Initial Thoughts

POC-iterations/M2Mlooping/LoopR&DIdeas.md

@@ -0,0 +1,49 @@
+November 10, 2023
+R&D Discussion
+Neil Beesley and Kip Twitchell
+
+Michael Shapiro talked this week of a need for many to many processing capabilities.  Neil asked how we might satisfy this need in the most simple way.  So we spent an hour discussing it.
+
+Objective: 
+    - Smallest number of changes to existing tools
+    - No required exits
+
+Proposal:
+    - Workbench changes:
+      - New type of LR Index: Non unique key
+      - New Logic Text Keyword:  Next Lookup {Join.ID}
+    - MR91
+      - New LUNX (replacing LUSM) to target LR
+      - Perhaps an additional LKC before LUNX giving search mode
+        - 0 = Start Search
+        - 1 = Next Record
+      - CFNX for Next Lookup 
+        - true row pointing back up to LUNX
+        - false row as the associated Join false row
+    - MR95
+      - Allow non-unique table load, removing binary search path build 
+      - Model code
+        - LUNX
+        - CFNX
+
+Sample Logic Table:
+    RENX
+    NV
+    JOIN
+    LKE....
+    LKC = 0 for start search
+    LUNX  
+        - Initial call finds first row in table for partial key given
+        - Branch not found if no match found (assuming beyond CFNX)
+        - Set next record pointer to next record in table
+    ....
+    WR..
+    CFNX
+        - Change LKC = 1 to Next Record
+        - Test pointer to next record for equal to lookup key
+        - Test true, branch back to LUNX
+        - Test False, fall through
+
+Assumptions:
+    - Each LKNX - CFNX operates a stack
+    - Nested sets are nested loops