Question

I have small script in bash, which is generating graphs via gnuplot.
Everything works fine until names of input files contain space(s).

Here’s what i’ve got:

INPUTFILES=("data1.txt" "data2 with spaces.txt" "data3.txt")

...

#MAXROWS is set earlier, not relevant.


for LINE in $( seq 0 $(( MAXROWS - 1 )) );do

gnuplot << EOF
reset
set terminal png
set output "out/graf_${LINE}.png"

filenames="${INPUTFILES[@]}"

set multiplot 

plot for [file in filenames] file every ::0::${LINE} using 1:2 with line title "graf_${LINE}"

unset multiplot
EOF
done

This code works, but only without spaces in names of input files.

In the example gnuplot evaluate this:

1 iteration: file=data1.txt  - CORRECT
2 iteration: file=data2  - INCORRECT
3 iteration: file=with  - INCORRECT
4 iteration: file=spaces.txt  - INCORRECT

Answer 1

The quick answer is that you can’t do exactly what you want to do. Gnuplot splits the string in an iteration on spaces and there’s no way around that (AFIK). Depending on what you want, there may be a “Work-around”. You can write a (recursive) function in gnuplot to replace a character string with another —

#S,C & R stand for STRING, CHARS and REPLACEMENT to help this be a little more legible.
replace(S,C,R)=(strstrt(S,C)) ? \
    replace( S[:strstrt(S,C)-1].R.S[strstrt(S,C)+strlen(C):] ,C,R) : S

Bonus points to anyone who can figure out how to do this without recursion…

Then your (bash) loop looks something like:

INPUTFILES_BEFORE=("data1.txt" "data2 with spaces.txt" "data3.txt")
INPUTFILES=()
#C style loop to avoid changing IFS -- Sorry SO doesn't like the #...
#This loop pre-processes files and changes spaces to '#_#'
for (( i=0; i < ${#INPUTFILES_BEFORE[@]}; i++)); do 
    FILE=${INPUTFILES_BEFORE[${i}]}
    INPUTFILES+=( "`echo ${FILE} | sed -e 's/ /#_#/g'`" ) #replace ' ' with '#_#'
done

which preprocesses your input files to add ‘#_#’ to the filenames which have spaces in them… Finally, the “complete” script:

...

INPUTFILES_BEFORE=("data1.txt" "data2 with spaces.txt" "data3.txt")
INPUTFILES=()
for (( i=0; i < ${#INPUTFILES_BEFORE[@]}; i++)); do 
    FILE=${INPUTFILES_BEFORE[${i}]}
    INPUTFILES+=( "`echo ${FILE} | sed -e 's/ /#_#/g'`" ) #replace ' ' with '#_#'
done

for LINE in $( seq 0 $(( MAXROWS - 1 )) );do
gnuplot <<EOF
filenames="${INPUTFILES[@]}"
replace(S,C,R)=(strstrt(S,C)) ? \
        replace( S[:strstrt(S,C)-1].R.S[strstrt(S,C)+strlen(C):] , C ,R) : S
#replace '#_#' with ' ' in filenames.
plot for [file in filenames] replace(file,'#_#',' ') every ::0::${LINE} using 1:2 with line title "graf_${LINE}"

EOF
done

However, I think the take-away here is that you shouldn’t use spaces in filenames 😉