Hokey (The Moderator) <hokey@plus5.uucp> (03/10/85)
std-mumps Digest Sat, 9 Feb 85 Volume 1 : Issue 10 Today's Topics: Exotic Language of the Month Club fun with setpieces (2 msgs) ---------------------------------------------------------------------- Date: 9 Mar 85 01:38:21 CST (Sat) From: hokey@plus5.uucp Subject: Exotic Language of the Month Club To: std-mumps Mumps was selected as the "Exotic Language of the Month Club" feature in the Feb. '85 issue of "Computer Language". The article was written by J. Edward Volkstorf Jr. ------------------------------ Date: Thu, 7 Mar 85 17:19:40 est From: ihnp4!watmath!wateng!jmmussi (Jose Mauro F. Mussi) Subject: fun with setpieces To: jmmussi, watmath!ihnp4!wucs!plus5!hokey From: hokey@plus5.uucp (Hokey) > According to the Standard, the behavior is: > > $P(glvn,d,m,n)=expr > > 1) scan the glvn and evaluate subscripts. (doesn't affect naked) ^^^^^^^ even if the subscripts are naked references? > 2) evaluate d, m, n (can affect naked) > 3) evaluate the expr to the right of the = (can affect naked) > 4) bounds check m and n > 5) evaluate glvn ("resolve" naked glvn) > In my opinion, the use of naked references should be very restricted. This construction was created to improve file access performance by reducing the number of hierarchical levels of a global variable that must be search to find a specific node. However, this improvement was only possible due to the nature of the internal file structure in earlier mumps systems. I imagine that all current implementations must use other tree structures (B* or AVL) for files where all subscripts are needed to produce the access key. In this case, naked references have only been kept around to provide backward compatibility. They do not improve performance and in most cases there is actually a loss of performance due to the concatenation of the subscripts in the naked reference with the subscripts of the last global variable reference (except the last subscript of course). I would like to receive comments on the correctness of the statements above and also would like to know the opinion of other mumps programmers about the "convenience" of using naked references. Jose Mauro Fialho Mussi Dept. of Electrical Engineering University of Waterloo Waterloo, Ont., N2L 3G1 Canada ------------------------------ Date: 9 Mar 85 01:33:39 CST (Sat) From: hokey@plus5.uucp Subject: fun with setpieces To: std-mumps Jose, you caught me. I was not explicit enough. If the glvn is a gvn, it does not affect the naked indicator when it is initially scanned. Any nakeds or globals present as subscripts within the glvn *do* affect the naked indicator. I shouldn't have been so brief. ------------------------------ End of std-mumps Digest ****************************** -- Hokey ..ihnp4!plus5!hokey 314-725-9492