SQL查询以比较单个表中彼此之间的记录行

Question

I'm trying to work with data in a table that in my opinion is somewhat incomplete and I cannot figure out how to approach this problem or how begin to frame the question to see if what I'm trying to accomplish is even possible using SQL. This is a hypothetical representation of the data I'm working with (I entered the data in CSV format since this text field doesn't support table formatting):

Date,Time,Traveler,Source,Destination,Travel Status
9/20/2014,1:00pm,James,Station A,Station B,Scheduled
9/20/2014,1:10pm,James,Station A,Station B,Traveling
9/20/2014,1:40pm,James,,Station B,Arrived
9/20/2014,1:00pm,Ann,Station B,Station A,Scheduled
9/20/2014,1:10pm,Ann,Station B,Station A,Traveling
9/20/2014,1:40pm,Ann,,Station A,Arrived
9/20/2014,1:00pm,Karl,Station A,Station B,Scheduled
9/20/2014,1:10pm,Karl,Station A,Station B,Traveling
9/20/2014,1:40pm,Karl,,Station B,Arrived
9/20/2014,1:00pm,Joyce,Station B,Station A,Scheduled
9/20/2014,1:10pm,Joyce,Station B,Station A,Traveling
9/20/2014,1:40pm,Joyce,,Station A,Arrived
9/20/2014,1:00pm,Kelly,Station B,Station B,Scheduled
9/20/2014,1:10pm,Kelly,Station B,Station B,Traveling
9/20/2014,1:40pm,Kelly,,Station B,Arrived
9/20/2014,1:00pm,Sam,Station A,Station A,Scheduled
9/20/2014,1:10pm,Sam,Station A,Station A,Traveling
9/20/2014,1:40pm,Sam,,Station A,Arrived

I'm trying to count how many "types" of arrivals we had, for example how many arrivals of type A->A, how many of type B->B and how many A->B and B->A.

If the data was like this:

Date,Time,Traveler,Source,Destination,Travel Status
9/20/2014,1:00pm,James,Station A,Station B,Scheduled
9/20/2014,1:10pm,James,Station A,Station B,Traveling
9/20/2014,1:40pm,James,Station A,Station B,Arrived
9/20/2014,1:00pm,Ann,Station B,Station A,Scheduled
9/20/2014,1:10pm,Ann,Station B,Station A,Traveling
9/20/2014,1:40pm,Ann,Station B,Station A,Arrived

this simple query would accomplish this for each type of arrival, ie for type A->B:

SELECT COUNT(*) FROM TRAVEL_TBL WHERE 
Travel Status = 'Arrived' AND Source = 'Station A'
AND Destination = 'Station B';

But since the Source field is missing from the record that contains the "Arrived" entry, how can I carry out a query to find the counts? I guess the only way is by comparing each record chronologically somehow for each traveler sequentially and keeping track of when a trip was scheduled by whom and if they arrived and increment the count of this basis. Is this possible using SQL or can you only accomplish this writing an application in Java, or PHP or whatever host language to carry out the logic?

Answer 1

One solution that works with MS SQL 2012+ is to use the LAG() function to access previous rows:

SELECT COUNT(*) AS "Count A-B"
FROM (
    SELECT 
       Date, Time, Traveler, 
       CASE 
          WHEN Source IS NULL THEN LAG(Source,1) OVER (PARTITION BY Date, Traveler ORDER BY Date) 
          ELSE Source 
       END AS Source, 
       Destination, 
       [Travel Status]
from TRAVEL_TBL) derived_table
WHERE [Travel Status] = 'Arrived' AND Source = 'Station A' AND Destination = 'Station B';

Or a more generic version using ROW_NUMBER() (which is a function that should be available in most major databases) in a cte with a self join:

;WITH cte AS (
    SELECT 
       Date, Time, Traveler, 
       ROW_NUMBER() OVER (ORDER BY Traveler, Date, Time) rn,
       Source,
       Destination, 
       [Travel Status]
    FROM TRAVEL_TBL
) 

SELECT COUNT(*) AS "Count A-B"
FROM (
    SELECT 
       c.Date, c.Time, c.Traveler, 
       CASE 
          WHEN c.Source IS NULL THEN c2.source 
          ELSE c.Source 
       END AS Source,
       c.Destination, 
       c.[Travel Status]
    FROM cte c
    LEFT JOIN cte c2 ON c.rn = c2.rn+1
) derived_table
WHERE [Travel Status] = 'Arrived' AND Source = 'Station A' AND Destination = 'Station B';