Front page | perl.perl5.porters |
Postings from November 2011
> From: Leon Timmermans
> On Thu, Nov 24, 2011 at 9:40 AM, Konovalov, Vadim wrote:
> > 1) as everyone knows "threading" should make "use" under better control - namely after threads creation.
> > How this will be handled?
>
> Why would it be different than now?
It is well-known adviced practice to use heavy "use something"
before any threads creation.
I am just wondered, this automated
use charnames; # :full or whatever
- will it be done before threads creation, or at what moment it will be
called WRT threads?
>
> > 2) PAR packer already generates huge executables (compared
> to e.g 5.6) because it happens now to load Math::BigInt
> everytime, due to some unfortunate dependency on some $^O.
> > Under new circumstances, the problem will be even worse.
>
> Again, I don't see this being different from now. If you need
> charnames it should be bundled, otherwise not.
Have you seen how modules dependency in PAR resolved?
There is some heuristic with false positives, and mostly the logic is to
package more modules than needed, to avoid any misses in packed
executables, so resulting in bigger executables than needed.
But with proposed new feature things will end up with adding
"use charnames;" to all PAR-packed executables, thus resulting
already oversized executables to be even bigger.
Needlessly bigger.
This is my concern.
>
> > I, personally, very much like for modules usage to be under
> better control, i.e. load what I demand.
> >
> > I think instead of loading of something when \N{} is seen
> would be a proper advice from warning system ("you probably
> need to do "use charnames::blabla" module")
>
> Telling the user "we know what you wanted from us, be we're not doing
> it" seems terribly user-unfriendly to me.
not to me :)
Yet, failed attempts to guess on what user needs is even worse.
Often this is unnoticeable waste, but in many situations this is hateful.
Regards,
Vadik.
Thread Previous
|
Thread Next