我一直在尝试将此sp从t-sql转换为http://hansolav.net/sql/graphs.html#dijkstra到Teradata sql。除了递归查询，我已经完成了大部分内容。我知道Teradata SPS无法处理递归查询。有人知道如何重写递归查询以使此SP起作用吗？
我们目前正在使用TD 16。
这是我的代码
CREATE VOLATILE TABLE VT_Nodes
(
Id INT, -- The Node Id
Estimate DECIMAL(10,3) , -- What is the distance to this node, so far?
Predecessor INT, -- The node we came from to get to this node with this distance.
Done SMALLINT -- Are we done with this node yet (is the estimate the final distance)?
)PRIMARY INDEX(id)
ON COMMIT PRESERVE ROWS;
CREATE PROCEDURE db.usp_Dijkstra
(
StartNode INT
, EndNode INT /*= NULL*/
)
SQL SECURITY INVOKER
BEGIN
/* -- Automatically rollback the transaction if something goes wrong.
SET XACT_ABORT ON
BEGIN TRAN
*/
-- Increase performance and do not intefere with the results.
--SET NOCOUNT ON;
DECLARE FromNode INT;
DECLARE CurrentEstimate INT;
-- Fill the temporary table with initial data
INSERT INTO VT_Nodes (Id, Estimate, Predecessor, Done)
SELECT Id, 9999999.999, NULL, 0 FROM db.T_NODE;
-- Set the estimate for the node we start in to be 0.
UPDATE VT_Nodes SET Estimate = 0 WHERE Id =StartNode;
/* IF @@rowcount <> 1
BEGIN
DROP TABLE VT_Nodes
RAISERROR ('Could not set start node', 11, 1)
ROLLBACK TRAN
RETURN 1
END*/
-- Run the algorithm until we decide that we are finished
label1:
WHILE (1 = 1) DO
BEGIN
-- Reset the variable, so we can detect getting no records in the next step.
SET FromNode = NULL;
-- Select the Id and current estimate for a node not done, with the lowest estimate.
SET FromNode = (SELECT TOP 1 Id FROM VT_Nodes WHERE Done = 0 AND Estimate < 9999999.999 ORDER BY Estimate);
SET CurrentEstimate = (SELECT TOP 1 Estimate FROM VT_Nodes WHERE Done = 0 AND Estimate < 9999999.999 ORDER BY Estimate);
-- Stop if we have no more unvisited, reachable nodes.
IF FromNode IS NULL OR FromNode = EndNode THEN
LEAVE label1;
END IF;
-- We are now done with this node.
UPDATE VT_Nodes
SET Done = 1
WHERE Id = FromNode;
-- Update the estimates to all neighbour node of this one (all the nodes
-- there are edges to from this node). Only update the estimate if the new
-- proposal (to go via the current node) is better (lower).
UPDATE VT_Nodes
FROM db.T_Edge AS e
SET Estimate = CurrentEstimate + e.Weight
, Predecessor = FromNode
WHERE VT_Nodes.Id = e.ToNode
AND e.FromNode = FromNode
AND (CurrentEstimate + e.Weight) < VT_Nodes.Estimate
AND Done = 0
;
END ;
END WHILE label1;
DECLARE CUR1 CURSOR WITH RETURN ONLY TO CLIENT FOR
-- Select the results. We use a recursive common table expression to
-- get the full path from the start node to the current node.
WITH RECURSIVE BacktraceCTE(Id, nm, Distance, PATH, NamePath)
AS
(
-- Anchor/base member of the recursion, this selects the start node.
SELECT n.Id
, nd.nm
, n.Estimate
, CAST(n.Id AS VARCHAR(8000)) AS PATH
,CAST(nd.nm AS VARCHAR(8000)) AS NamePath
FROM VT_Nodes AS n
JOIN db.T_NODE AS nd
ON n.Id = nd.Id
WHERE n.Id = StartNode
UNION ALL
-- Recursive member, select all the nodes which have the previous
-- one as their predecessor. Concat the paths together.
SELECT n.Id
, nd.nm
, n.Estimate
,CAST((cte.PATH || ',' ||TRIM(n.Id)) AS VARCHAR(8000))
,CAST((cte.NamePath || ',' || nd.nm) AS VARCHAR(8000))
FROM VT_Nodes n
JOIN BacktraceCTE AS cte
ON n.Predecessor = cte.Id
JOIN db.T_NODE AS nd
ON n.Id = nd.Id
)
SELECT Id
, nm
, Distance
, PATH
, NamePath
FROM BacktraceCTE
WHERE (Id = EndNode OR EndNode IS NULL) -- This kind of where clause can potentially produce
ORDER BY Id -- a bad execution plan, but I use it for simplicity here.
