* [ruby-core:95764] [Ruby master Feature#16336] Allow private constants to be accessed with absolute references
[not found] <redmine.issue-16336.20191108205406@ruby-lang.org>
@ 2019-11-08 20:54 ` dylan.smith
2019-11-08 23:26 ` [ruby-core:95767] " merch-redmine
2019-11-14 18:21 ` [ruby-core:95851] " dylan.smith
2 siblings, 0 replies; 3+ messages in thread
From: dylan.smith @ 2019-11-08 20:54 UTC (permalink / raw)
To: ruby-core
Issue #16336 has been reported by dylants (Dylan Thacker-Smith).
----------------------------------------
Feature #16336: Allow private constants to be accessed with absolute references
https://bugs.ruby-lang.org/issues/16336
* Author: dylants (Dylan Thacker-Smith)
* Status: Open
* Priority: Normal
* Assignee:
* Target version:
----------------------------------------
The purpose of constant privacy is to keep a constant from being accessed from outside the namespace. As such, it can be surprising at first when something like the following doesn't work
```ruby
# in foo/a.rb
module Foo
class A
end
private_constant :A
end
# in foo/b.rb
module Foo
class B
Foo::A # => NameError: private constant Foo::A referenced
end
end
```
Once the unexpected behaviour is understood, it can be worked around by using a relative constant reference (changing `Foo::A` to `A` in the example). However, the developer isn't just writing the code for the machine to understand, they are also writing it so it will be clear to other developers, where using the parent namespace in the constant reference may be appropriate for the readability of the code.
Instead, I think a reference to a private constant should be allowed through its parent namespace (as is the case with absolute constant references) if the parent namespace is in `Module.nesting` and the constant reference isn't made through a dynamic value (e.g. `foo = Foo; foo::A` wouldn't be allowed in the above example).
This will have the additional advantage that it will be easier to make a constant private, since these private constant references through their parent namespace won't have to be updated to remove the parent namespace from the constant reference.
I've opened pull request https://github.com/ruby/ruby/pull/2664 with a fix for this issue.
--
https://bugs.ruby-lang.org/
^ permalink raw reply [flat|nested] 3+ messages in thread
* [ruby-core:95767] [Ruby master Feature#16336] Allow private constants to be accessed with absolute references
[not found] <redmine.issue-16336.20191108205406@ruby-lang.org>
2019-11-08 20:54 ` [ruby-core:95764] [Ruby master Feature#16336] Allow private constants to be accessed with absolute references dylan.smith
@ 2019-11-08 23:26 ` merch-redmine
2019-11-14 18:21 ` [ruby-core:95851] " dylan.smith
2 siblings, 0 replies; 3+ messages in thread
From: merch-redmine @ 2019-11-08 23:26 UTC (permalink / raw)
To: ruby-core
Issue #16336 has been updated by jeremyevans0 (Jeremy Evans).
This would make private constant handling inconsistent with private method handling:
```
class A
def self.b
1
end
private_class_method :b
A.b
# NoMethodError (private method `b' called for A:Class)
end
```
That by itself isn't a reason to reject this feature, but it is something that should be considered.
----------------------------------------
Feature #16336: Allow private constants to be accessed with absolute references
https://bugs.ruby-lang.org/issues/16336#change-82583
* Author: dylants (Dylan Thacker-Smith)
* Status: Open
* Priority: Normal
* Assignee:
* Target version:
----------------------------------------
The purpose of constant privacy is to keep a constant from being accessed from outside the namespace. As such, it can be surprising at first when something like the following doesn't work
```ruby
# in foo/a.rb
module Foo
class A
end
private_constant :A
end
# in foo/b.rb
module Foo
class B
Foo::A # => NameError: private constant Foo::A referenced
end
end
```
Once the unexpected behaviour is understood, it can be worked around by using a relative constant reference (changing `Foo::A` to `A` in the example). However, the developer isn't just writing the code for the machine to understand, they are also writing it so it will be clear to other developers, where using the parent namespace in the constant reference may be appropriate for the readability of the code.
Instead, I think a reference to a private constant should be allowed through its parent namespace (as is the case with absolute constant references) if the parent namespace is in `Module.nesting` and the constant reference isn't made through a dynamic value (e.g. `foo = Foo; foo::A` wouldn't be allowed in the above example).
This will have the additional advantage that it will be easier to make a constant private, since these private constant references through their parent namespace won't have to be updated to remove the parent namespace from the constant reference.
I've opened pull request https://github.com/ruby/ruby/pull/2664 with a fix for this issue.
--
https://bugs.ruby-lang.org/
^ permalink raw reply [flat|nested] 3+ messages in thread
* [ruby-core:95851] [Ruby master Feature#16336] Allow private constants to be accessed with absolute references
[not found] <redmine.issue-16336.20191108205406@ruby-lang.org>
2019-11-08 20:54 ` [ruby-core:95764] [Ruby master Feature#16336] Allow private constants to be accessed with absolute references dylan.smith
2019-11-08 23:26 ` [ruby-core:95767] " merch-redmine
@ 2019-11-14 18:21 ` dylan.smith
2 siblings, 0 replies; 3+ messages in thread
From: dylan.smith @ 2019-11-14 18:21 UTC (permalink / raw)
To: ruby-core
Issue #16336 has been updated by dylants (Dylan Thacker-Smith).
https://bugs.ruby-lang.org/issues/16123 added support for making private methods calls explicitly through `self`, so in class `A` `self.b` could be used to call a private class method `b`. If consistency with private methods is desired, then we could support referring the `self` through a static constant reference, although I can't think of any reason why that would be useful.
----------------------------------------
Feature #16336: Allow private constants to be accessed with absolute references
https://bugs.ruby-lang.org/issues/16336#change-82685
* Author: dylants (Dylan Thacker-Smith)
* Status: Open
* Priority: Normal
* Assignee:
* Target version:
----------------------------------------
The purpose of constant privacy is to keep a constant from being accessed from outside the namespace. As such, it can be surprising at first when something like the following doesn't work
```ruby
# in foo/a.rb
module Foo
class A
end
private_constant :A
end
# in foo/b.rb
module Foo
class B
Foo::A # => NameError: private constant Foo::A referenced
end
end
```
Once the unexpected behaviour is understood, it can be worked around by using a relative constant reference (changing `Foo::A` to `A` in the example). However, the developer isn't just writing the code for the machine to understand, they are also writing it so it will be clear to other developers, where using the parent namespace in the constant reference may be appropriate for the readability of the code.
Instead, I think a reference to a private constant should be allowed through its parent namespace (as is the case with absolute constant references) if the parent namespace is in `Module.nesting` and the constant reference isn't made through a dynamic value (e.g. `foo = Foo; foo::A` wouldn't be allowed in the above example).
This will have the additional advantage that it will be easier to make a constant private, since these private constant references through their parent namespace won't have to be updated to remove the parent namespace from the constant reference.
I've opened pull request https://github.com/ruby/ruby/pull/2664 with a fix for this issue.
--
https://bugs.ruby-lang.org/
^ permalink raw reply [flat|nested] 3+ messages in thread