Here is a sample package and exported rule:
BeginPackage["PackageContext`"];
Rule1::usage = "Rule1 is a test exported rule.";
Begin["`Private`"];
Rule1 = Cos[x_]^2 + Sin[x_]^2 -> 1;
End[];
EndPackage[];
Then, if we evaluate Rule1 we obtain a very clumsy version of the rule,
although I suppose it will work.
Rule1
Cos[PackageContext`Private`x_]^2 + Sin[PackageContext`Private`x_]^2 -> 1
What I would like is the rule as written in the package. Is there a way to
do this?
David Park
<http://home.comcast.net/~djmpark> http://home.comcast.net/~djmpark/
Try with
BeginPackage["PackageContext`"];
Rule1::usage = "Rule1 is a test exported rule.";
Begin["`Private`"];
Rule1 = Cos[Global`x_]^2 + Sin[Global`x_]^2 -> 1;
End[];
EndPackage[];
In the Global` context, Rule1 will display as
Cos[x_]^2+Sin[x_]^2->1
Ingolf Dahl
Sweden
ToExpression[
StringReplace[ToString[InputForm[Rule1]],
"PackageContext`Private`" -> ""]]
Sin[x_]^2 + Cos[x_]^2 -> 1
Bob Hanlon
---- David Park <djm...@comcast.net> wrote:
=============
what about using the global symbol?
BeginPackage["PackageContext`"];
Rule1::usage = "Rule1 is a test exported rule.";
Begin["`Private`"];
Rule1 = Cos[Global`x_]^2 + Sin[Global`x_]^2 :> Global`x;
End[];
EndPackage[];
Cheers
Patrick
There are a number of issues I would like to raise because I'm not certain
of the best procedure if we want a package to generate an expression with
new symbols that are in the Context of the notebook using the package. There
are two solutions that might be useful but are unavailable.
1) WRI might be able to create a floating context, which had the property
that it took on the context of the notebook or cell in which it is created.
This could then be used by package writers to return created symbols.
2) WRI might provide a way for a package routine to retrieve the current
Context in the InputNotebook[]. This could then be used actively in
generating new symbols.
Otherwise we can obtain "Context free" output by putting the Rule symbols
into existing Contexts such as the Package Context itself, or the System`
context. But is it good practice to usurp simple symbols such as x, y, z
into some special Context? Maybe it is all right if users always read in the
package first. But suppose a number of packages start doing this?
In any case, here is the package routine again in a form that statically
determines the Context at the time it is read in. The problem is that if the
user changes evaluation from one notebook to another, that have different
Contexts, then the package must be read in anew. I am taking advantage here
of the fact that SymbolName strips any leading Context from a symbol so we
can just attach the new Context.
System`loadcontext=Context[];
BeginPackage["PackageContext`"];
Rule1::usage="Rule1 is a test exported rule.";
Begin["`Private`"];
cs[var_Symbol]:=var->Symbol[System`loadcontext<>SymbolName[var]]
cs[vars:{__Symbol}]:=cs/@vars
Rule1=(y_ Sin[x_ y_]/;AtomQ[y]->{x,y})//.cs[{x,y}];
End[];
EndPackage[];
Test:
Rule1
a Sin[a b] /. Rule1
y_ Sin[x_ y_] /; AtomQ[y] -> {x, y}
{b, a}
David Park
djm...@comcast.net
http://home.comcast.net/~djmpark/
From: David Park [mailto:djm...@comcast.net]
> I would like to export rules from a package in such a way that the pattern
> symbols did not include any contexts and in which I also did not have to
> export the symbols used in the pattern. In addition, If possible, I would
> like this to work even if the package was loaded from a notebook that had
> something other than the Global` context.
>
>
>
> Here is a sample package and exported rule:
>
>
>
> BeginPackage["PackageContext`"];
>
>
>
> Rule1::usage = "Rule1 is a test exported rule.";
>
>
>
> Begin["`Private`"];
>
>
>
> Rule1 = Cos[x_]^2 + Sin[x_]^2 -> 1;
>
>
>
> End[];
>
>
>
> EndPackage[];
>
>
>
> Then, if we evaluate Rule1 we obtain a very clumsy version of the rule,
> although I suppose it will work.
>
>
>
> Rule1
>
> Cos[PackageContext`Private`x_]^2 + Sin[PackageContext`Private`x_]^2 -> 1
>
>
>
> What I would like is the rule as written in the package. Is there a way to
> do this?
