Re: Second argument of BeginPackage, revisited
- To: mathgroup at smc.vnet.net
- Subject: [mg78537] Re: Second argument of BeginPackage, revisited
- From: Andrew Moylan <andrew.j.moylan at gmail.com>
- Date: Tue, 3 Jul 2007 06:51:59 -0400 (EDT)
- References: <f5td66$36j$1@smc.vnet.net><f62k72$bod$1@smc.vnet.net>
Right, there is no irregularity in the value of $ContextPath after
executing EndPackage[]. As stated in the documentation, after the
EndPackage[] statement corresponding to BeginPackage["a`"], only "a`"
has been added to the original value of $ContextPath.
The irregularity that I am referring to is in the different values of
$ContextPath *between* the calls to BeginPackage[] and EndPackage[].
Compare the output of these two examples:
BeginPackage["a`", {"PhysicalConstants`"}]
$ContextPath
Output: {"a`", "PhysicalConstants`", "System`"}
BeginPackage["a`"]
Needs["PhysicalConstants`"]
$ContextPath
Output: {"PhysicalConstants`", "Units`", "a`", "System`"}
As Don DuBois has pointed out in a private message, the correct way to
obtain the behaviour of Needs[] when declaring packages is not to use
the second argument of BeginPackage, but instead to use Needs[]
*twice*, like this:
BeginPackage["a`"]
Needs["b`"]
(* code block 2 *)
EndPackage[]
Needs["b`"]
On Jul 2, 9:08 pm, dh <d... at metrohm.ch> wrote:
> Hi Andrew,
>
> you are correct, BeginPackage["a`",{"b`",..}].. EndPackage[] will put
>
> all the additional contexts "b`".. into $ContextPath. Therefore, all
>
> variables exported by these contexts will be available.
>
> On the other hand, BeginPackage["a`"] Needs["b`"];.. Endpackage[] will
>
> only put "a`" into $ContextPath, therefore, variables from "b`".. are
>
> not available afterwards.
>
> Daniel
>
> Andrew Moylan wrote:
> > On Jun 28, 6:37 pm, dh <d... at metrohm.ch> wrote:
> >> Hi Andrew,
>
> >> you misunderstood the Package concept. BeginPackage["a`", {"b`"}] will
>
> >> place b in the $ContextPath so that its content can be found, but it
>
> >> does not put b in Context[], so that new variables are not created in b.
>
> >> The following may help. Start a new session and enter:
>
> >> BeginPackage["b`",{"PhysicalConstants`"}]
>
> >> Print[SpeedOfLight];
>
> >> Print[Context[]];
>
> >> Print[$ContextPath];
>
> >> EndPackage[]
>
> >> hope this helps, Daniel
>
> >> Andrew Moylan wrote:
> >>> Unfortunately (in my opinion), BeginPackage["a`", {"b`"}] does not *quite*
> >>> call Needs[] on its second argument(s).
> >>> Suppose b.m looks like this:
> >>> BeginPackage["b`", {"PhysicalConstants`"}]
> >>> (* some code *)
> >>> EndPackage[]
> >>> Then calling Needs["b`"] in a new kernel results in "PhysicalConstants`"
> >>> being on the list of contexts, so that e.g. SpeedOfLight works as expected.
> >>> But calling BeginPackage["a`", {"b`"}] instead does *not* put
> >>> PhysicalConstants on the list of contexts during the definition of the
> >>> package, so that e.g. SpeedOfLight results in the creation of a new, useless
> >>> symbol called a`SpeedOfLight.
> >>> Finally, for maximum confusion (in my opinion), once EndPackage[] is called
> >>> to finish loading the "a`" package, "PhysicalConstants`" *is* placed on the
> >>> list of contexts, causing shadowing between a`SpeedOfLight and
> >>> PhysicalConstants`SpeedOfLight.
> >>> Effectively, BeginPackage appears to
> >>> 1. Call Needs[] on its second arguments, then
> >>> 2. Temporarily remove from the context path any extra contexts those
> >>> packages added to the context path, then finally
> >>> 3. Put all those temporarily removed packages back onto the context path
> >>> once EndPackage[] is called.
> >>> Can anyone help me understand why this behaviour is useful?
>
> > Daniel, you have misunderstood the question in my post.
>
> > Here is the crucial irregularity to which I refer:
>
> > BeginPackage["a`", {"b`"}]
> > (* code block 1 *)
> > EndPackage[]
>
> > BeginPackage["a`"]
> > Needs["b`"]
> > (* code block 2 *)
> > EndPackage[]
> > Needs["b`"]
>
> > The above two alternatives are not identical, because (strangely, in
> > my opinion) code in code block 1 cannot refer to symbols defined in
> > any packages declared in the second argument of the BeginPackage
> > statement for b` (in b.m); whereas code in code block 2 *can* refer to
> > such symbols.
>
> > Here is a different way to understand the irregularity:
>
> > Suppose this code works correctly:
>
> > Needs["b`"]
> > (* code block 3 *)
>
> > Then, will this code generally also work correctly?
>
> > BeginPackage["whatever`", {"b`"}]
> > (* code block 3 *)
>
> > The answer is no, because of the way BeginPackage apparently modifies
> > the value of $ContextPath after calling Needs[] on its second argument.
- Follow-Ups:
- Re: Re: Second argument of BeginPackage, revisited
- From: Carl Woll <carlw@wolfram.com>
- Re: Re: Second argument of BeginPackage, revisited