views:

213

answers:

7

In Perl, how do I get this:

$VAR1 = { '999' => { '998' => [ '908', '906', '0', '998', '907' ] } }; 
$VAR1 = { '999' => { '991' => [ '913', '920', '918', '998', '916', '919', '917', '915', '912', '914' ] } }; 
$VAR1 = { '999' => { '996' => [] } }; 
$VAR1 = { '999' => { '995' => [] } }; 
$VAR1 = { '999' => { '994' => [] } }; 
$VAR1 = { '999' => { '993' => [] } }; 
$VAR1 = { '999' => { '997' => [ '986', '987', '990', '984', '989', '988' ] } }; 
$VAR1 = { '995' => { '101' => [] } }; 
$VAR1 = { '995' => { '102' => [] } }; 
$VAR1 = { '995' => { '103' => [] } }; 
$VAR1 = { '995' => { '104' => [] } }; 
$VAR1 = { '995' => { '105' => [] } }; 
$VAR1 = { '995' => { '106' => [] } }; 
$VAR1 = { '995' => { '107' => [] } }; 
$VAR1 = { '994' => { '910' => [] } }; 
$VAR1 = { '993' => { '909' => [] } }; 
$VAR1 = { '993' => { '904' => [] } }; 
$VAR1 = { '994' => { '985' => [] } }; 
$VAR1 = { '994' => { '983' => [] } }; 
$VAR1 = { '993' => { '902' => [] } }; 
$VAR1 = { '999' => { '992' => [ '905' ] } }; 

to this:

$VAR1 = { '999:' => [
 { '992' => [ '905' ] },
 { '993' => [
  { '909' => [] },
  { '904' => [] },
  { '902' => [] }
 ] },
 { '994' => [
  { '910' => [] },
  { '985' => [] },
  { '983' => [] }
 ] },
 { '995' => [
  { '101' => [] },
  { '102' => [] },
  { '103' => [] },
  { '104' => [] },
  { '105' => [] },
  { '106' => [] },
  { '107' => [] }
 ] },
 { '996' => [] },
 { '997' => [ '986', '987', '990', '984', '989', '988' ] },
 { '998' => [ '908', '906', '0', '998', '907' ] },
 { '991' => [ '913', '920', '918', '998', '916', '919', '917', '915', '912', '914' ] }
]};
A: 

I indented your wanted output as it was hard to read, for the benefit of other people who want to answer. I'm still thinking of an answer.

$VAR1 = { '999:' => [
                      { '992' => [ '905' ] },
                      { '993' => [
                                   { '909' => [] },
                                   { '904' => [] },
                                   { '902' => [] }
                                 ]
                      },
                      { '994' => [
                                   { '910' => [] },
                                   { '985' => [] },
                                   { '983' => [] }
                                 ]
                      },
                      { '995' => [
                                   { '101' => [] },
                                   { '102' => [] },
                                   { '103' => [] },
                                   { '104' => [] },
                                   { '105' => [] },
                                   { '106' => [] },
                                   { '107' => [] }
                                 ]
                      },
                      { '996' => [] },
                      { '997' => [ '986', '987', '990', '984', '989', '988' ] },
                      { '998' => [ '908', '906', '0', '998', '907' ] },
                      { '991' => [ '913', '920', '918', '998', '916', '919', '917', '915', '912', '914' ] }
                    ]
        };

I don't see the point of all those single entry hashes though, would not the following be better?

$VAR1 = { '999:' => {
                      '992' => [ '905' ],
                      '993' => {
                                 '909' => [],
                                 '904' => [],
                                 '902' => []
                               },
                      '994' => {
                                 '910' => [],
                                 '985' => [],
                                 '983' => []
                               },
                      '995' => {
                                 '101' => [],
                                 '102' => [],
                                 '103' => [],
                                 '104' => [],
                                 '105' => [],
                                 '106' => [],
                                 '107' => []
                               },
                      '996' => [],
                      '997' => [ '986', '987', '990', '984', '989', '988' ],
                      '998' => [ '908', '906', '0', '998', '907' ],
                      '991' => [ '913', '920', '918', '998', '916', '919', '917', '915', '912', '914' ]
                    }
        };
Christoffer Hammarström
A: 

Assuming the above data is in a file dump.txt, you can eval it piece by piece.

