Re^2: Algorithm To Select Lines Based On Attributes

Input Example:

DefectRecordSpec 17 DEFECTID XREL YREL XINDEX YINDEX XSIZE YSIZE DEFEC
+TAREA DSIZE CLASSNUMBER TEST CLUSTERNUMBER ROUGHBINNUMBER FINEBINNUMB
+ER REVIEWSAMPLE IMAGECOUNT IMAGELIST ;
DefectList
 1 466.458 4741.229 -2 24 2.000 1.725 3.451440 2.000 0 1 0 0 0 0 0 0
 2 5329.992 4264.499 -2 24 1.500 0.862 1.294290 1.500 0 1 0 0 0 0 0 0
 3 1469.965 4591.523 -1 24 0.500 0.431 0.215715 0.500 0 1 0 0 0 0 0 0
 4 7082.505 4283.913 -1 24 2.000 1.725 3.451440 2.000 0 1 0 0 0 0 0 0
 5 777.809 2623.219 -2 24 1.000 0.862 0.862860 1.000 0 1 0 0 0 0 0 0
 6 376.807 3904.135 -1 24 1.500 0.862 1.294290 1.500 0 1 0 0 0 0 0 0
 7 5345.841 3877.818 0 24 0.500 0.431 0.215715 0.500 0 1 0 0 0 0 0 0
[download]

I could make a rule XINDEX,-2 ( give me X-index of only -2. Negating this give me all values without XINDEX = -2 ). Output:

DefectRecordSpec 17 DEFECTID XREL YREL XINDEX YINDEX XSIZE YSIZE DEFEC
+TAREA DSIZE CLASSNUMBER TEST CLUSTERNUMBER ROUGHBINNUMBER FINEBINNUMB
+ER REVIEWSAMPLE IMAGECOUNT IMAGELIST ;
DefectList
 1 466.458 4741.229 -2 24 2.000 1.725 3.451440 2.000 0 1 0 0 0 0 0 0
 2 5329.992 4264.499 -2 24 1.500 0.862 1.294290 1.500 0 1 0 0 0 0 0 0
 5 777.809 2623.219 -2 24 1.000 0.862 0.862860 1.000 0 1 0 0 0 0 0 0
[download]

Comment on Re^2: Algorithm To Select Lines Based On Attributes Select or Download Code

Replies are listed 'Best First'.
Re^3: Algorithm To Select Lines Based On Attributes by gone2015 (Deacon) on Jan 15, 2009 at 21:36 UTC
You said the rules should be additive. If you're applying two rules, say XINDEX == -2, and XSIZE == 2, do you want to select defect 1 once or twice ? If twice, then I you're stuck with N * R work, where N is the number of records and R is the number of rules. If once, then one trick would be to run easy and/or most likely to match rules first. If the rules weren't additive, I'd suggest looking for ways to quickly exclude defects which definitely would not match any rule -- the idea being that you may be able to quickly hack away at the search space. So, as others have said, the problem is how to take your rules and render them into something that Perl can execute as quickly as possible. Here I think the basic trick is something along these lines: use strict ; use warnings ; my @DefectRecordSpec = qw(DEFECTID XREL YREL XINDEX YINDEX XSIZE YSIZE + DEFECTAREA DSIZE CLASSNUMBER TEST CLUSTERNUMBER ROUGHBINNUMBE +R FINEBINNUMBER REVIEWSAMPLE IMAGECOUNT IMAGELIST) ; my @DefectList = ( '1 466.458 4741.229 -2 24 2.000 1.725 3.451440 2.000 0 1 0 0 0 0 0 0 +', '2 5329.992 4264.499 -2 24 1.500 0.862 1.294290 1.500 0 1 0 0 0 0 0 0 +', '3 1469.965 4591.523 -1 24 0.500 0.431 0.215715 0.500 0 1 0 0 0 0 0 0 +', '4 7082.505 4283.913 -1 24 2.000 1.725 3.451440 2.000 0 1 0 0 0 0 0 0 +', '5 777.809 2623.219 -2 24 1.000 0.862 0.862860 1.000 0 1 0 0 0 0 0 0 +', '6 376.807 3904.135 -1 24 1.500 0.862 1.294290 1.500 0 1 0 0 0 0 0 0 +', '7 5345.841 3877.818 0 24 0.500 0.431 0.215715 0.500 0 1 0 0 0 0 0 0 +', ) ; my @rules = ( '$XINDEX == -2', '($XSIZE > 1.75) && ($XSIZE <= 2.15)', ) ; my $fields = '$'. join(', $', @DefectRecordSpec) ; my $rules = join("\n \|\| ", map "($_)", @rules) ; my $sub = join("\n", 'sub {', ' my ($line) = @_ ;', ' $line =~ s/^\s+// ;', # Trim leading spaces, so split is not foo +led # NB: the split will discard trailing spac +es and newline ' my ('. $fields .') = split(/\s+/, $line) ;', ' return '. $rules .' ;', '} ;' ) ; my $test = eval $sub ; print $sub, "\n" ; foreach my $d (@DefectList) { print "$d\n" if $test->($d) ; } ; [download] where your rules are each small Perl expressions, so can be arbitrarily complicated.	[reply] [d/l]

