Sorting bug in fo/pdf2index

["Paul A. Hoadley" <paulh@logicsquad.net>]

Hello,

I've been using fo/pdf2index to clean up FO indexes.  The results are
good, with the exception that pdf2index seems to be sorting numbers
lexically rather than numerically.  The culprit is in 'sub rangesort',
and, in fact, the following tiny patch will fix it:


--- pdf2index.orig      Wed Mar 19 20:56:11 2003
+++ pdf2index   Wed Mar 19 20:56:25 2003
@@ -114,7 +114,7 @@
     # Make sure roman pages come before arabic ones, otherwise sort them in order
     return -1 if ($apno !~ /^\d+/ && $bpno =~ /^\d+/);
     return  1 if ($apno =~ /^\d+/ && $bpno !~ /^\d+/);
-    return $apno cmp $bpno;
+    return $apno <=> $bpno;
 }
 
 sub rangelen {


I will file this at SourceForge.


-- 
Paul.

mailto:paulh@logicsquad.net
mailto:phoadley@maths.adelaide.edu.au