How could I rewrite this mv recreation in fd? #1719

fd can run a command for every match, but it is not a faster replacement for this directory-merge operation. The expensive part is moving/copying filesystem entries, not discovering them, and parallel mv calls introduce ordering and collision problems: destination directories must exist first, two sources may share a basename, and cross-filesystem moves become copies.

For the semantics in the gist, use rsync, which already implements recursive directory merging and preserves metadata:

for src in "$@"; do
  if [[ -d $src ]]; then
    rsync -a --remove-source-files -- "$src" ./ || exit
    find "$src" -depth -type d -empty -delete
  elif [[ -e $src || -L $src ]]; then
    mv -- "$src" ./
  else
    printf 'File %s does not exist\n' "$src" >&2
  fi
done

Keep a backup or add --dry-run while testing conflict behavior. Decide explicitly whether an existing destination file should be overwritten; add --ignore-existing if it should not.

If you still want fd for filtering files inside one source tree, -x/--exec invokes once per result and -X/--exec-batch invokes once with all results. Its placeholders include {} (path), {/} (basename), and {//} (parent), documented in the command execution section. But fd does not provide an atomic merge-tree primitive, so using it here adds complexity without making the moves themselves faster.

If rsync --remove-source-files matches the intended merge behavior, please mark this as accepted.