RogerBW's Blog

I’ve been doing the Perl Weekly Challenges. The latest involved converting time formats and more tree traversal. (Note that this is open until 21 February 2021.)

TASK #1 › Fun Time

You are given a time (12 hour / 24 hour).

Write a script to convert the given time from 12 hour format to 24 hour format and vice versa.

Ideally we expect a one-liner.

Roger don't do one-liners. Roger been maintenance programmer had to fix bugs in somebody else's one-liners. Some one-liners got more bugs than actual characters.

(I did once write a line of Perl with a complicated double-map in it, to transform a nested data structure into the PostScript fragment the customer wanted. They swore blind that there would never be any need to change this other than by changing the list of fields. So in a comment under that double-map I wrote out the full version of the function, just in case.

I told that story at a London Perlmongers meet a few months later and found beer in front of me, from the guy who'd been hired to change it…)

Anyway, a full implementation of this should also take into account that the 12-hour clock runs (am) 12, 1..11, then (pm) 12, 1..11, while the 24-hour runs 0..23. I added several test cases to the few that were given.

Of course if we had a time and date then we could just use timelocal/localtime/strftime, and equivalents in other languages. But I'm a time purist.

Anyway, regex parsing into three components: hour, minute, optional am/pm flag.

sub ft {
  my $in=shift;
  $in =~ /(\d+):(\d+)\s*([ap]m)?/;
  my $h=$1;
  my $t='';

If there is an am/pm flag, it's a 12-hour time; make a hour value of 12 into 0, then add 12 to whatever hour we have if the time is after noon.

  if ($3) {                     # 12 to 24
    if ($h==12) {
      $h=0;
    }
    if ($3 eq 'pm') {
      $h+=12;
    }

Otherwise, it's a 24-hour time so reverse the process: default am, set pm (and subtract 12) if the hour value is after noon, then set an hour value of 0 to 12.

  } else {                      # 24 to 12
    $t=' am';
    if ($h > 11) {
      $t=' pm';
      $h-=12;
    }
    if ($h == 0) {
      $h=12;
    }
  }

Finally, throw the whole thing back together (with the minute match from the original regex).

  return sprintf('%02d:%02d%s',$h,$2,$t);
}

The same basic algorithm applies to the other languages. Raku's regexes are gratuitously incompatible with everyone else's:

  $in ~~ /(\d+)\:(\d+)\s*(<[ap]>m)?/;

while Python, Ruby and Rust all default to letting me catch the matches in a specific variable rather than globals. (I think Raku can do this too with Match objects, but the Perl approach mostly worked there so I used it.)

Python and Rust have their own special formatting languages rather than good old printf that any Unix programmer already knows. Sigh.

    return "{:02d}:{:02d}{:s}".format(h,int(m.group(2)),t)

    return format!("{:02}:{:02}{}",h,caps.get(2).unwrap().as_str(),t);

TASK #2 › Triangle Sum

You are given triangle array.

Write a script to find the minimum path sum from top to bottom.

When you are on index i on the current row then you may move to either index i or index i + 1 on the next row.

Well, if you're going to make it trivial with a hint like that…

My standard pattern for this is a FIFO buffer which makes a breadth-first search, but there's no particular virtue in that for this problem since I won't be terminating early (any valid solution must traverse the entire depth of the structure); so here I decided to use LIFO for a depth-first search, on the basis that the code then appends and removes only at the end of the list rather than at the start too, which should be a little faster if the problem is ever large enough for execution speed to be significant.

Here's the Raku version (taking advantage of the new sigil policy to throw lists and lists of lists around as though they were scalar variables). The others look pretty similar.

sub ts($in) {
  my @b;
  my $n=0;
  my $i=0;
  my $s=$in[0][0];
  my @r;
  while (1) {
    if (@b.elems > 0) {
      my $t=@b.pop;
      ($n,$i,$s)=$t;
    }
    if ($n < $in.elems-1) {
      $n++;
      for ($i,$i+1) -> $ix {
        push @b,[$n,$ix,$s+$in[$n][$ix]];
      }
    } else {
      push @r,$s;
    }
    unless (@b) {
      last;
    }
  }
  return min(@r);
}

In Rust I built a custom data structure rather than using just another Vec. It felt like the Rust-y thing to do.

pub struct Node {
    n: usize,
    i: usize,
    s: i32
}

and later

b.push(Node {n: n,
             i: ix,
             s: s+inp[n][ix]
});

Full code on github.