Here is an example of a small report I threw together this evening. I decided to mix shell script and J because sometimes those are the right tools for the job.
I should note a discrepancy between http://www.jsoftware.com/help/jforc/input_and_output.htm and the addons I had installed: I had to use require 'format/printf' intead of require 'printf'. (I find it easier to locate something like printf from the command line. Something like find /usr/share/j -iname printf.ijs just rolls off my fingers - other people may have different sets of habits.) Anyways, here's a slightly cleaned up version of the script (I removed some directory names prefixing the file names and changed some file name spellings to make them easier to read, and removed a bit of noise code which probably would not interest this list.) #!/bin/sh # 0 :0 cd fgrep -vf done.list calendar.list >todo.list exec ijconsole "$0" "$@" ) 9!:29]1[9!:27'2!:55]1' NB. exit on error require 'format/printf' total=:_".(_4 {:: <;._2@,&'/');._2 fread 'calendar.list' done=:_".(_3 {:: <;._2@,&'/');._2 fread 'done.list' todo=:_".(_4 {:: <;._2@,&'/');._2 fread 'todo.list' uy=:~.total mat=: uy,.<:(#/.~>:uy,total),.(#/.~>:uy,todo),.(#/.~>:uy,done) mat=: mat,.100*(3{"1 mat)%1{"1 mat echo 'Year Total ToDo Done Pct Done' '%4d %4d %4d %4d %2.2g%%' printf mat exit 0 In my opinion, printf is a bit easier to use than 8!:x for something like this. Note also that when I was writing this code I had two windows open - one to test the script, the other to edit it in. Also, I commented out the exit 0 line and the 9!:29 line, and I had inserted set -vex in as the top line of the shell script (after the #0 :0 line, of course). In case anyone hasn't seen me presenting this idea before, what happens is /bin/sh runs the script (that's the standard unix command line processor) and it does a few things before turning control over to J. And then J runs the same script. What this gives me is a file that is small and easily describable (at least in terms of purpose - you run this file and you get a report - much easier to describe than "you run that file which is just a setup for this other file"). The result of running this looks like: Year Total ToDo Done Pct Done 2012 1464 1150 314 21.45% 2011 1163 950 213 18.31% 2010 1095 915 180 16.44% 2009 1096 841 255 23.27% 2008 703 411 292 41.54% 2013 730 288 442 60.55% 2007 186 1 185 99.46% And, yes, I ran it a couple times to adjust the position of the column headers over the numbers. That aspect was pure trial-and-error. Anyways, I like that these two parts of the script are glued together. Of course, another way to do this would be to run fgrep from inside J. I don't think we have an emulation of fgrep -vf though, that performs very well. But maybe I'm overlooking something? Here's a general description of what I was looking for there: given a list of boxed substrings X (e.g. 'ing';'cked';'lts') and another list of boxed strings Y, return Y with all boxes removed where a box in X contained a substring of box in Y. Also, for reference, X would have contained 6437 boxes (and something like 423000 characters split across those boxes) while Y would have contained 1885 boxes (and just under 29000 characters split across those boxes). The result would have had 4556 boxes (and a total of just under 300k characters). Also in this particular case each substring from X appeared exactly once in Y and was always in the middle of a string in Y (non-empty prefix and suffix). Anyways, fgrep -vf was a lot faster to type than even that description. Thanks, -- Raul ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
