[Feature Request] Resolution of namespaces

[Gravifer]

The current implementation supports document structure based on MMA cell structures (like (* ::Title:: *)). To my knowledge the official version is a pure front-end feature, and it is of much convenience to have it handled for code editors.
However, the kernel features - namespaces (contexts and packages) and scoping constructs like Module Block - are not resolved. Therefore, developers will have to go through the scripts manually to tell which context a symbol is in.

As most authors use namespaces concurrently along with cell structures, the current features are actually sufficient, and this issue may not be a necessity; if this is the case, this thread could be used to discuss the best practice for concurrent use of namespaces and cell structures.

[kenkangxgwe]

Cell structures are static while namespace and scoping functions are not. Begin["foo"<>ToString[Unique[]]] will give you a dynamic result. Resolving those will end up giving you wrong results. Thus, only if the file is evaluated once, will we know the actual context that a symbol is in. I am open to any effective, though not universally applicable, method to handle this.

[Gravifer]

Since the traditional dynamic namespace is difficult to resolve, would it be easier to resolve the static new style package derivatives? Does CodeParser handle this already?

[kenkangxgwe]

The new-style package is somehow not officially documented and CodeParser recognizes it but doesn't support the full features. I doubt I can run the following code directly from Wolfram Kernel:

Package["MyPackage`"]
PackageImport["CodeParser`"]

mySym := 1

Can you try it out? Thanks. I wish there will be more examples/resources of this package usage, if you'd like to provide them. Thanks

[Gravifer]

This feature has been added since 10.x or so, and seems to be still in common use even though remaining undocumented.
I found some related resources in the following links:

• tutorial by b3m2a1
• Leonid's description of the new package format
• Szabolcs' FAQ-style guide-line

I doubt I can run the following code directly from Wolfram Kernel: (code)

As mentioned above and explained in Leonid's answer, Package["MyPackage`"] and PackageImport["CodeParser`"] are only special static primitives, and are not really dynamically interpreted commands;. basically they are like PRAGMA macros in C/C++.

I see new-style packaging as a modernist attempt, but is not complete for designing nested packages and contexts. I won't fully switch to this practice, but I do think since the primitives are static, they can be easier to handle for the language server.

[kenkangxgwe]

Thanks for the information. I just tested the new-style and it seems working correctly (on Linux. No luck on my Windows' SimplifiedChinese version). Still gonna experiment with it for a while. I am starting to think about how we can leverage the namespace info now. Gonna leave this issue open for longer time.