There is a problem with your expectation: I think every symbol is per
definition in a context, so it is strictly not possible to define a
symbol which "does not include a context". Whether you see the "clumsy
version" or just the symbol name depends on whether the symbols context
is part of $ContextPath at the time the output is shown or not.
Considering this one way to achieve something that behaves as you
probably want is this:
Rule1 := Block @@ Join[
List /@ ToExpression["x", InputForm, Hold],
Hold[(Cos[y_]^2 + Sin[y_]^2 -> 1) /. y -> Symbol["x"]]
]
What this does is to create the symbol x not at the time the rule is
defined but only when the rule is called. So it will create the symbol x
in the context where Rule1 is used and hence should in all but very
exotic cases be shown as just x without its context shown. Since you
don't know the context where the x will be created it makes sense to
ensure it will not cause problems if there are definitions for this new
x. This makes the whole definition somewhat clumsy itself, but the user
will usually not see it.
The whole procedure is of course rather inefficient since it will run
the above code every time you call it. I don't know if this is worth the
advantage of a "nicer looking output" in your case, but depending on the
purpose of the package efficiency might not be so crucial...
hth,
albert
I'd try something like:
BeginPackage["PackageContext`"];
Rule1::usage = "Rule1 is a test exported rule.";
Begin["`Private`"];
Rule1 := (Cos[x_]^2 + Sin[x_]^2 -> 1)/.x->Unique["x"];
End[];
EndPackage[];
so you get something like
Rule1
Cos[x1_]^2 + Sin[x1_]^2 -> 1
cause the symbol is created only when the Rule1 is evaluated
David. I think you could do something like
Symbol[$Context<>"x"] to make the symbol belong to the current context
when the function is called.
However, I guess it makes sense to put it inside a block if you want
it for output and don't want conflics.
I gotta go now, but I think it should be simple to make a function
that behaves
SymbolOut[{list of symbols}, expression]
that treats every symbol inside the list of symbols as a symbol
belonging to the context where it was evaluated, taking local values
to prevent conflicts. I guess with a With to replace the symbols by
Symbol[$Context<>ToString[symbolss]] and a Block to localize them it
should be possible. I'll try it when I'm back.
That seems like a good solution. Many thanks for suggesting it.
Here is a new test package. newsymbols will be a list of rules for all of
the symbols used in exported rules. One precaution is that the definitions
have to be delayed so that $Context is not evaluated in the package.
This construction also works if one wishes to generate an expression in some
variable and we wish to give the variable a default value.
BeginPackage["PackageContext`"];
Rule1::usage="Rule1 is a test exported rule.";
Routine1::usage="Routine1[] returns a test polynomial with a default
variable x.";
Begin["`Private`"];
newsymbols:={x->Symbol[$Context<>"x"],y->Symbol[$Context<>"y"]};
Rule1:=(y_ Sin[x_ y_]/;AtomQ[y]->{x,y})//.newsymbols;
Routine1[x_:Automatic]:=
Module[{var},
var=If[x===Automatic,Symbol[$Context<>"x"],x];
var+var^2];
End[];
EndPackage[];
Rule1
a Sin[a b] /. Rule1
Routine1[]
I tested this with a PackageContext`m file, reading it in both in a regular
Global` context notebook, and in a notebook with its own Notebook context. I
could go back and forth and it evaluated properly in both notebooks and the
x,y symbols were in their respective notebook contexts.
Another solution for rules, suggested to me by Maxim Rytin, is to use Formal
symbols:
formalsymbols = {x -> \[FormalX], y -> \[FormalY]}
Rules seem to be a proper place to use Formal symbols because they are never
given values and being in the System` Context display without Context
information. But these were introduced in Version 6 and may not be familiar
to many users. They have gray dots above and below the characters. They are
not suitable for default variables because we may wish to later give those
values.
From: Rui [mailto:rui....@gmail.com]
On Apr 27, 5:04 am, "David Park" <djmp...@comcast.net> wrote:
> I want to thank Ingolf Dahl, Bob Hanlon and Patrick Scheibe for their
> answers. Using the Global` context is a good solution, except if the
package
> had been loaded in something other than the Global` context, say on a
> Function page. Also I would like to have a simple way to process simply
> written rules in the package such that it would return simple looking and
> usable rules to the user.
>
David. I think you could do something like