I. Basic Operations The distribution contains a plain text datafile of US postal codes in ./data/zipcodes: $ head ./data/zipcodes # Zip Code : Location # $NetBSD: zipcodes,v 1.6 2003/05/11 01:55:18 wiz Exp $ # @(#)zipcodes 8.1 (Berkeley) 6/8/93 00401:Pleasantville, NY 00501:Holtsville, NY 00544:Holtsville, NY 00601:Adjuntas, PR 00602:Aguada, PR 00603:Aguadilla, PR 00604:Aguadilla, PR Inspection of the file shows that it is directly parseable by hdb. Just use the "getline" input format specifying a `:' (colon) delimiter between key and value (this format also ignores the commented lines beginning with `#'): $ hdb make -c "US Zippers" -d: zipcodes.hdb <./data/zipcodes The -c option is used to add some useless descriptive comment into the datafile. In this kind of database, it wouldn't do at all if duplicate keys existed. Each zipcode should appear once and only once, or who knows where your mail could end up. To check for a unique key constraint on the database, use the "dupes" operation: $ hdb dupes -q zipcodes.hdb && echo "no dupes" no dupes Now perform a key lookup on the postal code "90210": $ hdb get 90210 zipcodes.hdb Beverly Hills, CA Wasn't that some tv series? Now search all record values beginning with "bone", case insensitive (the output shown is in the default format): $ hdb match -i "bone*" zipcodes.hdb 5:13 30806:Boneville, GA 5:13 57317:Bonesteel, SD 5:12 62815:Bone Gap, IL Nuts, I was trying to get a Boner. Repeat the match query, this time with an equivalent regular expression and the grep command. Note the explicit anchoring in this regular expression compared to the implicit anchoring in the wildcard expresson above. Pipeline the results directly into another hdb: $ hdb grep -i "^bone" zibcodes.hdb | hdb make -c "Dem Bones" bonez.hdb Check the new hdb file with a dump (with the -y option, the output format is in cdb "legacy" format): $ hdb dump -y bonez.hdb +5,13:30806->Boneville, GA +5,13:57317->Bonesteel, SD +5,12:62815->Bone Gap, IL Nope, still no Boner. II. Set Operations Everyone knows the zipcode for "The Dalles", Oregon, is 97058, right? $ hdb get 97058 <zipcodes.hdb The Dalles, OR Aside: what kind of town starts its name with "The"? Anyway, Google has one of their server farms located in The Dalles, sucking massive amounts of hydroelectric energy off the mighty Columbia river nearby. Let's suppose the town gets hit by a Google-Bomb, and is competely obliterated in a puff of virtual smoke. Sadly, then, it becomes necessary to remove The Dalles zipcode from the database. One could open and edit the original plaintext source file, then rebuild the hdb from the edited file. Or one could use hdb set operations directly on the command line as follows: $ echo "97058:" | hdb make -d: delete.hdb && \ hdb purge zipcodes_new.hdb zipcodes.hdb delete.hdb The "echo" command here is used simply to pipe in a single record with key "97058" (and empty value) that hdb make will compile into delete.hdb. Then zipcodes.hdb and delete.hdb are combined with the hdb purge command. The result of the "purge" is zipcodes_new.hdb, containing all the original records from zipcodes.hdb, except those with matching keys found in delete.hdb. For this example, then, the "97058" record will be deleted from zipcodes_new.hdb: $ hdb get 97058 zipcodes_new.hdb || echo "not found" not found The Dalles is history. The result file argument of a set operation command can safely name either of the operand files. The result file will then replace the operand file: $ echo "97058:" | hdb make -d: delete.hdb && \ hdb purge zipcodes.hdb zipcodes.hdb delete.hdb This example is the same as above, except that now the record has, effectively, been deleted from the zipcodes.hdb file directly: $ hdb get 97058 zipcodes.hdb || echo "not found" not found Everyone knows Montana is a big state, and seems to attract more than its fair share of whackjobs. For a time, even Frank Zappa settled the