Updated code below

use strict;
use File::Slurp;
my $final_data = {}; 
my @data = map {eval $_} (read_file("dump.txt") =~ /\$VAR1 = ([^;]+);/gs);
foreach my $element (@data) {
    my $key = (keys %$element)[0]; 
    $final_data->{$key} ||= []; 
    push @{$final_data->{$key}}, $element->{$key}
}; 
use Data::Dumper; 
print Data::Dumper->Dump([$final_data]);

If you want to completely deep merge, you can at the end pass $final_data through this (not tested!!!) deep merger:

# Merge an array of hashes as follows:
# IN:  [ { 1 => 11 }, { 1 => 12 },{ 2 => 22 } ]
# OUT: { 1 => [ 11, 12 ], 2 => [ 22 ] }
# This is recursive - if array [11,12] was an array of hashrefs, we merge those too
sub merge_hashes {
    my $hashes = @_[0];
    return $hashes unless ref $hashes eq ref []; # Hat tip to brian d foy
    return $hashes unless grep { ref @_ eq ref {} } @$hashes; # Only merge array of hashes
    my $final_hashref = {};
    foreach my $element (@$hashes) {
        foreach my $key (keys %$element) {
            $final_hashref->{$key} ||= [];
            push @{ $final_hashref->{$key} }, $element->{$key};
        }
    }
    foreach my $key (keys %$final_hashref) {
        $final_hashref->{$key} = merge_hashes($final_hashref->{$key});
    }
    return $final_hashref;
}
DVK
NOTE: I am assuming the only merging happens at the topmist key. If not, it's a bit harder to do though not too hard... left as excercize for the reader for now :)
DVK
Well, it kind of looks as if he wants a tree out of the whole thing as well, with what would be "top-level" keys being merged with their matching second-level keys.
Axeman
@Axeman - OK... I don't think I could parse that comment at all... need more sleep :) Do you mean he wants to merge second-level keys as well?
DVK
Updated my original code to account for the fact that real life Data::Dumper output would not necessarily be nicely lined up on 1 line per call :)
DVK
Added deep merger - hopefully it works ;)
DVK
@DVK: Sorry, I use shortcut verbs a lot nowadays. If you take the "wanted" result into a text editor and check for matching braces you'll see that the only top level key is '999' so it serves as the root of a tree-like structure.
Axeman
+1  A: 

Give this recursive solution a try:

#   XXX: doesn't handle circular problems...
sub deepmerge {
    my (@structs) = @_;
    my $new;

    # filter out non-existant structs
    @structs = grep {defined($_)} @structs;

    my $ref = ref($structs[0]);
    if (not all(map {ref($_) eq $ref} @structs)) { 
        warn("deepmerge: all structs are not $ref\n");
    } 

    my @tomerge = grep {ref($_) eq $ref} @structs;
    return qr/$tomerge[0]/ if scalar(@tomerge) == 1 and $ref eq 'Regexp';
    return $tomerge[0] if scalar(@tomerge) == 1;

    if ($ref eq '') { 
        $new = pop(@tomerge); # prefer farthest right
    } 
    elsif ($ref eq 'Regexp') { 
        $new = qr/$tomerge[$#tomerge]/;
    } 
    elsif ($ref eq 'ARRAY') { 
        $new = [];
        for my $i (0 .. max(map {scalar(@$_) - 1} @tomerge)) { 
            $new->[$i] = deepmerge(map {$_->[$i]} @tomerge);
        }
    } 
    elsif ($ref eq 'HASH') { 
        $new = {};
        for my $key (uniq(map {keys %$_} @tomerge)) { 
            $new->{$key} = deepmerge(map {$_->{$key}} @tomerge);
        }
    }
    else {
        # ignore all other structures...
        $new = '';
    }

    return $new;
}

Modify it to your hearts content to achieve the desired result.

Upon further investigation, I noticed you're merging them in some different way than the above algorithm. Maybe just use this as an example then. Mine does this:

deepmerge({k => 'v'}, {k2 => 'v2'});
# returns {k => 'v', k2 => 'v2'}

And similar things for arrays.

xyld
Wow, I hope you work as hard for your employer as you do for random strangers on the internet. :)
Ether
@Ether - may be he's just Jon Skeet's sock puppet account. (and i'm just Jon Skeet's bot... before you ask)
DVK
A: 

Use push and autovivification.

