Datadriven shell programming? Suggestions, please

Question

I'd like to convert my shell script to use a datadriven approach. But since there are no "tables" type values in any shell (that I know of), what are suggested alternative ways to do this?

What I'm looking for is solutions that would allow one to do things like:

animals = [['horse', 300 'brown'],
           ['cat', 3, 'black'],
           ['elephant', 3000, 'grey'],
           ['mouse', 0.3, 'grey']]
for a in animals ; do
  echo "A typical $a[1] is $a[3] and weighs about $a[2] kilograms."
done

More precisely, I'd like to try a number of commands and see if one of them is available, and then send arguments to it:

commands = [['c1', '-m', '-i'],
            ['c2', '-message', '-icon'],
            ['c3', '/m', '/i']]
for c in commands ; do
  if exists c[1] ; then
    command = c[1]
    message_option = c[2]
    icon_option = c[3]
    break;
  fi
done
$command $message_option "message" $icon_option icon

Why not put your 'table' data into a file and then read from it using `awk` or similar programs? — Chem-man17, Nov 24 '16 at 09:59
I was considering `python` but decided against it for portability issues. So my approach was `bash`. The script needs to do a lot of other things, but it would be interesting to explore a combination with `awk` since that is also available "everywhere". So how would a shell script call `awk` to do the example (which I'll update to pinpoint more precisely what I need to do)? — thoni56, Nov 24 '16 at 10:09
You don't need `awk`, can do with a simple `bash` built in arrays for this. — Inian, Nov 24 '16 at 10:11
`awk '{print "A typical",$1,"is", $3, "and weighs about" ,$2 "kilos."}' inputfile` — P...., Nov 24 '16 at 10:14
@thoni56: Refer my answer below and let me know if you were looking for a similar logic. — Inian, Nov 24 '16 at 10:17

Jens · Accepted Answer · 2016-11-24T10:35:51.033

4

No bashisms needed. This can be neatly solved with a here-document read by a while read loop:

#!/bin/sh
while read animal weigth colour; do
  printf '%s\n' "A typical $animal is $colour and weighs about $weight kilos."
done << EOF
horse 300 brown
cat 3 black
elephant 3000 grey
mouse 0.3 grey
EOF

Observe how this refers to the elements by name, not cryptic indexing with 1, 2, 3. It doesn't fork/exec any external commands and beats the 3 awks in the loop body seen in another answer hands down.

edited Nov 24 '16 at 10:35

answered Nov 24 '16 at 10:29

Jens

69,818
15
125
179

2

@Inian and Jens having the same profil picture is pretty confusing lol – Nov 24 '16 at 10:46
2

@Mayerz And we're not the only ones :-) – Jens Nov 24 '16 at 10:50
@Jens: Provided an approach using arrays only because the OP's post looked similar to that. – Inian Nov 24 '16 at 12:32

Inian · Answer 2 · 2016-11-24T10:22:20.460

2

You can define and use associative-arrays in bash for your requirement.

#!/bin/bash

# declaring the Associative array
declare -a animals

animals[0]="'horse':300:'brown'"
animals[1]="'cat':3:'black'"
animals[2]="'elephant':3000:'grey'"
animals[3]="'mouse':0.3:'grey'"

for animal in "${animals[@]}"
do
    myArray=(${animal//:/ })
    printf "A typical "${myArray[0]}" is "${myArray[2]}" and weighs about "${myArray[1]}" kilograms.\n"
done

The only tricky part in the above is parameter-expansion of type

${parameter/pattern/string}
          The pattern is expanded to produce a pattern just as in filename expansion. Parameter is expanded and the longest match of pattern against its value is replaced with string. 
If pattern begins with ‘/’, all matches of pattern are replaced with string. Normally only the first match is replaced.

So ideally the string 'horse':300:'brown' is split as individual elements and stored in the array myArray which is later used to access individual elements in a C-style loop.

edited Nov 24 '16 at 10:22

answered Nov 24 '16 at 10:16

Inian

80,270
14
142
161

@thoni56: And I thought you need an approach using arrays from your example! But now being un-accepted? :( – Inian Nov 24 '16 at 12:31
1

I wasn't specifically asking for how to implement arrays, but data-driven shell scripting. I accepted your answer because it was definitely the best at the time, and I'm still up-voting you answer for teaching me about bash arrays, but the answer from @Jens was an even better solution to my actual problem. Consider the acceptance of his answer a signal that his understanding of what I wanted to achieve was even better than yours. – thoni56 Nov 24 '16 at 12:39
@thoni56: fair enough! Cheers – Inian Nov 24 '16 at 12:40

Chem-man17 · Answer 3 · 2016-11-24T10:10:56.390

0

Try this?

cat file
'horse', 300, 'brown'
'cat', 3, 'black'
'elephant', 3000, 'grey'
'mouse', 0.3, 'grey'

for i in {1..4} ; 
do 
    animal=$(awk -F, -v var="$i" 'NR== var {print $1}' file) 
    weight=$(awk -F, -v var="$i" 'NR== var {print $2}' file) 
    colour=$(awk -F, -v var="$i" 'NR== var {print $3}' file) 
    echo "A typical "$animal" is "$colour" and weighs about "$weight" kilos." 
done

Output-

A typical 'horse' is 'brown' and weighs about  300 kilos.
A typical 'cat' is  'black' and weighs about  3 kilos.
A typical 'elephant' is  'grey' and weighs about  3000 kilos.
A typical 'mouse' is  'grey' and weighs about  0.3 kilos.

edited Nov 24 '16 at 10:10

answered Nov 24 '16 at 10:04

Chem-man17

1,700
1
12
27

This is a waste of CPU cycles. It forks 12 processes and reads the same file 12 times. – Jens Nov 24 '16 at 10:39
Performance was not one of the requirements, but that might be valid observation in some circumstances. – thoni56 Nov 24 '16 at 13:12

Datadriven shell programming? Suggestions, please

3 Answers3