Builtin `Begin` not adding variable in context and `End` leaves traces of variables defined by `Begin[]`

[rocky]

Description

Begin is not adding context, and End might not be removing variables added.

How to Reproduce and Output Given

In[1]:= ??f
Out[1]= Global`f

In[2]:= Begin["MyContext`"]
Out[2]= "MyContext`"

In[3]:= f[x_] := x^2

In[4]:= ??f
Out[4]= Global`f  (* Wrong. Should be MyContext` *)       
        f[x_] = x^2

In[5]:= End[]
Out[5]= "MyContext`"

In[6]:= ??f
Out[6]= Global`f (* Wrong. Should have been removed *)
        
        f[MyContext`x_] = MyContext`x^2

Additional context

Noticed while trying to understand why certain Mathematica packages are failing.

[mmatera]

The problem seems to be in Information: it defines the symbol, which hides the MyContext definition. In WMA:

In[1]:= ?? f                                                                    

Out[1]= Missing[UnknownSymbol, f]

In[2]:= Begin["MyContext`"]                                                     

Out[2]= MyContext`

In[3]:= f[x_]=2 x                                                               

Out[3]= 2 x

In[4]:= End[]                                                                   

Out[4]= MyContext`

In[5]:= ?? f                                                                    

Out[5]= Missing[UnknownSymbol, f]

In[6]:= ?? MyContext`f                                                          

Out[6]= MyContext`f
        MyContext`f[MyContext`x_] = 2*MyContext`x
         

In[7]:= f[z_]:=3                                                                

In[8]:= ?? f                                                                    

Out[8]= Global`f
        f[z_] := 3

I am planning to fix this over the weekend.

[mmatera]

@rocky, if I am right, what is reported here is not an issue of the Begin/BeginPackage implementations, but of how ?? ... is parsed as an expression.

[rocky]

@rocky, if I am right, what is reported here is not an issue of the Begin/BeginPackage implementations, but of how ?? ... is parsed as an expression.

@mmatera ?? parsing is not the essential problem reported here, but instead a side effect of how I was conveying the problem.

In[1]:= ?? f                                                                    
Out[4]= Global`f

is indeed a problem with the prefix operator ??. However, I just added that to try demonstrate that the variable wasn't previously defined. There is a second bug that is what the recent PRs have been focusing on.

The import part is of this bug is what followed:

In[2]:= Begin["MyContext`"]                                                     
Out[2]= MyContext`

In[3]:= f[x_]=2 x    
Out[3]= 2 x

Now, if we continue and ask about f:

In[4]:= ??f
Out[4]= Global`f
        
        f[x_] = 2*x

we see that it got defined in Global, not MyContext.

[mmatera]

Replacing the ?? query by an explicit call to Information, using a string as an argument, do you get the same result?

[rocky]

Replacing the ?? query by an explicit call to Information, using a string as an argument, do you get the same result?

Interesting! I get the correct result using Information[f].

Hmm, what's up here?

[mmatera]

What happens is that each time you use a symbol in an expression, it gets defined. If you use Information[f] (with the symbool, not a string) in WMA, you also get the same issue.

[rocky]

I've redone using Information calls, and I am still seeing a bug:

In[1]:= Information["f"]
Out[1]= Missing[UnknownSymbol, f]

In[2]:= Begin["MyContext`"]
Out[2]= "MyContext`"

In[3]:= f[x_] := x^2

In[4]:= Information["f"]
Out[4]= Global`f
        
        f[x_] = x^2

Now in WMA:

In[1]:= Information["f"]                                               

Out[1]= Missing[UnknownSymbol, f]

In[2]:= Begin["MyContext`"]                                            

Out[2]= MyContext`

In[3]:= f[x_] := x^2                                                   

In[4]:= Information["f"]                                               

Out[4]= MyContext`f
        f[x_] := x^2

[rocky]

And furthermore, in Mathics3:

In[5]:= Information["MyContext`f"]
Out[5]= Missing[UnknownSymbol, f]

versus WMA:

In[5]:= Information["MyContext`f"]                                     

Out[5]= MyContext`f
        f[x_] := x^2

[rocky]

Just to be clear. The original problem I was seeing and trying to report was that on an EndPackage, I am getting:

EndPackage::noctx: No previous context defined.

The stuff with Information was to try to understand how to isolate the problem further.

Ok, so there are bugs in Information.

But given the above, the data is consistent with the symbol f getting put in the Global context rather than in the MyContext context. If that's the case, then when we get to EndPackage, it is not finding the package name variable in MyContext.

[mmatera]

In[1]:= Information["f"]

Again, the behavior seems to be the expected one if we do not execute the first line. When Information is evaluated, the symbol is created, opposite to what is expected in WMA.

[mmatera]

Just to be clear. The original problem I was seeing and trying to report was that on an EndPackage, I am getting:

EndPackage::noctx: No previous context defined.

This could be another problem. This error happens if we call End[] more times than the times we call Begin[]. But I am not being able to reproduce this error with the example in this issue.