Start with the usual front matter:

#! /usr/bin/perl

use warnings;
use strict;

Read your sample input from the DATA filehandle and create a datastructure similar to the one you dumped:

my @hashes;
while (<DATA>) {
  my $VAR1;
  $VAR1 = eval $_;
  die $@ if $@;
  push @hashes => $VAR1;
}

Your input has two cases:

  1. A reference to an array that contains data to be merged with its cousins that have the same "key path."
  2. Otherwise, it's a reference to a hash that contains a reference to an array from case 1 at some depth, so we strip off the outermost layer and keep digging.

Note the use of $_[0]. The semantics of Perl subroutines are such that the values in @_ are aliases rather than copies. This lets us call merge directly without having to first create a bunch of scaffolding to hold the merged contents. The code will break if you copy the value instead.

sub merge {
  my $data = shift;

  if (ref($data) eq "ARRAY") {
    push @{ $_[0] } => @$data;
  }
  else {
    foreach my $k (%$data) {
      merge($data->{$k} => $_[0]{$k});
    }
  }
}

Now we walk @hashes and incrementally merge their contents into %merged.

my %merged;    
foreach my $h (@hashes) {
  foreach my $k (keys %$h) {
    merge $h->{$k} => $merged{$k};
  }
}

We don't know in what order the values arrived, so run a final cleanup pass to sort the arrays:

sub sort_arrays {
  my($root) = @_;
  if (ref($root) eq "ARRAY") {
    @$root = sort { $a <=> $b } @$root;
  }
  else {
    sort_arrays($root->{$_}) for keys %$root;
  }
}

sort_arrays \%merged;

The Data::Dumper module is great for quick debugging!

use Data::Dumper;
$Data::Dumper::Indent = 1;
print Dumper \%merged;

Place a copy of the input from your question into the special DATA filehandle:

__DATA__
$VAR1 = { '999' => { '998' => [ '908', '906', '0', '998', '907' ] } };
$VAR1 = { '999' => { '991' => [ '913', '920', '918', '998', '916', '919', '917', '915', '912', '914' ] } };
$VAR1 = { '999' => { '996' => [] } };
$VAR1 = { '999' => { '995' => [] } };
$VAR1 = { '999' => { '994' => [] } };
$VAR1 = { '999' => { '993' => [] } };
$VAR1 = { '999' => { '997' => [ '986', '987', '990', '984', '989', '988' ] } };
$VAR1 = { '995' => { '101' => [] } };
$VAR1 = { '995' => { '102' => [] } };
$VAR1 = { '995' => { '103' => [] } };
$VAR1 = { '995' => { '104' => [] } };
$VAR1 = { '995' => { '105' => [] } };
$VAR1 = { '995' => { '106' => [] } };
$VAR1 = { '995' => { '107' => [] } };
$VAR1 = { '994' => { '910' => [] } };
$VAR1 = { '993' => { '909' => [] } };
$VAR1 = { '993' => { '904' => [] } };
$VAR1 = { '994' => { '985' => [] } };
$VAR1 = { '994' => { '983' => [] } };
$VAR1 = { '993' => { '902' => [] } };
$VAR1 = { '999' => { '992' => [ '905' ] } };

A sample of the output is below:

  '994' => {
    '910' => [],
    '985' => [],
    '983' => []
  },
  '999' => {
    '993' => [],
    '992' => [
      '905'
    ],
    '997' => [
      '984',
      '986',
      '987',
      '988',
      '989',
      '990'
    ],
Greg Bacon
+1  A: 

Use CPAN! Try Hash::Merge

# OO interface.  
my $merge = Hash::Merge->new( 'LEFT_PRECEDENT' );
my %c = %{ $merge->merge( \%a, \%b ) };

See CPAN for more info, it pretty much does everything you would want to, and is fully customizable.

Evan Carroll
+2  A: 

I think this is closer than anybody else has gotten:

This does most of what you want. I did not store things in arrays of singular hashes, as I don't feel that that is useful.

Your scenario is not a regular one. I've tried to genericize this to some extent, but was not possible to overcome the singularity of this code.

  • First of all because it appears you want to collapse everything with the same id into a merged entity (with exceptions), you have to descend through the structure pulling the definitions of the entities. Keeping track of levels, because you want them in the form of a tree.

