How to (Dirty) Pair DateTimes Across Two Tables - sql-server

Question

I am looking at a SQL Server 2008 Database with two Tables, each with a PK (INT) column and a DateTime column.
There is no explicit relationship between the Tables, except I know the application has a heuristic tendency to insert to the database in pairs, one row into each Table, with DateTimes that seem to never match exactly but are usually pretty close.
I am trying to match back up the PKs in each table by finding the closest matching DateTime in the other table. Each PK can only be used once for this matching.
What is the best way to do this?
EDIT: Sorry, please find at bottom some example input and desired output.
+-------+-------------------------+
| t1.PK | t1.DateTime |
+-------+-------------------------+
| 1 | 2016-08-11 00:11:03.000 |
| 2 | 2016-08-11 00:11:08.000 |
| 3 | 2016-08-11 11:03:00.000 |
| 4 | 2016-08-11 11:08:00.000 |
+-------+-------------------------+
+-------+-------------------------+
| t2.PK | t2.DateTime |
+-------+-------------------------+
| 1 | 2016-08-11 11:02:00.000 |
| 2 | 2016-08-11 00:11:02.000 |
| 3 | 2016-08-11 22:00:00.000 |
| 4 | 2016-08-11 11:07:00.000 |
| 5 | 2016-08-11 00:11:07.000 |
+-------+-------------------------+
+-------+-------+-------------------------+-------------------------+
| t1.PK | t2.PK | t1.DateTime | t2.DateTime |
+-------+-------+-------------------------+-------------------------+
| 1 | 2 | 2016-08-11 00:11:03.000 | 2016-08-11 00:11:02.000 |
| 2 | 5 | 2016-08-11 00:11:08.000 | 2016-08-11 00:11:07.000 |
| 3 | 1 | 2016-08-11 11:03:00.000 | 2016-08-11 11:02:00.000 |
| 4 | 4 | 2016-08-11 11:08:00.000 | 2016-08-11 11:07:00.000 |
+-------+-------+-------------------------+-------------------------+

Answer 1

JOIN to the row with lowest DATEDIFF (in seconds) between t1.DateTime and t2.DateTime.

Answer 2

You can achieve the result you are looking for by cross joining table 1 with table 2 and then getting the difference of the dates in seconds as per Tab Alleman’s suggestion. The next step would then be to rank each match using the ROW_NUMBER() function. Final step is to select out only rows which Rank = 1.
The following example demonstrates using your example data:
DECLARE #t1 TABLE
(
ID INT PRIMARY KEY
,[DateTime] DATETIME
);
DECLARE #t2 TABLE
(
ID INT PRIMARY KEY
,[DateTime] DATETIME
)
INSERT INTO #t1
(
ID
,[DateTime]
)
VALUES
(1 ,'2016-08-11 00:11:03.000'),
(2 ,'2016-08-11 00:11:08.000'),
(3 ,'2016-08-11 11:03:00.000'),
(4 ,'2016-08-11 11:08:00.000');
INSERT INTO #t2
(
ID
,[DateTime]
)
VALUES
(1, '2016-08-11 11:02:00.000'),
(2, '2016-08-11 00:11:02.000'),
(3, '2016-08-11 22:00:00.000'),
(4, '2016-08-11 11:07:00.000'),
(5, '2016-08-11 00:11:07.000');
WITH CTE_DateDifference
AS
(
SELECT t1.ID AS T1_ID
,t2.ID AS T2_ID
,t1.[DateTime] AS T1_DateTime
,t2.[DateTime] AS T2_DateTime
,ABS(DATEDIFF(SECOND, t1.[DateTime], t2.[DateTime])) AS Duration -- Determine the difference between the dates in seconds.
FROM #t1 t1
CROSS JOIN #t2 t2
),CTE_RankDateMatch
AS
(
SELECT T1_ID
,T2_ID
,T1_DateTime
,T2_DateTime
,ROW_NUMBER() OVER (PARTITION BY T1_ID ORDER BY Duration) AS [Rank] -- Rank each match, the row numbers generated will be order based on the duration between the dates. Thus rows with a number of 1will be the closest match between the two tables.
FROM CTE_DateDifference
)
-- Finally select out the rows with a Rank equal to 1.
SELECT *
FROM CTE_RankDateMatch
WHERE [Rank] = 1