;
OPEN CUR1;
DROP TABLE VT_Nodes;
/* COMMIT TRAN
RETURN 0*/
END ;
Tables
CREATE TABLE DB.T_NODE
(
Id INT
,Nm VARCHAR(50)
)PRIMARY INDEX(id)
;
CREATE TABLE DB.T_Edge
(
FromNode INT ,
ToNode INT,
Weight DECIMAL (10, 3)
) PRIMARY INDEX(FromNode ,ToNode)
;
DATA
INSERT DB.T_NODE (Id, Nm) VALUES (1, 'Seattle');
INSERT DB.T_NODE (Id, Nm) VALUES (2, 'San Francisco');
INSERT DB.T_NODE (Id, Nm) VALUES (3, 'Las Vegas');
INSERT DB.T_NODE (Id, Nm) VALUES (4, 'Los Angeles');
INSERT DB.T_NODE (Id, Nm) VALUES (5, 'Denver');
INSERT DB.T_NODE (Id, Nm) VALUES (6, 'Minneapolis');
INSERT DB.T_NODE (Id, Nm) VALUES (7, 'Dallas');
INSERT DB.T_NODE (Id, Nm) VALUES (8, 'Chicago');
INSERT DB.T_NODE (Id, Nm) VALUES (9, 'Washington DC');
INSERT DB.T_NODE (Id, Nm) VALUES (10, 'Boston');
INSERT DB.T_NODE (Id, Nm) VALUES (11, 'New York');
INSERT DB.T_NODE (Id, Nm) VALUES (12, 'Miami');
INSERT DB.T_Edge (FromNode, ToNode, Weight) VALUES (1, 2, 1306.000);
INSERT DB.T_Edge (FromNode, ToNode, Weight) VALUES (1, 5, 2161.000);
INSERT DB.T_Edge (FromNode, ToNode, Weight) VALUES (1, 6, 2661.000);
INSERT DB.T_Edge (FromNode, ToNode, Weight) VALUES (2, 1, 1306.000);
INSERT DB.T_Edge (FromNode, ToNode, Weight) VALUES (2, 3, 919.000);
INSERT DB.T_Edge (FromNode, ToNode, Weight) VALUES (2, 4, 629.000);
INSERT DB.T_Edge (FromNode, ToNode, Weight) VALUES (3, 2, 919.000);
INSERT DB.T_Edge (FromNode, ToNode, Weight) VALUES (3, 4, 435.000);
INSERT DB.T_Edge (FromNode, ToNode, Weight) VALUES (3, 5, 1225.000);
INSERT DB.T_Edge (FromNode, ToNode, Weight) VALUES (3, 7, 1983.000);
INSERT DB.T_Edge (FromNode, ToNode, Weight) VALUES (4, 2, 629.000);
INSERT DB.T_Edge (FromNode, ToNode, Weight) VALUES (4, 3, 435.000);
INSERT DB.T_Edge (FromNode, ToNode, Weight) VALUES (5, 1, 2161.000);
INSERT DB.T_Edge (FromNode, ToNode, Weight) VALUES (5, 3, 1225.000);
INSERT DB.T_Edge (FromNode, ToNode, Weight) VALUES (5, 6, 1483.000);
INSERT DB.T_Edge (FromNode, ToNode, Weight) VALUES (5, 7, 1258.000);
INSERT DB.T_Edge (FromNode, ToNode, Weight) VALUES (6, 1, 2661.000);
INSERT DB.T_Edge (FromNode, ToNode, Weight) VALUES (6, 5, 1483.000);
INSERT DB.T_Edge (FromNode, ToNode, Weight) VALUES (6, 7, 1532.000);
INSERT DB.T_Edge (FromNode, ToNode, Weight) VALUES (6, 8, 661.000);
INSERT DB.T_Edge (FromNode, ToNode, Weight) VALUES (7, 3, 1983.000);
INSERT DB.T_Edge (FromNode, ToNode, Weight) VALUES (7, 5, 1258.000);
INSERT DB.T_Edge (FromNode, ToNode, Weight) VALUES (7, 6, 1532.000);
INSERT DB.T_Edge (FromNode, ToNode, Weight) VALUES (7, 9, 2113.000);
INSERT DB.T_Edge (FromNode, ToNode, Weight) VALUES (7, 12, 2161.000);
，除非您运行旧的teradata版本，否则您可以使用递归，但是任何声明都必须在块中首先完成。只需要一些小修改：
REPLACE PROCEDURE usp_Dijkstra
(
StartNode INT
, EndNode INT /*= NULL*/
)
SQL SECURITY INVOKER
BEGIN
/* -- Automatically rollback the transaction if something goes wrong.
