in reply to Re: Hacking Perl Code
in thread Hacking Perl Code

thanks LanX, as suggested I added a link to this post in the original one.

I also added an update to the same post to explain better my optimistic sentence parsing code without executing it. Doing this i discovered that my solution executes even less code than the perl -c -d:TraceUse script.pl one.

HtH
L*
There are no rules, there are no thumbs..
Reinvent the wheel, then learn The Wheel; may be one day you reinvent one of THE WHEELS.

Replies are listed 'Best First'.
Re^3: Hacking Perl Code
by LanX (Saint) on Apr 09, 2014 at 10:48 UTC
    > Doing this i discovered that my solution executes even less code than the perl -c -d:TraceUse script.pl one.

    Could you please elaborate what you mean with less code?

    Can you provide an example showing the difference?

    TraceUse hooks into @INC with a callback like recently discussed.

    With this approach its even possible to report before anything is included and to avoid the import.

    Cheers Rolf

    ( addicted to the Perl Programming Language)

      Dear LanX,
      i only meant, as explained in the update of the original post that my solution only executes any BEGIN block and no more, while with perl -c -d:TraceUse script.pl all the BEGIN UNITCHECK CHECK blocks are executed.
      I was wrong (so thanks for your question), because my code also executes any END blocks and even other UNITCHECK already present, because those blocks run LIFO.
      Here some code to demonstrate the concept. With a script called test_of_executed_code.pl as follow:
      BEGIN{print qq(1-begin\n)}; UNITCHECK {print qq(2-unitcheck\n)}; CHECK {print qq(3-check\n)}; INIT {print qq(4-init\n)}; print qq(5-main\n); END{print qq(6-end\n)};
      You can now compare the two solutions:
      some_win>perl -c -d:TraceUse test_of_executed_code.pl 1-begin 2-unitcheck 3-check Modules used from test_of_executed_code.pl: test_of_executed_code.pl syntax OK
      while with my script you'll see:
      some_win>perl used_modules.pl test_of_executed_code.pl 1-begin 2-unitcheck strict 1.04 warnings 1.12 6-end
      The only way i found to prevent the END blocks to be executed is to change, in the UNITCHECK part of the original script, exit; with an ugly exec qq(perl -e "exit"); that produces:
      some_win>perl used_modules.pl test_of_executed_code.pl 1-begin 2-unitcheck strict 1.04 warnings 1.12
      But, considering the LIFO aptitude of the UNITCHECK blocks, i can do it better and change eval $begin.$content; with eval $content.$begin; producing:
      some_win>perl used_modules.pl test_of_executed_code.pl 1-begin strict 1.04 warnings 1.12
      Thanks for the opportunity to learn something more.
      L*

      There are no rules, there are no thumbs..
      Reinvent the wheel, then learn The Wheel; may be one day you reinvent one of THE WHEELS.
        Well OK I see, you exit before CHECKs are executed and -c does all UNITCHECKs and CHECKs.

        But please be aware that any BEGIN and UNITCHECK within a required module is executed, so you are only disabling CHECK and other UNITCHECKS on the top level script.

        As I said, IMHO does Devel::TraceUse have the more flexible and extendable approach, because it catches all includes before they are done (and immediately printing a nested tree output)

        You are parsing the (hopefully unmanipulated) entries from %INC at the top-level.

        Both approaches have their pro and cons... :)

        update

        your approach is not much different from a source filter, so if you consider to release it as a use-able module you could implement it that way. But please be aware that both approaches have to deal with trailing __DATA__ and __END__ sections when appending (instead of prepending) a UNITCHECK.

        Cheers Rolf

        ( addicted to the Perl Programming Language)