rsync rename duplicated files in dest directory - bash

Question

I have implemented a rsync based system to move files from different environments to others.
The problem I'm facing now is that sometimes, there are files with the same name, but different path and content.
I want to make rsync (if possible) rename duplicated files because I need and use --no-relative option.
Duplicated files can occur in two ways:
There was a file with same name in dest directory already.
In the same rsync execution, we are transferring file with same name in a different location. Ex: dir1/file.txt and dir2/file.txt
Adding -b --suffix options, allows me to have at least 1 repetition for the first duplicated file's type mentioned.
A minimum example (for Linux based systems):
mkdir sourceDir1 sourceDir2 sourceDir3 destDir;
echo "1" >> sourceDir1/file.txt;
echo "2" >> sourceDir2/file.txt;
echo "3" >> sourceDir3/file.txt;
rsync --no-relative sourceDir1/file.txt destDir
rsync --no-relative -b --suffix="_old" sourceDir2/file.txt sourceDir3/file.txt destDir
Is there any way to achieve my requirements?

Answer 1

I don't think that you can do it directly with rsync.
Here's a work-around in bash that does some preparation work with find and GNU awk and then calls rsync afterwards.
The idea is to categorize the input files by "copy number" (for example sourceDir1/file.txt would be the copy #1 of file.txt, sourceDir2/file.txt the copy #2 and sourceDir3/file.txt the copy #3) and generate a file per "copy number" containing the list of all the files in that category.
Then, you just have to launch an rsync with --from-file and a customized --suffix per category.
Pros
fast: incomparable to firing one rsync per file.
safe: it won't ever overwrite a file (see the step #3 below).
robust: handles any filename, even with newlines in it.
Cons
the destination directory have to be empty (or else it might overwrite a few files).
the code is a little long (and I made it longer by using a few process substitutions and by splitting the awk call into two).
Here are the steps:
0)   Use a correct shebang for bash in your system.
#!/usr/bin/env bash
1)   Create a directory for storing the temporary files.
tmpdir=$( mktemp -d ) || exit 1
2)   Categorize the input files by "duplicate number", generate the files for rsync --from-file (one per dup category), and get the total number of categories.
read filesCount < <(
find sourceDir* -type f -print0 |
LANG=C gawk -F '/' '
BEGIN {
RS = ORS = "\0"
tmpdir = ARGV[2]
delete ARGV[2]
}
{
id = ++seen[$NF]
if ( ! (id in outFiles) ) {
outFilesCount++
outFiles[id] = tmpdir "/" id
}
print $0 > outFiles[id]
}
END {
printf "%d\n", outFilesCount
}
' - "$tmpdir"
)
3)   Find a unique suffix — generated using a given set of chars — for rsync --suffix => the string shall be appended to it.
note: You can skip this step if you know for sure that there's no existing filename that ends with _old+number.
(( filesCount > 0 )) && IFS='' read -r -d '' suffix < <(
LANG=C gawk -F '/' '
BEGIN {
RS = ORS = "\0"
charsCount = split( ARGV[2], chars)
delete ARGV[2]
for ( i = 1; i <= 255; i++ )
ord[ sprintf( "%c", i ) ] = i
}
{
l0 = length($NF)
l1 = length(suffix)
if ( substr( $NF, l0 - l1, l1) == suffix ) {
n = ord[ substr( $NF, l0 - l1 - 1, 1 ) ]
suffix = chars[ (n + 1) % charsCount ] suffix
}
}
END {
print suffix
}
' "$tmpdir/1" '0/1/2/3/4/5/6/7/8/9/a/b/c/d/e/f'
)
4)   Run the rsync(s).
for (( i = filesCount; i > 0; i-- ))
do
fromFile=$tmpdir/$i
rsync --no-R -b --suffix="_old${i}_$suffix" -0 --files-from="$fromFile" ./ destDir/
done
5)   Clean-up the temporary directory.
rm -rf "$tmpdir"

Answer 2

Guess it's not possible with only rsync. You have to make a list of files first and analyze it to work around dupes. Take a look at this command:
$ rsync --no-implied-dirs --relative --dry-run --verbose sourceDir*/* dst/
sourceDir1/file.txt
sourceDir2/file.txt
sourceDir3/file.txt
sent 167 bytes received 21 bytes 376.00 bytes/sec
total size is 6 speedup is 0.03 (DRY RUN)
Lets use it to create list of source files:
mapfile -t list < <(rsync --no-implied-dirs --relative --dry-run --verbose sourceDir*/* dst/)
Now we can loop through this list with something like this:
declare -A count
for item in "${list[#]}"; {
[[ $item =~ ^sent.*bytes/sec$ ]] && break
[[ $item ]] || break
fname=$(basename $item)
echo "$item dst/$fname${count[$fname]}"
((count[$fname]++))
}
sourceDir1/file.txt dst/file.txt
sourceDir2/file.txt dst/file.txt1
sourceDir3/file.txt dst/file.txt2
Change echo to rsync and that is it.