Replies are listed 'Best First'.

Re^3: Algorithm To Select Lines Based On Attributes
by gone2015 (Deacon) on Jan 15, 2009 at 21:36 UTC

You said the rules should be additive. If you're applying two rules, say XINDEX == -2, and XSIZE == 2, do you want to select defect 1 once or twice ? If twice, then I you're stuck with N * R work, where N is the number of records and R is the number of rules. If once, then one trick would be to run easy and/or most likely to match rules first.

If the rules weren't additive, I'd suggest looking for ways to quickly exclude defects which definitely would not match any rule -- the idea being that you may be able to quickly hack away at the search space.

So, as others have said, the problem is how to take your rules and render them into something that Perl can execute as quickly as possible. Here I think the basic trick is something along these lines:

use strict ; use warnings ;

my @DefectRecordSpec = qw(DEFECTID XREL YREL XINDEX YINDEX XSIZE YSIZE
+ DEFECTAREA DSIZE
                          CLASSNUMBER TEST CLUSTERNUMBER ROUGHBINNUMBE
+R FINEBINNUMBER
                          REVIEWSAMPLE IMAGECOUNT IMAGELIST) ;

my @DefectList = (
 '1  466.458 4741.229 -2 24 2.000 1.725 3.451440 2.000 0 1 0 0 0 0 0 0
+',
 '2 5329.992 4264.499 -2 24 1.500 0.862 1.294290 1.500 0 1 0 0 0 0 0 0
+',
 '3 1469.965 4591.523 -1 24 0.500 0.431 0.215715 0.500 0 1 0 0 0 0 0 0
+',
 '4 7082.505 4283.913 -1 24 2.000 1.725 3.451440 2.000 0 1 0 0 0 0 0 0
+',
 '5  777.809 2623.219 -2 24 1.000 0.862 0.862860 1.000 0 1 0 0 0 0 0 0
+',
 '6  376.807 3904.135 -1 24 1.500 0.862 1.294290 1.500 0 1 0 0 0 0 0 0
+',
 '7 5345.841 3877.818  0 24 0.500 0.431 0.215715 0.500 0 1 0 0 0 0 0 0
+',
) ;

my @rules = (
  '$XINDEX == -2',
  '($XSIZE > 1.75) && ($XSIZE <= 2.15)',
) ;

my $fields = '$'. join(', $', @DefectRecordSpec) ;
my $rules  = join("\n      || ", map "($_)", @rules) ;

my $sub = join("\n",
  'sub {',
  '  my ($line) = @_ ;',
  '  $line =~ s/^\s+// ;',  # Trim leading spaces, so split is not foo
+led
                            # NB: the split will discard trailing spac
+es and newline
  '  my ('. $fields .') = split(/\s+/, $line) ;',
  '  return '. $rules .' ;',
  '} ;'
) ;

my $test = eval $sub ;                    

print $sub, "\n" ;

foreach my $d (@DefectList) {
  print "$d\n"  if $test->($d) ;
} ;
[download]

[reply]
[d/l]