note ambrus <blockquote> IMO, the DB hack is a cleaner solution, because it is faster, uses a documented feature ... </blockquote> <P> I don't think any of them would be much faster than the other. The source filter is documented too. <blockquote> and has clean syntax for using it. To achieve such clean syntax with a source filter, you need to write a complex regex. </blockquote> <P> It's not that simple. The difference is that the code I gave above does the include in compile time, and its syntax is <code>use Filter::Include "file"</code> (the do is not needed when including complete statements). The DB way includes the code at run-time, that's why it's possible to use a simple subroutine <code>include "file"</code> is possible. While it is indeed not possible to make my solution work with such a simple syntax, without actually interpreting the code (incidentally that's what the actual [cpan://Filter::Include] cpan module does); if you wanted to modify the DB solution so that it includes the code in compile time (which can be a difference in semantics, depending on what the include file contains), you'll have to use a <code>use</code> or <code>BEGIN</code> syntax too, or try to parse the code. <blockquote> Besides that, source filters don't work everywhere (like in eval)... </blockquote> That's true. More generally, source filters can be used only at compile-time, not runtime. Also, source filters can not be used from command line (-e) it seems. <blockquote> and I really think the included file should by itself be syntactically correct. Including code is bad, but including partitial expressions is, IMHO, even worse. </blockquote> <P> True. I just wanted to show that this is really including the file. <P> Finally let me note that some include facility is already built in perl: the -P switch. If the file <b>third</b> contains <code> #!perl -w use warnings; use strict; my(@a, @b); @a = ( 1, 2, #include "./second" 7, 8); print "a(@a) b(@b)\n"; __END__ </code> and you run it with <code>perl -P third</code>, you get the same results. Of course, [perlrun] warns you that there are lots of problems with the -P switch. 393426 393465