3 l-MAR-75 09:48: 14,6278 Date: 31 MAR 1975 0948-PST From: LEDERBERG Subject: FIRST BLOCK OF COMMENTS ON CONGEN ORGANIZATION AND DOCUMENTATIO? - - - - PRUNE should not eliminate, but rather segregate a set of graphs into the "accepted" /"rejected" categories. The defaukt selection for further processing would be the 1Ast accepted set; but the others would still be available unless cleared by the user. + 1; ECL-DUMP.SET; 1 Fri 22-Sep-78 8:09PM > The User should also have the option of requesting a reverse- or de-IMBED function (when the constraint is a BL or GL item, the program has already done almost all the necessary work. A new command -- clear up redundancies -- may then be needed, to be sure. In complex structures, it may not be so easy for the user to recogbize a superatom. Each new aggregate should have a name to be used as a handle perhaps == an external file name; As well as names for single structures in an aggregate CONSTRAINTS might well include some graph-oriented criteria. Besides: NON-Planar already mentioned, perha@s also SINGLY-connected or with specific refrence to a pArticular **re (if l-connected) or even a singly-connected composition. The ms-ist might even want to specify a l-connected low-res mol. wt. as something he can infer from his data. PAGE 411:78 SUBstructu GRAPHS I have been running some tests on CONGEN by way of building n12 and ? n14 vertex graphs. What is the generator algorithm now implemented ? Do you have some #SAVEd files of such graphs? DETACH: I did try (GCGAG NIL) and detach...continue appatently successfully on one of these biggish problems. INTERRUPT Is it possible to break during generation And resume? 2 IS THERE ANY EASY WAY TO GET THE EQUIVALENT OF SYSTAT FOR JUST ONES OWN JOBS --E.G. TO MONITOR A DETACHED ONE? WHO WASTES TIME PRINTIMG THE WHOLE USER SET. (Not for implementation) A complete structure-manipulating language, on analogy of TECO would allow for the user construction ofmacros from the primitive commands. Not too virtuous here. But I hAd in mind the analogy of a "line" in a text editor, to a structure U/e superatom in CONGEN. For example, in ES, a command should allow the recopying of a given structure n times, or other simple sequences of operations. It wouldn't hurt if the general format of the ES interaction more generally resembled that of a polished text-edit language like TECO or TV-EDIT. Is that a pipe-dream? CONGEN.DOC + COMMENTS on the documentation 1; ECL-DUMP.SET; 1 Fri 22-Sep-78 8:09PM PAGE 411:79 p.6:2 ALL is not that clearly explained. One soon discovers that it means the other commands will be prompted. p.7:3 para. "bond multiplicities are taken literally" - should be more carefully explained. I assume it means that a = is regarded as an additive conjunction of two -. p.7:5 ATNAME This is the offending statement that Polynames... should not be used for superatoms. By implication, for superatoms to be used in composition lists. ES does Also refer to superatoms. p.7:1 last line/ "no structures can possess" -- would be clearer as "no structures are permitted to contain". U "can" may be ambiguous. Indeed "may possess" perhaps already suffices.e p.7:10 HINTS --par. 3 I hAven't tried recursion What happens? p.7:12 2) Explain significance of the query for intra-superatom bonds How does CONGEN compare with PROPHET (what does latter do that may be similar)? 3 CONGEN will not handle H2 correctly. (Nor have most versions of DENDRAL. But do you understand just why?) Would it be possible, during ES, to get disconnected graphs ADRAWn or NDRAWn across the page instead of vertically? A small point, but it is rather frustrating tryibg to keep track of these complex structures while ESing them from a CRT terminal., and they are rolling by. B lREAKPOINT These computations are so long that we really do have to take account of that in engineering. There should be a breakpoint facility thatcan do autosaves of the current state of the generation perhaps every n garbage-collects, and which can be interrogated by the user to see what is happending. I realize all of these features take additional time and space; but much more will be wasted in tracking mysterious faults. All of this obviously will require further attention to comoact representations. Now that I have A better feel for the operation, I would very much like to dig into more of the details with you, viz-a-viz possible improvements in ti,e-efficacy. Perhaps we need some more heuristic approaches to problems like solving the group; as well as deeper study of dictionary/compute-time tradeoffs. STEREOISOMERS I pervesevere that there are really exciting possibilities here. Perhaps cis-trans geometric isomerism around X=C< is the simpler caseto analyze and display. + 1; ECL-DUMP.SET; 1 Fri 22-Sep-78 8:09PM PAGE 411:80 DISPLAY. I really do favor more open formats. Look as what happens to RING 8+ in the ES ADRAW. ----- Comments on memo 26-Mar Much of above was composed before remarks about TCONGEN; which I have yet to try, but soon will. I was glad to note things like RAISE and the d effect. Yes, doubtlessthere should be an explicit RESET command. 4 If we could write macros like do n drawsome (n) if tty: = then jump . . . . . or == in editor-like language, we would not have to make rigid decisions. But I think the SCAN alternative suggested is in the right ball park. If SCAN pauses for input at each DRAWing, you don't need SCANSOME -- that is one of the options available at the pause, i.e., what to go to next. EXTENSIONS I forsee that one application of CONGEN will be to filter a file of known strucrtures Ue.g. in interactive lookupe -- and this might be on quite a large scAle. That's why I raised questions about merging files, etc. UCall previous marks about analogy with a text-edit languagee. Perhaps we should again review what CA, IS1 ARE doing in this field. UAnd IS1 --surely not CA -- and perhaps also NLM might be interested in suppor ting further work to make CONGEN more efficient if it could serve in that kind of intelligent retrieval mide as well.]]e