Replace strings in multiple files with corresponding caps using bash on MacOSX - macos

Question

I have multiple .txt files, in which I want to replace the strings
old -> new
Old -> New
OLD -> NEW
The first step is to only replace one string Old->New. Here is my current code, but it does not do the job (the files remain unchanged). The sed line works only if I replace the variables with the actual strings.
#!/bin/bash
old_string="Old"
new_string="New"
sed -i '.bak' 's/$old_string/$new_string/g' *.txt
Also, how do I convert a string to all upper-caps and all lower-caps?
Thank you very much for your advice!

Answer 1

To complement #merlin2011's helpful answer:
If you wanted to create the case variants dynamically, try this:
# Define search and replacement strings
# as all-lowercase.
old_string='old'
new_string='new'
# Loop 3 times and create the case variants dynamically.
# Build up a _single_ sed command that performs all 3
# replacements.
sedCmd=
for (( i = 1; i <= 3; i++ )); do
case $i in
1) # as defined (all-lowercase)
old_string_variant=$old_string
new_string_variant=$new_string
;;
2) # initial capital
old_string_variant="$(tr '[:lower:]' '[:upper:]' <<<"${old_string:0:1}")${old_string:1}"
new_string_variant="$(tr '[:lower:]' '[:upper:]' <<<"${new_string:0:1}")${new_string:1}"
;;
3) # all-uppercase
old_string_variant=$(tr '[:lower:]' '[:upper:]' <<<"$old_string")
new_string_variant=$(tr '[:lower:]' '[:upper:]' <<<"$new_string")
;;
esac
# Append to the sed command string. Note the use of _double_ quotes
# to ensure that variable references are expanded.
sedCmd+="s/$old_string_variant/$new_string_variant/g; "
done
# Finally, invoke sed.
sed -i '.bak' "$sedCmd" *.txt
Note that bash 4 supports case conversions directly (as part of parameter expansion), but OS X, as of 10.9.3, is still on bash 3.2.51.
Alternative solution, using awk to create the case variants and synthesize the sed command:
Aside from being shorter, it is also more robust, because it also handles strings correctly that happen to contain characters that are regex metacharacters (characters with special meaning in an regular expression, e.g., *) or have special meaning in sed's s function's replacement-string parameter (e.g., \), through appropriate escaping; without escaping, the sed command would not work as expected.
Caveat: Doesn't support strings with embedded \n chars. (though that could be fixed, too).
# Define search and replacement strings as all-lowercase literals.
old_string='old'
new_string='new'
# Synthesize the sed command string, utilizing awk and its tolower() and toupper()
# functions to create the case variants.
# Note the need to escape \ chars to prevent awk from interpreting them.
sedCmd=$(awk \
-v old_string="${old_string//\\/\\\\}" \
-v new_string="${new_string//\\/\\\\}" \
'BEGIN {
printf "s/%s/%s/g; s/%s/%s/g; s/%s/%s/g",
old_string, new_string,
toupper(substr(old_string,1,1)) substr(old_string,2), toupper(substr(new_string,1,1)) substr(new_string,2),
toupper(old_string), toupper(new_string)
}')
# Invoke sed with the synthesized command.
# The inner sed command ensures that all regex metacharacters in the strings
# are escaped so that sed treats them as literals.
sed -i '.bak' "$(sed 's#[][(){}^$.*?+\]#\\&#g' <<<"$sedCmd")" *.txt

Answer 2

If you want to do bash variable expansion inside the argument to sed, you need to use double quotes " instead of single quotes '.
sed -i '.bak' "s/$old_string/$new_string/g" *.txt
In terms of getting matches on all three of the literal substitutions, the cleanest solution may be just to run sed three times in a loop like this.
declare -a olds=(old Old OLD)
declare -a news=(new New NEW)
for i in `seq 0 2`; do
sed -i "s/${olds[$i]}/${news[$i]}/g" *.txt
done;
Update: The solution above works on Linux, but apparently OS X has different requirements. Additionally, as #mklement0 mentioned, my for loop is silly. Here is an improved version for OS X.
declare -a olds=(old Old OLD)
declare -a news=(new New NEW)
for (( i = 0; i < ${#olds[#]}; i++ )); do
sed -i '.bak' "s/${olds[$i]}/${news[$i]}/g" *.txt
done;

Answer 3

Assuming each string is separated by spaces from your other strings and that you don't want partial matches within longer strings and that you don't care about preserving white space on output and assuming that if an "old" string matches on a "new" string after a previous conversion operation, then the string should be changed again:
$ cat tst.awk
BEGIN {
split(tolower(old),oldStrs)
split(tolower(new),newStrs)
}
{
for (fldNr=1; fldNr<=NF; fldNr++) {
for (stringNr=1; stringNr in oldStrs; stringNr++) {
oldStr = oldStrs[stringNr]
if (tolower($fldNr) == oldStr) {
newStr = newStrs[stringNr]
split(newStr,newChars,"")
split($fldNr,fldChars,"")
$fldNr = ""
for (charNr=1; charNr in fldChars; charNr++) {
fldChar = fldChars[charNr]
newChar = newChars[charNr]
$fldNr = $fldNr ( fldChar ~ /[[:lower:]]/ ?
newChar : toupper(newChar) )
}
}
}
}
print
}
.
$ cat file
The old Old OLD smOLDering QuICk brown FoX jumped
$ awk -v old="old" -v new="new" -f tst.awk file
The new New NEW smOLDering QuICk brown FoX jumped
Note that the "old" in "smOLDering" did not get changed. Is that desirable?
$ awk -v old="QUIck Fox" -v new="raBid DOG" -f tst.awk file
The old Old OLD smOLDering RaBId brown DoG jumped
$ awk -v old="THE brown Jumped" -v new="FEW dingy TuRnEd" -f tst.awk file
Few old Old OLD smOLDering QuICk dingy FoX turned
Think about whether or not this is your expected output:
$ awk -v old="old new" -v new="new yes" -f tst.awk file
The yes Yes YES smOLDering QuICk brown FoX jumped
A few lines of sample input and expected output in the question would be useful to avoid all the guessing and assumptions.