Path: utzoo!utgpu!news-server.csri.toronto.edu!bonnie.concordia.ca!uunet!samsung!caen!ox.com!ox.com!emv From: emv@ox.com (Ed Vielmetti) Newsgroups: comp.lang.perl Subject: Re: Perl-Users Digest #595 (Re: combined split modes? and inverted indexes) Message-ID: Date: 4 Feb 91 04:19:35 GMT References: <1991Feb4.012922.7962@uvaarpa.Virginia.EDU> Sender: usenet@ox.com (Usenet News Administrator) Organization: OTA Limited Partnership, Ann Arbor MI. Lines: 46 In-Reply-To: eichin@athena.mit.edu's message of 4 Feb 91 01:29:22 GMT In article <1991Feb4.012922.7962@uvaarpa.Virginia.EDU> eichin@athena.mit.edu (Mark W. Eichin) writes: look at the latest Usenix proceedings - "A New Hashing Package for UNIX", Margo Seltzer (UCB) and Ozan Yigit (York U.) p. 173, USENIX Winter '91. It's fast and machine independent (the disk files can be read on any machine, whereas dbm indices are not portable; the files tend to be smaller, or at least have fewer "holes"; and it's typically faster than dbm and *much* faster than GNU dbm.) The new hashing package has not been released yet, as far as I know. But might do you well to pick up Oz's slightly older "sdbm" package; it should do you better than ndbm, and there are enough things tunable about it to let you experiment. A perl built with it won't read old dbm files though... Here's a pointers to sdbm Archive: decuac.dec.com:/public/sources/sdbm.shar.Z [16.67.32.1] Archive: tut.cis.ohio-state.edu:/pub/sdbm/sdbm.Z [128.146.8.60] and part of the release announcement: From: oz@nexus.YorkU.CA (Ozan Yigit) Subject: sdbm is available. Keywords: dbm,ndbm,sdbm Date: 15 Dec 90 22:02:21 GMT A complete clone of Berkeley's ndbm, called sdbm is available. It is currently ftp-able from nexus.yorku.ca [130.63.9.1] as pub/oz/sdbm.shar.Z. points of interest: + A complete functional clone of ndbm, using a fast and simple external hashing algorithm: "Dynamic Hashing" (1978) by P.-A. Larson. + sdbm is fully public domain. There is no pedantic, overbearing or infectious copyright/left to put up with. It is unconditionally yours, and everybody-else's. You can do whatever you like with it. + sdbm is faster and safer than ndbm, but shares the same shortcomings: limited size records, holes. On the other hand, ndbm needs no apologies: it has proven itself to be extremely useful within its limitations. The sdbm pagefiles are smaller than ndbm, and contain fewer holes.