Hello,
I'm running into a bit of confusion between the results I'm getting and
what the docs on the web site read, specifically:
parse
$req->parse()
Forces the request to be parsed immediately. In void context, this will throw
an APR::Request::Error should the either the query-string or body parser fail.
In all other contexts it will return the two parsers' combined APR status code
$req->body_status || $req->args_status
However parse should be avoided in most normal situations. For example, in a
mod_perl content handler it is more efficient to write
sub handler {
my $r = shift;
my $req = Apache2::Request->new($r);
$r->discard_request_body; # efficiently parses the request body
my $parser_status = $req->body_status;
#...
}
Calling $r->discard_request_body outside the content handler is generally a
mistake, so use $req->parse there, but only as a last resort. The Apache2::Request
API is designed around a lazy-parsing scheme, so calling parse should not affect the
behavior of any other methods.
The results of which are "Not Good!(tm)", or so it seems. I reduced it to a
very
example of what's going on, which I'm going to C&P here so you can look at it.
The
basic problem is if that suggestion is followed, the POST parameters get lost
when
you issue a subrequest!!
As I read the docs, issuing the discard should slurp in the rest of the params
and
hold them for the duration of this request. I put some debugging statements
into
the little package I made up and this is what I get out:
[Tue Dec 27 16:13:52 2005] [notice] Apache/2.2.0 (Fedora) configured --
resuming normal operations
n1: list: d, e, f, A, B, C, submit
n2: list: d, e, f
n3: list: d, e, f
n4: list: d, e, f
It matters little if I call discard_request_body on '_request' or '_query'.
The result
is always the same. When I comment out the call in the handler so that it is
never called,
I get what I suspect is the expected overall behavior:
[Tue Dec 27 16:10:04 2005] [notice] Apache/2.2.0 (Fedora) configured --
resuming normal operations
n1: list: d, e, f, A, B, C, submit
n2: list: d, e, f, A, B, C, submit
n3: list: d, e, f, A, B, C, submit
n4: list: d, e, f, A, B, C, submit
If I've missed something, please let me know. My purpose in calling that
routine is to
make sure that there's no leftovers from one request to another on KeepAlive
(persistant)
connections. If that's unnecessary in Apache2/ModPerl2/libapreq2 then I'll
drop the
method call and we'll call it a day. :)
Thanks again,
Michael
This is what .htaccess in / says:
PerlModule Test::HndlAdmin
<Files "test.html">
PerlResponseHandler Test::HndlAdmin virtual
SetHandler perl-script
</Files>
Coupled with this package:
package Test::HndlAdmin;
use Apache2::Request;
use Apache2::Const;
sub handler
{
my ($request) = shift;
my $self = bless({}, __PACKAGE__);
$self->{'_request'} = $request;
$self->{'_query'} = Apache2::Request->new($self->{'_request'});
$self->{'_query'}->discard_request_body;
$self->{'_request'}->content_type('text/html');
my $sub = $self->{'_request'}->lookup_uri('/header.html');
print STDERR "n1: list: " . join(', ', keys(%{scalar $self->{'_query'}->param})) .
"\n";
$sub->run;
print STDERR "n2: list: " . join(', ', keys(%{scalar $self->{'_query'}->param})) .
"\n";
print '<FORM METHOD=POST ACTION="/test.html?d=5&e=113&f=69">';
print '<INPUT TYPE=HIDDEN NAME="A" VALUE=7>';
print '<INPUT TYPE=HIDDEN NAME="B" VALUE=-4>';
print '<INPUT TYPE=HIDDEN NAME="C" VALUE=12>';
print '<INPUT TYPE=SUBMIT NAME="submit" VALUE="submit">';
print '</FORM>';
my $sub = $self->{'_request'}->lookup_uri('/footer.html');
print STDERR "n3: list: " . join(', ', keys(%{scalar $self->{'_query'}->param})) .
"\n";
$sub->run;
print STDERR "n4: list: " . join(', ', keys(%{scalar $self->{'_query'}->param})) .
"\n";
return Apache2::Const::OK;
}
1;
The header.html and footer.html files contain only a <BODY> & </BODY> tag
respectively.
My tests here show it doesn't much matter what you put in those two files I
tried simple
SGML comment in each to our main site's rather extensive mod_including calling
files.
