updated kv to support inverted tables. https://github.com/Pascal-J/kv/blob/main/kv.ijs
output for these functions are near bottom of https://raw.githubusercontent.com/Pascal-J/kv/main/README.md 'Id Name Job Status' kv ,.&.:>"1 |: maybenum each > ','cut each cutLF 0 : 0 NB. from your example 3,Jerry,Unemployed,Married 6,Jan,CEO,Married 5,Frieda,student,Single 1,Alex,Waiter,Separated ) A one line dsl version: itdsl kvdsL 'Id Name Job Status `3 6 5 1 ; Jerry Jan Frieda Alex ;Unemployed:CEO:student:Waiter; Married Married Single Separated' NB. note extra garbage spaces In a dictionary, inverted tables are stored as key(fieldname), and associated boxed table (field value that can be longer than 1 char per value). I have an inverted table display (keys as column headers) too. That can optionally filter just the fields you wish A query that retrieves records where Job is 'CEO' or 'student' then displays in IT format itdisp ('CEO;student' padstrmatch 'Job' kvget ]) kvQ it_kvtest_ or partial 3 column display without the special ; parsing 'Id Name Job' itdisp1 (('CEO';'student') padstrmatch 'Job' kvget ]) kvQ it kvQ is a "query engine" (too simple to call it that) that takes a boolean function to filter records "on set of properties/fields" Basically an inverted table as a dictionary is a dictionary that only contains the properties corresponding to fields. Because kv can store other dictionaries, if you have additional information to store, it can be in "peer" keys to your IT dictionary. On Sunday, February 13, 2022, 08:49:02 a.m. EST, Ric Sherlock <tikk...@gmail.com> wrote: Inspired by recent threads, I've started experimenting with a DataFrame structure in J. I began by building off the 'general/misc/inverted' utilities so that a DataFrame is just a 2-row table, where the first row is a list of labels & the 2nd is an inverted table. It can be installed as 'tables/dataframe' from the following github repository: https://github.com/tikkanz/jdataframe The JArrow bindings that Aaron pointed me at a while back represent a similar data structure as a 2-column table where the labels are in the first column and the "columns" are rows in the 2nd column. (https://github.com/tikkanz/JArrow) I think I need to play with both for a while to see which works better. Obviously there is a lot more to do, but I thought I'd share in case anyone has any thoughts, suggestions or wants to help take the idea further. Ric ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
