Re^2: Stats for super search look buggy

[jdporter]

Great. Now I feel even worse about that. :-(

[pryrt]

Well, the percentage calculator could be presumably changed to something like

use constant jump => 9866219;
my $delta = $last_node_in_search_range - $first_node_in_search_range;
$delta -= jump if $delta > jump;
my $percent = 100 * $delta / ($most_recent_node_id - jump);

[jdporter]

Right, but I don't think that the delta merely being larger than the gap is a sufficient test; we should check whether the search actually spans the gap:

$delta -= jump
  if  $first_node_in_search_range <= 1233781 
  &&   $last_node_in_search_range >= 11100000;

[LanX]

Diminishing the bigger a ° value by the gap-distance, if it exceeds the threshold, is sufficient.

$bigger_node_in_search_range -= $jump if 
    $bigger_node_in_search_range > 1233781;

for ($node_min, $node_max) {
    $_ -= $jump if  $_ > 1233781;
}

Calculate percentage then

update

°) sorry, both could have jumped.

[pryrt]

Good point. As of today, I don't see a way that the delta would be over 9 million if it wasn't crossing the gap without another 9million posts or another gap, but might as well future-proof it ... and just because I don't see something doesn't mean it doesn't exist.

My reply was a rather off-the-cuff, first-level approximation of how I thought I might solve it, rather than a fully fleshed-out solution, with implications and edge cases considered. Also, I also realized that my reply could have come across as "why haven't you come up with this", which isn't what I intended; rather, I meant something more like "but at least the fix shouldn't be too difficult to come up with (famous last words), something like /this/".

[LanX]

I've already identified the code in question, but it's a little more complicated.

Furthermore there is also the "remaining" part, and the search could go backwards.

Interested pmdevs can submit patches here

 511:           . qq[ (searched ] . sprintf( "%.2f",
 512:               100*( abs($n0-$start)+1 ) / $lastNode
 513:             ) . qq[% of DB).</div><br />\n<div class="ss-criteri
+a-summary">\n];

and

 569:         } else {
 570:             my( $min, $max )= $oldFirst ? ( $n0, $lastNode ) : (
+ 1, $n0 );
 571:             my $pct= sprintf "%.2f%%", 100*($max-$min+1)/$lastNo
+de;
 572:             if(  1 == $doSearch  ) {
 573:                 $html .= qq[Press ]
 574:                   . $q->submit( "nx", "Next >" )
 575:                   . qq[ to <b>continue</b> searching remaining $
+pct of DB.];
 576:             } else {

[LanX]

Hey relax, 💩 happens! :)