SET XACT_ABORT ON
BEGIN TRAN
*/
-- Increase performance and do not intefere with the results.
--SET NOCOUNT ON;
DECLARE FromNode INT;
DECLARE CurrentEstimate INT;
-- Fill the temporary table with initial data
INSERT INTO VT_Nodes (Id, Estimate, Predecessor, Done)
SELECT Id, 9999999.999, NULL, 0 FROM T_NODE;
-- Set the estimate for the node we start in to be 0.
UPDATE VT_Nodes SET Estimate = 0 WHERE Id =StartNode;
/* IF @@rowcount <> 1
BEGIN
DROP TABLE VT_Nodes
RAISERROR ('Could not set start node', 11, 1)
ROLLBACK TRAN
RETURN 1
END*/
-- Run the algorithm until we decide that we are finished
label1:
WHILE (1 = 1) DO
BEGIN
-- Reset the variable, so we can detect getting no records in the next step.
SET FromNode = NULL;
-- Select the Id and current estimate for a node not done, with the lowest estimate.
-- replacing two SET with a single SELECT
SELECT TOP 1 Id, Estimate INTO FromNode, CurrentEstimate
FROM VT_Nodes WHERE Done = 0 AND Estimate < 9999999.999 ORDER BY Estimate;
-- Stop if we have no more unvisited, reachable nodes.
IF FromNode IS NULL OR FromNode = EndNode THEN
LEAVE label1;
END IF;
-- We are now done with this node.
UPDATE VT_Nodes
SET Done = 1
WHERE Id = FromNode;
-- Update the estimates to all neighbour node of this one (all the nodes
-- there are edges to from this node). Only update the estimate if the new
-- proposal (to go via the current node) is better (lower).
UPDATE VT_Nodes
FROM T_Edge AS e
SET Estimate = :CurrentEstimate + e.Weight
, Predecessor = :FromNode
WHERE VT_Nodes.Id = e.ToNode
AND e.FromNode = :FromNode
AND (:CurrentEstimate + e.Weight) < VT_Nodes.Estimate
AND Done = 0
;
END ;
END WHILE label1;
BEGIN -- wrap the cursor within a new BEGIN/END block
DECLARE CUR1 CURSOR WITH RETURN ONLY TO CLIENT FOR
-- Select the results. We use a recursive common table expression to
-- get the full path from the start node to the current node.
WITH RECURSIVE BacktraceCTE(Id, nm, Distance, Path, NamePath)
AS
(
-- Anchor/base member of the recursion, this selects the start node.
SELECT n.Id
, nd.nm
, n.Estimate
, Cast(n.Id AS VARCHAR(8000)) AS Path
,Cast(nd.nm AS VARCHAR(8000)) AS NamePath
FROM VT_Nodes AS n
JOIN T_NODE AS nd
ON n.Id = nd.Id
WHERE n.Id = StartNode
UNION ALL
-- Recursive member, select all the nodes which have the previous
-- one as their predecessor. Concat the paths together.
SELECT n.Id
, nd.nm
, n.Estimate
,Cast((cte.Path || ',' ||Trim(n.Id)) AS VARCHAR(8000))
,Cast((cte.NamePath || ',' || nd.nm) AS VARCHAR(8000))
FROM VT_Nodes n
JOIN BacktraceCTE AS cte
ON n.Predecessor = cte.Id
JOIN T_NODE AS nd
ON n.Id = nd.Id
)
SELECT Id
, nm
, Distance
, Path
, NamePath
FROM BacktraceCTE
WHERE (Id = EndNode OR EndNode IS NULL) -- This kind of where clause can potentially produce
ORDER BY Id -- a bad execution plan, but I use it for simplicity here.
;
OPEN CUR1;
END;
DROP TABLE VT_Nodes;
/* COMMIT TRAN
RETURN 0*/
END ;