I have a hash of complex structure and I want to perform a search and replace. The first hash is like the following:

$VAR1 = {
  abc => { 123 => ["xx", "yy", "zy"], 456 => ["ab", "cd", "ef"] },
  def => { 659 => ["wx", "yg", "kl"], 456 => ["as", "sd", "df"] },
  mno => { 987 => ["lk", "dm", "sd"] },
}

and I want to iteratively search for all '123'/'456' elements, and if a match is found, I need to do a comparison of the sublayer, i.e. of ['ab','cd','ef'] and ['as','sd','df'] and in this case, keep only the one with ['ab','cd','ef']. So the output will be as follows:

$VAR1 = {
  abc => { 123 => ["xx", "yy", "zy"], 456 => ["ab", "cd", "ef"] },
  def => { 659 => ["wx", "yg", "kl"] },
  mno => { 987 => ["lk", "dm", "sd"] },
}

So the deletion is based on the substructure, and not index. How can it be done? Thanks for the help!!

Lets assume that I will declare the values to be kept, i.e. I will keep 456 => ["ab", "cd", "ef"] based on a predeclared value of ["ab", "cd", "ef"] and delete any other instance of 456 anywhere else. The search has to be for every key. so the code will go through the hash, first taking 123 => ["xx", "yy", "zy"] and compare it against itself throughout the rest of the hash, if no match is found, do nothing. If a match is found, like in the case of 456 => ["ab", "cd", "ef"], it will compare the two, and as I have said that in case of a match the one with ["ab", "cd", "ef"] would be kept, it will keep 456 => ["ab", "cd", "ef"] and discard any other instances of 456 anywhere else in the hash, i.e. it will delete 456 => ["as", "sd", "df"] in this case.