CODEWITHSUNDEEP
sqlgoogle-bigquery
AmirCS
AmirCS

Reputation: 321

BigQuery - Find the closest region

I have two tables, and for each region in A, I want to find the closest regions in B. 

A:
------------------------
ID | Start | End | Color 
------------------------
 1 |  400  | 500 | White
------------------------
 1 |  10   | 20  | Red 
------------------------
 2 |   2   |  10 | Blue 
------------------------
 4 |   88  |  90 | Color 
------------------------

B:
------------------------
ID | Start | End | Name 
------------------------
 1 |  1    | 2   | XYZ1 
------------------------
 1 |  50   | 60  | XYZ4 
------------------------
 2 |  150  | 160 | ABC1 
------------------------
 2 |  50   | 60  | ABC2 
------------------------
 4 |  100  | 120 | EFG 
------------------------

RS:
---------------------------------------
ID | Start | End | Color | Closest Name
---------------------------------------
 1 |  400  | 500 | White |   XYZ4
---------------------------------------
 1 |  10   | 20  | Red   |   XYZ1
---------------------------------------
 2 |   2   |  10 | Blue  |   ABC2
---------------------------------------
 4 |   88  |  90 | Color |   EFG
---------------------------------------

Currently, I first find min distance by joining two tables:

MinDist Table:

SELECT   A.ID,   A.Start,  A.End,   
MIN(CASE  
WHEN (ABS(A.End-B.Start)>=ABS(A.Start - B.End)) 
THEN ABS(A.Start-B.End)     
ELSE ABS(A.End - B.Start) 
END) AS distance 
FROM ( Select A ... ) 
Join B On A.ID=B.ID) 
Group By A.ID,   A.Start,  A.End

Then recompute distance for by joining table A and B again, GlobDist Table (Note, the query retrieves B.Name in this case):

SELECT   A.ID,   A.Start,  A.End,   
CASE  
WHEN (ABS(A.End-B.Start)>=ABS(A.Start - B.End)) 
THEN ABS(A.Start-B.End)     
ELSE ABS(A.End - B.Start) 
END AS distance,
B.Name 
FROM ( Select A ... ) 
Join B On A.ID=B.ID)

Finally join these two tables MinDist and GlobDist Tables on 

GlobDist.ID= MinDist.ID, 
GlobDist.Start=MinDist.Start, 
GlobDist.End= MinDist.End, 
GlobDist.distance= MinDist.distance.

I tested ROW_NUMBER() and PARTITION BY over (ID, Start, End), but it took much longer. So, what's the fastest and most efficient way of solving this problem? How can I reduce duplicate computation?

Thanks!

Upvotes: 2

Views: 609

Answers (1)

Mikhail Berlyant
Mikhail Berlyant

Reputation: 172974

Below solution is for BigQuery Standard SQL and as simple and short as below 

#standardSQL
SELECT a_id, a_start, a_end, color,  
  ARRAY_AGG(name ORDER BY POW(ABS(a_start - b_start), 2) + POW(ABS(a_end - b_end), 2) LIMIT 1)[SAFE_OFFSET(0)] name
FROM A JOIN B ON a_id = b_id
GROUP BY a_id, a_start, a_end, color
-- ORDER BY a_id

You can test / play with above using dummy data in your question 

#standardSQL
WITH A AS (
  SELECT 1 a_id, 400 a_start, 500 a_end, 'White' color UNION ALL
  SELECT 1, 10,  20  , 'Red' UNION ALL
  SELECT 2, 2,   10, 'Blue' UNION ALL
  SELECT 4, 88,  90, 'Color'
), B AS (
  SELECT 1 b_id, 1 b_start, 2 b_end, 'XYZ1' name UNION ALL
  SELECT 1, 50, 60,  'XYZ4' UNION ALL
  SELECT 2, 150, 160,'ABC1' UNION ALL
  SELECT 2, 50, 60,  'ABC2' UNION ALL
  SELECT 4, 100, 120,'EFG'
)
SELECT a_id, a_start, a_end, color,  
  ARRAY_AGG(name ORDER BY POW(ABS(a_start - b_start), 2) + POW(ABS(a_end - b_end), 2) LIMIT 1)[SAFE_OFFSET(0)] name
FROM A JOIN B ON a_id = b_id
GROUP BY a_id, a_start, a_end, color
ORDER BY a_id

with result as below 

Row a_id    a_start a_end   color   name     
1   1       400     500     White   XYZ4     
2   1       10      20      Red     XYZ1     
3   2       2       10      Blue    ABC2     
4   4       88      90      Color   EFG

Upvotes: 2

Related Questions