Hi,

you have been given some good solutions, I won't give you any other. I would just like to make some comments on your code.

my ($TEXTIN,$HTMLOUT);
my $input2 = $outputfile;
my $output2 = "FTFIMS.html";
my @records=();
my $inrecord=0;
my $rxRecStart = qr{STORE\d+};
my $rxRecStop = qr{\n\s};
my $recordStr = q{};
[download]

It is not a very good idea to declare all your lexical variables at the top of your program, because you are essentially making them global to the whole file and this negates a large part of the advantages of lexical variables. Try to limit scope of variables to the enclosing block where they belong.

This will not die if the program fails to open the file (say, if the file does not exists), because of precedence problems.

You should either have parens:

or use the lower precedence operator or:

But it would even be better to declare your filehandler within that statement:

Same thing of course for the other file opening statement.

Otherwise, I think that the algorithm within your while loop is too complicated, error prone and not very robust. Especially, the $rxRecStop regexp is very weak and might match where you don't expect. Also, it will probably not match anything at the end of the file, so that the last section will not be recorded.

Rather than having a beginning and end regexp, I think it would probably be better to have only one break regexp (the /STORE\d{3}/ is a good candidate). When you meet it, you do the house cleaning of the previous section (storing data) and the preparation of the next section (reinitializing the variables). Something like this:

my $header = "";
$header .= <$TEXTIN> for 0..1; # record the first two lines for later 
+use
while(<$TEXTIN>){
     next if m/^\s*$/; # get rid of empty lines
     if (m{$rxRecStart}){
          # do what you need to finish off the previous section
          # (saving the data) and start the new one
     } else {
         $recordStr .= $_;
     }
}
# add code here for storing the last section
[download]

I just wanted to propose some improvements on the basis of your code, to help you think about it. I think that the solution with the modification of the input record separator proposed by Rolf and others is probably better.