  • Next, you assemble the ID table, merging entities as possible. Note that you had 995 defined as an empty array one place and as a level another. So given your output, I wanted to overwrite the empty list with the hash.

  • After that, we need to move the root to the result structure, descending that in order to assign canonical entities to the identifiers at each level.

Like I said, it's not anything that regular. Of course, if you still want a list of hashes which are no more than pairs, that's an exercise left to you.

use strict;
use warnings;

# subroutine to identify all elements
sub descend_identify {
    my ( $level, $hash_ref ) = @_;
    # return an expanding list that gets populated as we desecend 
    return map {
        my $item = $hash_ref->{$_};
        $_ => ( $level, $item )
            , ( ref( $item ) eq 'HASH' ? descend_identify( $level + 1, $item ) 
              :                          ()
              )
           ;
    } keys %$hash_ref
    ;
}

# subroutine to refit all nested elements
sub descend_restore { 
    my ( $hash, $ident_hash ) = @_;

    my @keys        = keys %$hash;
    @$hash{ @keys } = @$ident_hash{ @keys };
    foreach my $h ( grep { ref() eq 'HASH' } values %$hash ) {
        descend_restore( $h, $ident_hash );
    }
    return;
}

# merge hashes, descending down the hash structures.
sub merge_hashes {
    my ( $dest_hash, $src_hash ) = @_;
    foreach my $key ( keys %$src_hash ) {
        if ( exists $dest_hash->{$key} ) {
            my $ref = $dest_hash->{$key};
            my $typ = ref( $ref );
            if ( $typ eq 'HASH' ) {
                merge_hashes( $ref, $src_hash->{$key} );
            }
            else { 
                push @$ref, $src_hash->{$key};
            }
        }
        else {
            $dest_hash->{$key} = $src_hash->{$key};
        }
    }
    return;
}

my ( %levels, %ident_map, %result );

#descend through every level of hash in the list
# @hash_list is assumed to be whatever you Dumper-ed.
my @pairs = map { descend_identify( 0, $_ ); } @hash_list;

while ( @pairs ) {
    my ( $key, $level, $ref ) = splice( @pairs, 0, 3 );
    $levels{$key} |= $level;

    # if we already have an identity for this key, merge the two
    if ( exists $ident_map{$key} ) {
        my $oref = $ident_map{$key};
        my $otyp = ref( $oref );
        if ( $otyp ne ref( $ref )) {
            # empty arrays can be overwritten by hashrefs -- per 995
            if ( $otyp eq 'ARRAY' && @$oref == 0 && ref( $ref ) eq 'HASH' ) {
                $ident_map{$key} = $ref;
            }
            else { 
                die "Uncertain merge for '$key'!";
            }
        }
        elsif ( $otyp eq 'HASH' ) {
            merge_hashes( $oref, $ref );
        }
        else {
            @$oref = sort { $a <=> $b || $a cmp $b } keys %{{ @$ref, @$oref }};
        }
    }
    else {
        $ident_map{$key} = $ref;
    }
}

# Copy only the keys that do not appear at higher levels to the 
# result hash
if ( my @keys = grep { !$levels{$_} } keys %ident_map ) { 
    @result{ @keys } = @ident_map{ @keys } if @keys;

}
# then step through the hash to make sure that the entries at
# all levels are equal to the identity
descend_restore( \%result, \%ident_map );
Axeman
wow. thanks so much; sorry for the lack of code or clarification, I was trying to generate a tree, and did try Hash::Merge, but could not for the life of me resolve the coined-995 problem of replacing the empty 995 with the non-empty 995; this works beautifully and i really appreciate the help! (also tried the others and it either did the same thing as Hash::Merge, or it actually got rid of some branches)
Nick
A: 

wow. thanks so much everyone (especially Axeman)! sorry for the lack of code or clarification, I was trying to generate a tree, and did try Hash::Merge, but could not for the life of me resolve the coined-995 problem of replacing the empty 995 with the non-empty 995; Axeman's solution works beautifully and I really appreciate the help/collaboration! (also tried the others and it either did the same thing as Hash::Merge, or it actually got rid of some branches).

some background on the input: had a set of hashes, each had keys (all same level) and two of which defined a) a parent to another, and b) itself (the rest were children), and so with a tree, i figured a hash was perfect, came up with a set of new hashes {a}->{b}->[c], and here we are...

again, thanks everyone and Axeman!

Nick