Question

I have a skeleton text file with placeholder strings:

blah blah blah
blah $PLACEHOLDER_1$
blah
$PLACEHOLDER_2$

and so on. Specific "form" of placeholders does not matter -- I may change them to whatever most comfortable for specific implementation.

I have a bash script where I know values for placeholders, and I need to generate a new file, with placeholders replaced with values.

#! /bin/sh
PLACEHOLDER_1 = 'string 1'
PLACEHOLDER_2 = 'multiline 
string 
2'
# TODO: Generate file output.txt from file output.template 
#       using placeholders above.

I may do this in multiple passes with sed, but it is not fun. I do not want to use Perl. I want to use textutils and bash itself only.

What is the best way to do what I want in a single pass?

Was it helpful?

Solution

You can still use sed to do the replace in a single pass. You just need to specify all the replacements in one command.

eg.

sed -i 's/PLACEHOLDER_1/string 1/g;s/PLACEHOLDER_2/string 2/g' <file>

OTHER TIPS

Here's a way to do it without sed:

First, a slightly modified template file in which the placeholders are bash variables:

blah blah blah
blah $PLACEHOLDER_1
blah
$PLACEHOLDER_2

And the script:

#! /bin/sh
templatefile=output.template
outputfile=output.txt

PLACEHOLDER_1='string 1'

PLACEHOLDER_2='multiline 
string 
2'

# DONE: Generate file output.txt from file output.template 
#       using placeholders above.

echo "$(eval "echo \"$(cat $templatefile)\"")" > $outputfile

Here's a version that demonstrates a template contained within the script, but with a twist. It also demonstrates default values, which can also be used in the template file version, plus you can do math in the template:

#! /bin/sh
template='blah blah blah
blah $PLACEHOLDER_1
blah
${PLACEHOLDER_2:-"some text"} blah ${PLACEHOLDER_3:-"some
lines
of
text"} and the total is: $((${VAL_1:-0} + ${VAL_2:-0}))'
# default operands to zero (or 1) to prevent errors due to unset variables
outputfile=output.txt

# gears spin, bells ding, values for placeholders are computed

PLACEHOLDER_1='string 1'

PLACEHOLDER_2='multiline 
string 
2'

VAL_1=2

VAL_2=4

unset PLACEHOLDER_3 # so we can trigger one of the defaults

# Generate file output.txt from variable $template 
#       using placeholders above.

echo "$(eval "echo \"$template\"")" > $outputfile

No sed, no loops, just hairy nesting and quotes. I'm pretty sure all the quoting will protect you from malicious stuff in a template file, but I'm not going to guarantee it.

Building on the previous answer, perhaps use an array and compute the sed string?

#!/bin/sh
PLACEHOLDER[0]='string 1'
PLACEHOLDER[1]='multiline 
string 
2'

s="sed -i "
for(( i=0 ; i<${#PLACEHOLDER[*]} ; i++ )) ; do 
    echo ${PLACEHOLDER[$i]}
    s=$s"s/PLACEHOLDER_$i/${PLACEHOLDER[$i]}/g;"
done
echo $s

Seems to fail on the multi-line strings, though.

I don't know how portable Bash arrays might be. Above snippet tested with "GNU bash, version 3.2.17(1)-release (i386-apple-darwin9.0)"

My bash only solution:

TEMPLATE='
foo
$var1
bar
$var2'
eval "echo \"$TEMPLATE\""

I just stumbled upon this question because I was just looking for the exact same, and I found envsubst(1).

You can use envsubst if you don't mind using environment variables:

PLACEHOLDER_1='string 1' PLACEHOLDER_2='multiline 
string 
2' envsubst < output.template 

If you have a lot of variables you can store them in a file and just source it (remember to use export at the end of the sourced file!)

Licensed under: CC-BY-SA with attribution
Not affiliated with StackOverflow
scroll top