When merging directories with similarly-named files, how do you ensure that only smaller+earlier files are replaced?

Question

E.g., if you want to merge two directories with each other (by moving everything from directory 1 into directory 2), and both directory 1 and directory 2 have some files with the same name.

Then how can you write code such that if SharedFile is found in both directories, then replace SharedFile in directory 2 with SharedFile in directory 1 IF SharedFile is bigger in directory 1 and SharedFile has a later modification date in directory 1? (but do not replace SharedFile otherwise).

I'm fine with both tcsh and bash scripts.

Answer 1

Here's a bash/ksh93/zsh script that emulates the core behavior of rsync, where you can easily tune the decision to copy or not copy a source file. Here a copy is made only if the source file is both larger and newer. In bash, add shopt -s globdots before the script. Untested.

target=/path/to/destination
cd source-directory
skip=
err=0
for x in **/*; do
  # Skip the contents of a directory that has been copied wholesale
  case $x in $skip/*) continue;; *) skip=;; esac
  if [[ -d $x ]]; then
    # Recreate source directory on the target.
    # Note that existing directories do not have their permissions or modification times updated.
    if [[ -e $target/$x ]]; then continue; fi
    skip=$x
    if [[ -e $target/$x ]]; then
      echo 1>&2 "Not overwriting non-directory $target/$x with a directory."
      err=1
    else
      # The directory doesn't exist on the destination, so copy it
      cp -Rp -- "$x" "$target/$x" || err=1
    fi
  elif [[ -f $x ]]; then
    # We have a regular file. Copy it over if desired.
    if [[ $x -nt $target/$x ]] && [ $(wc -c <"$x") -gt $(wc -c <"$target/$x") ]; then
      cp -p -- "$x" "$target/$" || err=1
    fi
  else
    # neither a file nor a directory. Overwrite the destination
    cp -p -- "$x" "$target/$x" || err=1
  fi
done

Answer 2

According to rsync --help:

 -u, --update                skip files that are newer on the receiver

So, the command you're probably looking for is:

rsync -auPsourcedir/ destdir/

Then delete sourcedir after the fact.

Of course, remember the significance of trailing /'s in rsync.

I don't know that there is an rsync behavior for treating larger files as better...