Remove @aimglobal@(11) node maintenance from $$XREFDATA^%YDBAIM as it seems to serve no purpose

Final Release Note

Description

This is something I noticed while investigating a TRANS2BIG error in Octo (YottaDB/DBMS/YDBOcto#1083 (closed)).

As line 525 states below, the xref global node @aimglobal@(11) contains the TOTAL number of nodes being tracked. This is maintained in lines 1147-1162 below as well as in various other places in _YDBAIM.m.

_YDBAIM.m

    525 ; - (11) if stat is 2, this contains the total number of nodes being tracked

But I don't understand what purpose it serves. Let me take the below example M program where ^names global has a id subscript and firstname|lastname as its value.

$ cat aimstat2.m
 do UNXREFDATA^%YDBAIM
 kill ^names
 set ^names(1)="First1|Last1"
 set base="^names",subs(1)=":"
 set aimglobal=$$XREFDATA^%YDBAIM(base,.subs,"|",1,0,0,1,2,0,1)
 set aimglobal=$$XREFDATA^%YDBAIM(base,.subs,"|",2,0,0,1,2,0,1)
 write "# xref of firstname=First1 and lastname=Last1",!
 do show
 write "# xref of firstname=First2 and lastname=Last2",!
 set ^names(2)="First2|Last2" do show
 write "# xref of firstname=First3 and lastname=Last2",!
 set ^names(3)="First3|Last2" do show
 write "# xref of firstname=First4 and lastname=Last2",!
 set ^names(4)="First4|Last1" do show
 quit

show
 zwrite ^%ydbAIMDERzGTIrQTVBoOhdanoKcC7(-2)
 zwrite ^%ydbAIMDERzGTIrQTVBoOhdanoKcC7(-1)
 zwrite ^%ydbAIMDERzGTIrQTVBoOhdanoKcC7(11)
 write !

Running the M program gives the below output.

$ mumps -run aimstat2
# xref of firstname=First1 and lastname=Last1
^%ydbAIMDERzGTIrQTVBoOhdanoKcC7(-2)=1
^%ydbAIMDERzGTIrQTVBoOhdanoKcC7(-1)=1
^%ydbAIMDERzGTIrQTVBoOhdanoKcC7(11)=2

# xref of firstname=First2 and lastname=Last2
^%ydbAIMDERzGTIrQTVBoOhdanoKcC7(-2)=2
^%ydbAIMDERzGTIrQTVBoOhdanoKcC7(-1)=2
^%ydbAIMDERzGTIrQTVBoOhdanoKcC7(11)=4

# xref of firstname=First3 and lastname=Last2
^%ydbAIMDERzGTIrQTVBoOhdanoKcC7(-2)=2
^%ydbAIMDERzGTIrQTVBoOhdanoKcC7(-1)=3
^%ydbAIMDERzGTIrQTVBoOhdanoKcC7(11)=6

# xref of firstname=First4 and lastname=Last2
^%ydbAIMDERzGTIrQTVBoOhdanoKcC7(-2)=2
^%ydbAIMDERzGTIrQTVBoOhdanoKcC7(-1)=4
^%ydbAIMDERzGTIrQTVBoOhdanoKcC7(11)=8

^%ydbAIMDERzGTIrQTVBoOhdanoKcC7(-2) clearly gives us the number of unique last names at each stage. And this is useful information to have.

Similarly, ^%ydbAIMDERzGTIrQTVBoOhdanoKcC7(-1) gives us the number of unique first names at each stage. And this is also useful information to have.

But ^%ydbAIMDERzGTIrQTVBoOhdanoKcC7(11) seems like useless information. It seems to be the total sum of ALL cross referenced pieces. So if there are N nodes in the ^names global, and we are cross referencing both firstname and lastname pieces, ^%ydbAIMDERzGTIrQTVBoOhdanoKcC7(11) seems to be always 2*N. This is totally derived information. And to actually compute it, we create a tstart/tcommit fence which is what caused a TRANS2BIG error in YottaDB/DBMS/YDBOcto#1083 (closed).

Octo will be temporarily fixed in this case to use stat=1 to avoid the TRANS2BIG error (instead of stat=2). But stat=2 provides more information for optimizations in Octo (e.g. YottaDB/DBMS/YDBOcto#317 and YottaDB/DBMS/YDBOcto#275). So it would be nice to remove this unnecessary code in YDBAIM thereby we can revert to Octo using stat=2 afterwards.

Initial Release Note