Caching a hash - am I missing something?

W. Tyler Gee wrote:
> On Tue, Aug 19, 2008 at =
5:35 PM, Chris Faust <cfaust@doyougot.com> wrote:
>> =
Hi,
>>
>>
>>
>> This might be a =
little off topic, I hope it's OK to post. I'm not positive
>> =
if mod_perl matters or not because it's a little confusing to =
me.
>>
>>
>>
>> I've taken over some =
pretty old code that I'm updating and making mp2
>> content =
handlers out of. The main script is a standard cgi script
>> =
"start.cgi" there is nothing special in the apache conf for =
it.
>>
>>
>>
>>   &nbs=
p;     <Directory =
/xxx/>
>>
>>      &nbs=
p;  SetHandler =
perl-script
>>
>>      &n=
bsp;  PerlFixupHandler =
My::Fixup
>>
>>      &nbs=
p;  PerlResponseHandler =
ModPerl::PerlRun
>>
>>     &nb=
sp;   PerlOptions =
+ParseHeaders
>>
>>      =
   DirectoryIndex =
start.cgi
>>
>>      &nbs=
p;  Options +ExecCGI =
+Indexes
>>
>>       =
;  allow from =
all
>>
>>       &nbs=
p; </Directory>
>>
>>
>>
>> =
start.cgi calls a custom module (use CustomModule;) which exports a =
bunch of
>> subs, for example foobar and all over the place in =
the subs that are
>> exported from CustomModule I see code =
like
>>
>>
>>
>>
>>
>&=
gt; sub foobar =
{
>>
>>        =
     my $key =3D =
@_;
>>
>>
>>
>>   &nbs=
p;         if ($cache{$key}) =
{
>>
>>        =
            &=
nbsp;    return =
$cache{$key};
>>
>>      =
       } else =
{
>>
>>        =
            &=
nbsp;    my $do_some_query =3D =
xxxx;
>>
>>       &n=
bsp;           &nb=
sp;     $cache{$key} =3D =
$do_some_query_results
>>
>>    &nb=
sp;           &nbs=
p;        return =
$cache{$key};
>>
>>      =
       }
>>
>> =
}
>>
>>
>>
>> My question is isn't =
the "else" in foobar always going to be true anyplace
>> where =
start.cgi is calling =
"&foobar('somekey')"??????
>>
>> I don't =
understand how %cache could already be populate from a =
previous
>> browser request or something? I'm I just missing =
something stupid?
>
> %cache is defined outside the scope of =
the sub so it will persist for
> the lifetime of the apache =
server.  The very first time
> foobar('somekey') is called it =
will do the query lookup, the next time
> it will return from =
cache.
>
I believe the above is almost, but not totally =
true.  It should probably
be "for the lifetime of this =
particular apache child".
Each apache child process has it's own copy =
of the above code, and it's
own copy of the above "global" (*) %cache =
hash.  Thus whether the first
or second part of the if will run, =
depends of the previous history of
the particular apache child which =
handles the current request.  If this
particular child has =
already previously accessed the same hash key, it
will server it from =
(it's own) cache, and otherwise it will execute the
query to create =
the key (in it's own cache).
Apache children are created, and die, as =
directed by the main Apache
process configuration, and HTTP requests =
are served more or less at
random, by whichever child is available =
when the request comes in.
Is is thus quite possible for instance =
that the first 10 requests for a
particular hash key would each be =
handled by a different apache child,
and would each result in a =
query; then the 11th request would be handled
by a child that has =
already accessed this same key before, and thus
served from cache; =
the 13th request would be handled by a brand-new
apache child just =
created, thus would re-do the query, etc..

I this would be =
useful, I believe that it would be possible to avoid
this, by =
"priming" the mod_perl module during the initial start of
Apache, =
before it forks into children.  Then each new child (being =
a
copy of the main apache) would start it's life with a number of =
keys
already in its cache.  Of course this would work only if =
the contents of
the keys of the hash are never modified while apache =
is running.
And it probably involves some delicate mod_perl =
programming to do the
priming process.

(*) like the previous =
contributor, I guess that the %cache hash is
somehow global, because =
it is not declared within the sub that you show.
  Whether it =
really is though, may depend on other code that we don't
see =
here.  But anyway, "global" would mean only "global to this =
apache
child", not to the whole apache server.

And something =
that I don't know at all, is how this all works out with a
threaded =
apache, such as under Windows =
e.g.

Andr=E9