html-template-users

[htmltmpl] H-T strangeness with CGI-Session

[Puneet K. <pk...@ei...>]

Two apologies --

1. goofy subject line, because I couldn't think of a concise, precise 
description.

2. might be considered OT, but it seems H-T is a part of this 
strangeness (and there seems to be no active mailing list for 
CGI-Session), so...



I am using CGI-Session to write to a session file a record of pages 
visited by a user. The application is made with H-T with 
loop_context_vars turned on.

As a page is visted, its identifying info and a descriptive text is 
stored in an AoH that is used to build a widget so the user may visit 
recently seen pages, etc.

The problem is thus -- I am storing only the identifying info and a 
descriptive text. But the AofH in the Session file is showing additional 
info that looks suspiciously like the loop_context_vars. Here is an 
excerpt from my Session file --

"rolodex" => [{"ROLODEXVAL" => 3330,"__last__" => 0,"ROLODEXTEXT" => 
"Puneet Kishor (My Org, Inc.)","__counter__" => undef,"__first__" => 
0,"__odd__" => 0,"__inner__" => 0}],


See... I only meant to store ROLODEXVAL and ROLODEXTEXT. Where the heck 
did the __counter__, __first__, etc. come from?

Here is the relevant code --

     # Create rolodexval
     my $rolodexval = $res->[0]->{'contact_id'};

     # Create rolodextext
     my $fn= $res->[0]->{'firstname'};
     my $ln= $res->[0]->{'lastname'};
     my $fullname = ($fn or $ln) ? "$fn $ln" : 'unnamed';
     my $org = $res->[0]->{'org'};
     my $rolodextext = "$fullname ($org)";

     # A hash to hold the current rolodex entry
     my %rolodexentry = (
	ROLODEXVAL => $rolodexval,
	ROLODEXTEXT => $rolodextext
     );

     # Add the current rolodexentry to the existing rolodex
     push(@rolodex, \%rolodexentry);

    # Store the entire AoH in the session
    $session->param("rolodex", \@rolodex);

Re: [htmltmpl] H-T strangeness with CGI-Session

[<m.s...@gm...>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi Puneet, hi list

You are on the right list, it's an H::T issue and has nothing to with
CGI::Session. I discovered the same some weeks ago.

It seems than H::T can contaminate your data structure - for the sake
of speed, no copying is done in H::T. If it really bothers you, you
might consider using a clone package (try
http://search.cpan.org/search?query=clone&mode=all) to first copy
your data structure, then let H::T work on the clone and discard it
after displaying.

Kind regards - Markus
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (GNU/Linux)
Comment: Processed by Mailcrypt 3.5.6 and Gnu Privacy Guard <http://www.gnupg.org/>

iD8DBQFAZJ1txxUzQSse11ARAt0yAJ40Ttc/pymQYuzsYuSi/yHesvoK8QCeOEe6
OuQqmIfNBzj6FhPNStx9+TU=
=1d+T
-----END PGP SIGNATURE-----

Re: [htmltmpl] H-T strangeness with CGI-Session

[Puneet K. <pk...@ei...>]

m.s...@gm... wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Hi Puneet, hi list
> 
> You are on the right list, it's an H::T issue and has nothing to with
> CGI::Session. I discovered the same some weeks ago.
> 
> It seems than H::T can contaminate your data structure - for the sake
> of speed, no copying is done in H::T. If it really bothers you, you
> might consider using a clone package (try
> http://search.cpan.org/search?query=clone&mode=all) to first copy
> your data structure, then let H::T work on the clone and discard it
> after displaying.
> 


Markus,

Thanks for the reply. This is very informative, and I will follow your 
suggestion (and hope the H-T developers will consider this as a problem 
to be solved).

