[SOLVED] BASH Arrays

dsartain · 05-07-2012, 09:56 AM

I'm running BASH version 3.2.25(1)-release

I need to be able to read a line of user input and store that into an array. Currently, what I have technically works, but not the way I want it to.

Code:

 echo "How many documents do you want to create?"
        read COUNT

        echo "Enter the names of the documents you wish to build surrounded by single quotes, separated by spaces."
        echo "See the following format: '01 Financial Statement ct' '02 Insurances ct' '03 Tax Returns ct'"


read -ea DOCUMENTS

        echo "${DOCUMENTS[0]}"
        echo "${DOCUMENTS[1]}"

THE PROBLEM:

1) Even when I surround what I'd like each element of the array to be by single quotes, I still have to use a backslash to escape the space between words within the single quotes. And then when I view the output, it shows the single quotes at the beginning and end, which is not desirable.

INPUT 1:

Code:

'01\ Financial\ Statement\ ct' '02\ Insurances\ ct'

OUTPUT 1:

Code:

'01 Financial Statement ct'
'02 Insurances ct'

Problem 2
In attempts to not have the single quotes show up, I didn't use them and only escapes backslashes between spaces of the same elements. This output is better, but still not practical as I'm sure other users will put slashes in the wrong place, etc.

INPUT 2:

Code:

01\ Financial\ Statement\ ct 02\ Insurances\ ct

OUTPUT 2:

Code:

01 Financial Statement ct
02 Insurances ct

Is there a way to separate array elements by using a comma instead, negating the need for backslashes? Or a way to use the single quotes to isolate the elements from each other, without requiring the backslashes as well??

catkin · 05-07-2012, 10:40 AM

Maybe read -ead',' DOCUMENTS

dsartain · 05-07-2012, 11:05 AM

Changed to

Code:

read -ead',' DOCUMENTS

Resulted in: ./test.sh: line 20: read: `d,': not a valid identifier

grail · 05-07-2012, 11:41 AM

How about:

Code:

mapfile -t DOCUMENTS

No quotes required, but must advise user to press Ctrl+D after last line and that each record to be on its own line

Code:

01 Financial Statement ct
02 Insurances ct
<Ctrl+D here>

dsartain · 05-07-2012, 12:15 PM

That looks like a great option, but I'm running BASH 3.2.25. Mapfile wasn't introduced until BASH 4

David the H. · 05-07-2012, 12:23 PM

Yes, mapfile would've been a good choice.

But really, the big problem here is that a single iteration of read only works on a single "line" of input (as delimited by the "-d" option). Then to split the line into an array, you have to change the IFS variable to define the field breaks.

More properly, to capture multiple lines like this you should use a loop.

And it's no use adding quotes to the input. Once the text is stored in the variable, the shell parsing order ensures that they will always be treated as literal. You'd have to parse the line manually in order to handle them.

I think what I would do is something more like this:

Code:

echo "Enter the names of the documents you wish to build."
echo "Press 'enter' after each entry.  When finished, enter the word 'done'."

n=0
while true; do

	read -e -p 'Enter a document name or "done": ' name

	case $name in

		[Dd][oO][nN][eE]) break ;;

		*) documents[n++]=$name ;;

	esac

done

echo "${documents[@]}"

This also adds another benefit in that you can also add tests to validate each entry by other criteria (e.g. file exists, or matches a certain pattern) before continuing.

P.S. Environment variables are generally all upper-case. So while not critical, it's good practice to keep your own user variables in lower-case or mixed-case, to help differentiate them.

David the H. · 05-07-2012, 12:33 PM

By the way, for a simple comma-delimited version, change the IFS setting like I mentioned above:

Code:

echo "Enter the names of the documents you wish to build, separated by commas."

IFS=',' read -e -a documents

The -e option tells bash to use the readline library, btw. This makes filename tab completion available.

grail · 05-07-2012, 12:34 PM

@David - I think you mixed up the variable being assigned to the array ... name vs line

David the H. · 05-07-2012, 12:41 PM

Ah thanks. Minor mistake there. I wanted to change the variable name to something more appropriate, but forgot to change every entry.

It's fixed now, and I've also added a bit of extra info.

dsartain · 05-07-2012, 01:08 PM

Ok, the IFS change is going to work best for what I need. Here's the updated code as of now. Thanks to everyone for the help so far.

Code:

      1         echo "How many documents do you want to create?"
      2         read COUNT
      3
      4         echo "Enter the names of the documents you wish to build surrounded by single quotes, separated by spaces."
      5         echo "See the following format: '01 Financial Statement ct' '02 Insurances ct' '03 Tax Returns ct'"
      6
      7         OIFS=$IFS #Saves original field separator
      8         IFS=',' read -ea DOCUMENTS
      9
     10         for i in {0..$((COUNT-1))}
     11         do
     12                 echo "${DOCUMENTS[$i]}"
     13         done
     14
     15         IFS=$OIFS #Restores original field separator

The problem I'm running into now is on line 12.

Input:

Code:

01 Financial Statement ct, 02 Insurances ct, 03 Tax Returns ct

Error:

Code:

./test.sh: line 12: {0..2}: syntax error: operand expected (error token is "{0..2}")

I'm not sure what's causing the problem here, because when I change the $((COUNT-1)) in the for loop to a static number, it works just fine...

Any thoughts?

EDIT:
Scratch that. The issue was brace expansion preceeding variable expansion.

New line for the for loop control:

Code:

for i in $(eval echo {0..$((COUNT-1))})

Consider this one solved, guys. Thanks!

grail · 05-07-2012, 01:31 PM

Please don't use eval when something easier can be used (here are some reasons not use eval).
Just use a standard for loop construct:

Code:

for (( i = 0; i <= COUNT - 1; i++ ))

David the H. · 05-07-2012, 01:34 PM

Oh for goodness sakes, no. Never use eval for something like this. That's like using a howitzer to hunt mosquitoes.

There are several (much) more appropriate ways to handle looping through arrays.

Code:

for i in "${DOCUMENTS[@]}"; do
	echo "$i"
done

for i in "${!DOCUMENTS[@]}"; do
	echo "${DOCUMENTS[i]}"
done

for (( i=0; i<=COUNT; i++ )); do
	echo "${DOCUMENTS[i]}"
done


printf '%s\n' "${DOCUMENTS[@]}"

How can I use array variables?
http://mywiki.wooledge.org/BashFAQ/005

dsartain · 05-07-2012, 01:35 PM

That works for me. That's the syntax I'm used to anyway, I just didn't see it pop up when I searched for for loops in bash. #bashnewbie

grail · 05-07-2012, 02:06 PM

Sorry if David and I came across a little harsh ... not trying to pick on you as we are aware that you were new. Just trying to help you move away from starting a bad habit

Nominal Animal · 05-07-2012, 02:19 PM

From an user interface perspective, instead of asking how many documents the user intends to specify up front, you could just accept new document names from the user, until they supply an empty line:

Code:

DOCUMENTS=()
while read -p "Document $[${#DOCUMENTS[@]}+1]: " DOCUMENT ; do

    # Break the loop if DOCUMENT is empty
    [ -n "$DOCUMENT" ] || break

    # Append DOCUMENT to DOCUMENTS array.
    DOCUMENTS[${#DOCUMENTS[@]}]="$DOCUMENT"
done

printf 'You specified %d documents:' ${#DOCUMENTS[@]}
printf '\t\047%s\047\n' "${DOCUMENTS[@]}"