In a message dated 3/22/2012 6:02:00 PM Central Daylight Time, [email protected] writes: > Question 1 **************************** > Dr. Codd appears to be freeing us from having to indicate 'levels' with > column structures. In his text: > > "... Q1, Q2, ...,Qn denote immediate properties of each particular > SUBORDINATION" > > So in his construction, these might be reasonable values for Qx: > > Q1: Quantity of SUB_P# in SUP_P# > Q2: Joining method (threading, welding, painting) > Q3: Resultant Weight >
Okay, I understand that. In my table, I have a "Q1" table that indicates the quantity of the subpart in the upper level part. > > Question 2 ***************************** > > That's a thought of mine, extrapolating from yesterday's 'SQL Help' > example and a few snippets from Dr. Codd. (Let's just say I'm hopeful). > > " I claim to have a solution to the general bill-of-materials problem, > one that is very concise, that protects the user from iterative and > recursive programming, and that provides pertinent integrity constraints > as well as manipulative power. However, the recursive join described in > Chapter 5 is not a complete solution to this problem." > > And then, maddeningly: > > "The more complete solution will, of course, be published later." > Wow, that's a cop-out. Bad Codd... > The recursive join of Chapter 5: > > "The recursive join is an operator with one operand. This operand is a > relation that represents a directed graph [map of parent-to-child > relationships]. One of the columns of this relation plays a subordinate > (SUB) role, while another plays a superior (SUP) role. Each tuple [row] > represents an edge [connection] of a directed graph [parent-value-node > to child-value-node], and by convention this edge is directed from the > node identified by the SUP component down to the node identified by the > SUB component. Because joins are normally applied to pairs of relations, > it is convenient to think of the single operand as two identical > relations. The recursive join acts on this pair of identical relations > by matching each SUB value in one operand to a SUP value in the second > operand. It yields all of the pairs of identifiers for nodes that are > connected by paths in the acylic graph [a map of all the node > interconnections], NO MATTER WHAT THE PATH LENGTHS [levels] ARE > [emphasis mine]." [Codd 1990, Ch.5, Sect.5-10, p140] > So in his Chapter 5, does he show the syntax of any recursive join? Wait -- In another email that Bruce sent me, he scanned in pages from Chapter 5 for me. Been a long time since I sat down and tried to wrap my head around such a tome, but I swear I'm going to try. And there are NO examples of a self-join in there. Come on, how do you run on and on about them and then not show a single example.... Karen
