Should Perl introduce a language-level private attribute for subroutines?

Answer

Yes, Perl could benefit from a language-level private attribute for subroutines, provided it is introduced via a feature pragma and designed to coexist with existing tools like Exporter and symbol-table introspection.

Confirmed facts

• All globally declared subroutines are entered into a package's symbol table; there is no built-in privacy mechanism.
• Developers currently rely on naming conventions (leading underscore) or lexical closures, which are only advisory.
• Perl already supports subroutine attributes such as :lvalue and :method that can affect compile-time behavior.
• The Exporter module builds @EXPORT and @EXPORT_OK lists from the symbol table and can be taught to ignore subroutines bearing a specific attribute.
• A feature pragma (e.g., use feature 'private_sub';) can enable new syntax on newer Perls while older versions simply ignore the attribute, preserving backward compatibility.

Likely explanation of impact

Introducing an attribute like :private would cause the compiler to:

1. Forbid calls to the subroutine via its fully qualified name (Package::helper) when the feature is enabled, producing a compile-time error.
2. Optionally omit the subroutine's glob from the symbol table, making it invisible to tools that enumerate globs unless they explicitly look for private entries.
3. Be skipped by Exporter when constructing export lists, so private subs would not accidentally become part of a module's public API.

This would enforce encapsulation, reducing the chance that a helper routine is mistakenly used as part of the public interface.

Migration steps for CPAN authors

1. Add use feature 'private_sub'; near the top of the module (requires Perl 5.38+ or whatever version implements the feature).
2. Mark helper subroutines with :private, e.g., sub helper :private { ... }.
3. Ensure any tests or documentation that previously called the helper via its fully qualified name are updated to call it only within the module.
4. Run the module's test suite against both the new Perl (with the feature) and an older Perl (without the feature) to confirm that the attribute is ignored and the subroutine remains callable as before on older versions.

Open diagnostic detail

The recommendation hinges on how many CPAN modules currently depend on symbol-table introspection to access non-exported subroutines (e.g., for plugins, serialization, or debugging). If that proportion is large, hiding subs from the symbol table could break existing workflows and might require a more gradual approach. Please provide the estimated percentage of CPAN modules that rely on such introspection so the trade-off can be weighed accurately.