to be unique (or not unique in the columns you select). Futhermore it gives you the chance to actually select a column that does not exists in the left table.

SELECT T1.* FROM T1
INNER JOIN
(SELECT DISTINCT a, b FROM T2) AS T2sub
ON T1.a=T2sub.a AND T1.b=T2sub.b

Most common mistakes:

0. Joining directly on T2, without a safe subquery. Resulting in the risk of duplication)
1. SELECT * (Guaranateed to get columns from T2)
2. SELECT c (Does not guarantee that your column comes and always will come from T1)
3. No DISTINCT or DISTINCT in the wrong place

#CONCATENATION OF COLUMNS WITH SEPARATOR (Not very safe, horrible performance)
The functional problem is that if you use a separator which might occur in a column, it gets tricky to ensure that the outcome is 100% accurate. The technical problem is that this method often incurs type conversions and completely ignores indexes, resulting in possibly horrible performance.
Despite these problems, I have to admit that I sometimes still use it for ad-hoc queries on small datasets.

SELECT * FROM T1
WHERE CONCAT(a,"_",b) IN
(SELECT CONCAT(a,"_",b) FROM T2)

Note that if your columns are numeric, some SQL dialects will require you to cast them to strings first. I believe SQL server will do this automatically.

---
To wrap things up: As usual there are many ways to do this in SQL, using safe choices will avoid suprises and save you time and headaces in the long run.

Concatenating the columns together in some form is a "hack", but when the product doesn't support semi-joins for more than one column, sometimes you have no choice.

Example of where inner/outer join solution would not work:

select * from T1
where <boolean expression>
and (<boolean expression> OR (ColA, ColB) in (select A, B ...))
and <boolean expression>
...

When the queries aren't trivial in nature sometimes you don't have access to the base table set to perform regular inner/outer joins.

If you do use this "hack", when you combine fields just be sure to add enough of a delimiter in between them to avoid misinterpretations, e.g. `ColA + ":-:" + ColB`

A simple EXISTS clause is cleanest

select *
from table1 t1
WHERE
EXISTS
(
Select * --or 1. No difference...
From CRM_VCM_CURRENT_LEAD_STATUS Ex
Where Lead_Key = :_Lead_Key
-- correlation here...
AND
t1.CM_PLAN_ID = Ex.CM_PLAN_ID AND t1.CM_PLAN_ID = Ex.Individual_ID
)

If you have multiple rows in the correlation then a JOIN gives multiple rows in the output, so you'd need distinct. Which usually makes the EXISTS more efficient.

Note `SELECT *` with a JOIN would also include columns from the row limiting tables

*** T-SQL ***

I used the string_agg as a cheap hack to get some pseudo-normalization on the cheep. (okay, its multi-plexing, I know its awful, I compensated your pain with some carefully crafted 80's style box drawing, enjoy!~)

here's a sample (had fun coming up with name replacements :D)

select
vendorId,
affiliate_type_code,
parent_vendor_id,
state_abbr,
county_abbr,
litigation_activity_indicator,
string_agg(employee_id,',') as employee_ids,
string_agg(employee_in_deep_doodoo,',') as 'employee-inventory connections'
from (
select distinct top 10000 -- so I could pre-order my employee id's - didn't want mixed sorting in those concats
mi.missing_invintory_identifier as rqid,
vendorId,
affiliate_type_code,
parent_vendor_id,
state_abbr,
county_abbr,
litigation_activity_indicator,
employee_identifier as employee_id,
concat(employee_identifier,'-',mi.missing_invintory_identifier) as employee_in_deep_doodoo
from
missing_invintory as mi
inner join vendor_employee_view as ev
on mi.missing_invintory_identifier = ev.missing_invintory_identifier
where ev.litigation_activity_indicator = 'N'
order by employee_identifier desc

) as x
group by
vendorId,
affiliate_type_code,
parent_vendor_id,
state_abbr,
county_abbr,
litigation_activity_indicator
having count(employee_id) > 1



┏━━━━━━━━━━┳━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓
┃ vendorId ┃ affiliate_type ┃ parent_vendor_id ┃ state_abbr ┃ county_abbr ┃ litigation_indicator ┃ employee_ids ┃ employee-inventory connections ┃
┣━━━━━━━━━━╋━━━━━━━━━━━━━━━━╋━━━━━━━━━━━━━━━━━━╋━━━━━━━━━━━━╋━━━━━━━━━━━━━━━━━━━╋━━━━━━━━━━━━━━━━━━━━━━╋━━━━━━━━━━━━━━━━━━━━━╋━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┫
┃ 123 ┃ EXP ┃ 17 ┃ CA ┃ SDG ┃ N ┃ 112358,445678 ┃ 112358-1212,1534490-1212 ┃
┣━━━━━━━━━━╋━━━━━━━━━━━━━━━━╋━━━━━━━━━━━━━━━━━━╋━━━━━━━━━━━━╋━━━━━━━━━━━━━━━━━━━╋━━━━━━━━━━━━━━━━━━━━━━╋━━━━━━━━━━━━━━━━━━━━━╋━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┫
┃ 4567 ┃ PRI ┃ 202 ┃ TX ┃ STB ┃ Y ┃ 998754,332165 ┃ 998754-4545,332165-4545 ┃
┗━━━━━━━━━━┻━━━━━━━━━━━━━━━━┻━━━━━━━━━━━━━━━━━━┻━━━━━━━━━━━━┻━━━━━━━━━━━━━━━━━━━┻━━━━━━━━━━━━━━━━━━━━━━┻━━━━━━━━━━━━━━━━━━━━━┻━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┛

```
with free_seats AS(SELECT seat_id , free as cureent_seat, LAG(free,1,0) OVER() previous_seat,LEAD(free,1,0) OVER() next_seat FROM seat)SELECT seat_id ,cureent_seat, previous_seat,next_seat from free_seats WHERE cureent_seat=1 AND ( previous_seat=1 OR next_seat=1) ORDER BY seat_id;
```