Find and replace string and print file directory on change - shell

Question

I am using find and sed to replace a string in multiple files. Here is my script:
find ./ -type f -name "*.html" -maxdepth 1 -exec sed -i '' "s/${REPLACE_STRING}/${STRING}/g" {} \; -print
The -print always prints the file no matter if something was changed or not. What I would like to see what files are changed. Ideally I would like the output to be something like this(as the files are changing):
/path/to/file was changed
- REPLACE STRING line 9 was changed
- REPLACE STRING line 12 was changed
- REPLACE STRING line 26 was changed
/path/to/file2 was changed
- REPLACE STRING line 1 was changed
- REPLACE STRING line 6 was changed
- REPLACE STRING line 36 was changed
Is there anyway of doing something like this?

Answer 1

Cool idea. I think -print is a deadend for the reason you mention, so it needs to be done in the exec. I think sed is also a deadend due to the challenge of printing to STDOUT as well as modifying the file. So a natural extension is to wrap some Perl around it.
What if this was your exec statement:
perl -p -i -e '$i=1 if not defined($i); print STDOUT "$ARGV, line $i: $_" if s/REPLACE_STRING/STRING/; $i++' {} \;
-p wraps the Perl statements in a standard while(<>) loop so the file is processed line by line just like sed.
-i does in-place replacement, just like sed.
-e means execute the following Perl statements.
if not defined is a sneaky way of initialising a line count variable, even though it's executed for every line.
STDOUT tells print to output to the console instead of the file.
$ARGV is the current filename, when reading from <>.
$_ is the line being processed.
if means the print only gets executed if a match is found.
For an input file text.txt containing:
line 1
token 2
line 3
token 4
line 5
The statement perl -p -i -e '$i=1 if not defined($i); print STDOUT "$ARGV, line $i: $_" if s/token/sub/; $i++' text.txt gives me:
text.txt, line 2: sub 2
text.txt, line 4: sub 4
Leaving text.txt containing:
line 1
sub 2
line 3
sub 4
line 5
So you don't get your introductory "file was changed" line, but for a one-liner I think it's a pretty good compromise.
Operating on a couple of files it looks like this:
find ./ -type f -name "*.txt" -maxdepth 1 -exec perl -p -i -e '$i=1 if not defined($i); print STDOUT "$ARGV, line $i: $_" if s/token/sub/; $i++' {} \;
.//text1.txt, line 2: sub 2
.//text1.txt, line 4: sub 4
.//text2.txt, line 1: sub 1
.//text2.txt, line 3: sub 3
.//text2.txt, line 5: sub 5

Answer 2

You could chain -exec actions and take advantage of the exit status. For example:
find . \
-maxdepth 1 \
-type f \
-name '*.html' \
-exec grep -Hn "$REPLACE_STRING" {} \; \
-exec sed -i '' "s/${REPLACE_STRING}/${STRING}/g" {} \;
This prints, for each matching file, the path, the line number and the line:
./file1.html:9:contents of line 9
./file1.html:12:contents of line 12
./file1.html:26:contents of line 26
./file2.html:1:contents of line 1
./file2.html:6:contents of line 6
./file2.html:36:contents of line 36
For files without a match, nothing else happens; for files with a match, the sed command will be called.
If you wanted output closer to what you have in your question, you could add a few actions:
find . \
-maxdepth 1 \
-type f \
-name '*.html' \
-exec grep -q "$REPLACE_STRING" {} \; \
-printf '%p was changed\n' \
-exec grep -n "$REPLACE_STRING" {} \; \
-exec sed -i '' "s/${REPLACE_STRING}/${STRING}/g" {} \; \
| sed -E "s/^([[:digit:]]+):.*/ - $REPLACE_STRING line \1 was changed/"
This now first checks if the file contains the string, silently, with grep -q, then prints the filename (-printf), then all the matching lines with line numbers (grep -n), then does the substitution with sed and finally modifies the output slightly with sed.
Since you're using sed -i '', I assume you're on macOS; I'm not sure if the stock find on there supports the printf option.
By now, we're pretty close to running a complex-ish script on each file that matches, so we might as well do that directly:
shopt -s nullglob
for f in ./*.html; do
if grep -q "$REPLACE_STRING" "$f"; then
printf '%s\n' "$f was changed"
grep -n "$REPLACE_STRING" "$f" \
| sed -E "s/^([[:digit:]]+):.*/ - $REPLACE_STRING line \1 was changed/"
sed -i '' "s/${REPLACE_STRING}/${STRING}/g" "$f"
fi
done

Answer 3

Replace your find+sed command:
find ./ -type f -name "*.html" -maxdepth 1 -exec sed -i '' "s/${REPLACE_STRING}/${STRING}/g" {} \; -print
with this GNU awk command (needs gawk for inplace editing):
gawk -i inplace -v old="$REPLACE_STRING" -v new="$STRING" '
FNR==1 { hdr=FILENAME " was changed\n" }
gsub(old,new) { printf "%s - %s line %d was changed\n", hdr, old, FNR | "cat>&2"; hdr="" }
1' *.html
You could also make it much more robust with awk than with sed if necessary since awk can support literal strings while sed can't

Answer 4

Alright, always defer to Ed's awk script for efficiency, but continuing with the sed + helper script using a preliminary call to grep to determine whether your file contains the word to replace, you could use a short helper script taking your ${REPLACE_STRING}, ${STRING} and filename as the first three positional parameters as follows:
Helper Script named helper.sh
#!/bin/sh
test -z "$1" && exit
test -z "$2" && exit
test -z "$3" && exit
findw="$1"
replw="$2"
fname="$3"
grep -q "$findw" "$fname" || exit
echo "$(readlink -f $fname) was changed"
grep -n "$findw" "$fname" | {
while read line; do
printf -- " - REPLACE STRING line %d was changed\n" "${line%:*}"
done }
sed -i "s/$findw/$replw/g" "$fname"
Then your call to find could be, e.g.:
find . -type f -name "f*" -exec ./helper.sh "dog" "cat" '{}' \;
Example Use/Output
Starting with a couple of files named f containing:
$ cat f
my
dog
dog
has
fleas
In a file structure containing the script in the present directory with a subdirectory d1 and multiple copies of f, e.g.
$ tree .
.
├── d1
│   └── f
├── f
└── helper.sh
Running the script results in the following:
$ find . -type f -name "f*" -exec ./helper.sh "dog" "cat" '{}' \;
/tmp/tmp-david/f was changed
- REPLACE STRING line 2 was changed
- REPLACE STRING line 3 was changed
/tmp/tmp-david/d1/f was changed
- REPLACE STRING line 2 was changed
- REPLACE STRING line 3 was changed
and the contents of f are changed accordingly
$ cat f
my
cat
cat
has
fleas
If there is no search term found in any of the files located by find, the modification times on those files are left unchanged.
Now with all that in mind, if you have gawk available, follow Ed's advise, but -- you can do it with sed and a helper :)

Answer 5

install Perl easily for free, define your own strings on bash shell and test here:
STRING=
REPLACE=
perl -ne 'foreach(`find . -maxdepth 1 -type f -iname "*.html"`){ open IH,$_ or die "Error $!"; print "Processing: $_";while (<IH>) {$s=$_;$t=s/$REPLACE/$STRING/; print "$s --> $_" if $t };print "Nothing replaced" if !$t}'
to truly edit it add -i option so it'd be perl -i -ne....