But, my question is -- where on earth did I let H-T "contaminate" my 
AofH? I am hand-crafting my own AofH --

     # A hash to hold the current rolodex entry
     my %rolodexentry = (
     ROLODEXVAL => $rolodexval,
     ROLODEXTEXT => $rolodextext
     );

     # Add the current rolodexentry to the existing rolodex
     push(@rolodex, \%rolodexentry);

Where, o, where is H-T even touching all this?

Thanks,

Puneet.

Re: [htmltmpl] H-T strangeness with CGI-Session

[Cees H. <ce...@si...>]

Puneet Kishor wrote:
> The problem is thus -- I am storing only the identifying info and a 
> descriptive text. But the AofH in the Session file is showing additional 
> info that looks suspiciously like the loop_context_vars. Here is an 
> excerpt from my Session file --
> 
> "rolodex" => [{"ROLODEXVAL" => 3330,"__last__" => 0,"ROLODEXTEXT" => 
> "Puneet Kishor (My Org, Inc.)","__counter__" => undef,"__first__" => 
> 0,"__odd__" => 0,"__inner__" => 0}],

This is probably because you are passing a reference to a data structure 
to both CGI::Session and HTML::Template.  In other words they both get 
the exact in memory data structure.  HTML::Template is adding things to 
the hashes that are stored in the array ref, and CGI::Session doesn't 
save it's parameters to the session until it goes out of scope (probably 
at the end of the request) so it sees the changes that HTML::Template 
made to the data.

There are a couple of things you can do to solve it.

- Don't pass the same data structure to both CGI::Session and 
HTML::Template.  Make a copy of the data before passing it by 
dereferencing the data (see the dclone() method in the Storable module 
for copying complex data structures if you have a deep data structure).

- Or you can make CGI::Session save it's data right away by making it go 
out of scope, or calling close or flush to sync the data to the session 
store.

Although HTML::Template is causing this problem, I don't think you cna 
call it a bug.  Although many modules will make a copy of the data that 
is passed in if they intend to make any changes, that can be very 
expensive if there is a lot of data (which there tends to be when 
dealing with templates).

Cheers,

Cees

Re: [htmltmpl] H-T strangeness with CGI-Session

[Puneet K. <pk...@ei...>]

Cees Hek wrote:

> Puneet Kishor wrote:
> 
>> The problem is thus -- I am storing only the identifying info and a 
>> descriptive text. But the AofH in the Session file is showing 
>> additional info that looks suspiciously like the loop_context_vars. 
>> Here is an excerpt from my Session file --
>>
>> "rolodex" => [{"ROLODEXVAL" => 3330,"__last__" => 0,"ROLODEXTEXT" => 
>> "Puneet Kishor (My Org, Inc.)","__counter__" => undef,"__first__" => 
>> 0,"__odd__" => 0,"__inner__" => 0}],
> 
> 
> This is probably because you are passing a reference to a data structure 
> to both CGI::Session and HTML::Template.  In other words they both get 
> the exact in memory data structure.  HTML::Template is adding things to 
> the hashes that are stored in the array ref, and CGI::Session doesn't 
> save it's parameters to the session until it goes out of scope (probably 
> at the end of the request) so it sees the changes that HTML::Template 
> made to the data.
> 
> There are a couple of things you can do to solve it.
> 
> - Don't pass the same data structure to both CGI::Session and 
> HTML::Template.  Make a copy of the data before passing it by 
> dereferencing the data (see the dclone() method in the Storable module 
> for copying complex data structures if you have a deep data structure).
> 
> - Or you can make CGI::Session save it's data right away by making it go 
> out of scope, or calling close or flush to sync the data to the session 
> store.
> 
> Although HTML::Template is causing this problem, I don't think you cna 
> call it a bug.  Although many modules will make a copy of the data that 
> is passed in if they intend to make any changes, that can be very 
> expensive if there is a lot of data (which there tends to be when 
> dealing with templates).
> 


That explains it. You hit it right on the nose, Cees. It is indeed 
happening because I am sending the same ref to the AofH. Thanks much for 
you advice as I now know how to solve it.

Puneet.