ApacheBench Perl API changes - ver 0.6x

The ApacheBench Perl API has changed since versions 0.5x, such that
scripts that use the 0.5x API will not work correctly using 0.6x.
I apologize for the forced upgrade, but the 0.5x API was non-intuitive
and buggy.  The new 0.6x API will remain stable from now on.  Any future
changes will be accompanied by backwards compatibility fixes.

To upgrade scripts that were written for ApacheBench 0.5x:

1) When calling configuration methods, 'filesize' has been renamed to
   'buffersize'.  For individual runs, 'cookie' has been renamed to 'cookies'.

   The fix:
   - replace all $b->config({ filesize => XXX }) occurrences with
     $b->config({ buffersize => XXX }), or better yet: $b->buffersize(XXX)

2) The run configuration add() method no longer exists.  You now have to
   instantiate a new Run object and call add_run() on it.

   The fix:
   - replace all $b->add({ ... }) occurrences with:
     my $r = HTTPD::Bench::ApacheBench::Run->new({ ... });
     $b->add_run($r);

3) The execute() method no longer returns a regression hash, it returns
   an HTTPD::Bench::ApacheBench::Regression object.  The old regression
   hash can be found in this object's variable 'regression'.

   The fix:
   - replace all $re = $b->execute() with:
     $re = $b->execute();
     $rhash = $re->{'regression'};

4) The regression hash no longer contains any per-URL information.
   Half of the data keyed on URL was wrong, and the other half was
   fairly useless, so it is no longer returned at all.

   The fix:
   - remove all references to URLs in the regression hash, e.g.
     anything that looks like $re->{'run0'}->{'http://test.url/'}

5) What was stored in the 'threads' array reference (for each run),
   is now in the top-level of the regression hash for each run.
   The term threads was confusing, especially considering an upcoming
   rewrite will probably include multi-threaded request sending.

   The fix:
   - replace all $re->{'run0'}->{'threads'} with $rhash->{'run0'}

   Notice that each runI entry in the regression hash is now an ARRAY ref, not
   a HASH ref.  Your Perl script will fail unless you treat them as such.

6) Read the documentation.  There is a much more intuitive way of
   getting regression data using object method calls, rather than poking
   around through the huge regression hash.  Configuration can also now be
   done with method calls.

--
Adi Fairbank
February 23, 2001