RE: RE: Re: Sorting on Section Numbers

[jimt]

There is a slight bug in this solution, it drops letters.
Using the above test data, the following list is generated:
1, 1.2, 2, 2.1.7, 2.2, 2.13, 3.4, 10.1, 10.1, 10.10
which dropped the letters from the section headings.
Here's my version:

@sorted = 
    map {$_->[0], ", "}
    sort {$a->[0] <=> $b->[0]}
    map {[$_, pack ("N*", $_)]}
    
@unsorted;

it works the same way, but I substituted in a Schwartzian transform to cache the original value of the heading instead of unpacking it again later. According to some tests with Benchmark, it's about 40% faster than doing the unpack later.
Plus it keeps letters intact.

[DrManhattan]

  @sorted = 
          map {$_->[0], ", "}
          sort {$a->[0] <=> $b->[0]}
          map {[$_, pack ("N*", $_)]}
          
  @unsorted;

For anyone who doesn't understand what the above code does, here are some nodes with explanations of the technique and some interesting discussion:

<list>
• Schwartzian Transform
• Sorting IP Addresses Quickly
• A Fresh Look at Efficient Perl Sorting (thanks davorg)

-Matt

[turnstep]

> it works the same way, but I substituted in a Schwartzian transform to cache the original
> value of the heading instead of unpacking it again later. According to some tests with
> Benchmark, it's about 40% faster than doing the unpack later. Plus it keeps letters intact.

Unfortunately, there are two problems with this code:

@sorted = 
  map {$_->[0], ", "}
  sort { $a->[0] <=> $b->[0]}
  map {[$_, pack ("N*", $_)]}
@unsorted;

First, the map arguments should be reversed or the sort subscripts should be "1" not "0". As it is, all the code does is a simple sort, and ignores the whole pack part! Second, even when it is fixed, it does not quite sort correctly:

@unsorted = qw(2 1.2a 2.2 1.2 1.1a);
## A fixed version:
@sorted = 
  map {$_->[0], ", "}
  sort { $a->[1] <=> $b->[1]}
  map {[$_, pack ("N*", $_)]}
@unsorted;

print @sorted, "\n";
## produces:
1.2a, 1.1a, 1.2, 2, 2.2

[chip]

Of course that solution drops letters. That was the whole point, demonstrating that things get a lot simpler without them. Quoting myself:

If we simplify the problem space by eliminating the alphanumerics, things get even neater.

    -- Chip Salzenberg, Free-Floating Agent of Chaos