directory naming conventions

[mkobetic]

I updated STIG to align with what we agreed on at the last ESUG (not quite everything yet, but getting there). I'm trying to sync up with what Jan did for ST/X. I see a major issue though in terms of how we name the package and class directories. For Xtreams I'm generating following directory structure (sample):

Xtreams-Core.package/
Object.extension/
Incomplete.class/
...
Xtreams-Core-Tests.package/
...

In ST/X a similarly packaged structure is coming out as

stx_goodies_xtreams_core.package/
Object.extension/
Xtreams__Incomplete.class/
...
stx_goodies_xtreams_core_tests.package/
...

Both are obviously valid choices, but the ultimate result is that there's no way we can usefully manage these two structures as parallel branches in git. So I'm wondering if we should agree on some additional guidelines for cross-dialect compatibility? Thoughts?

[janvrany]

On 11/04/13 03:27, Martin Kobetic wrote:

I updated STIG to align with what we agreed on at the last ESUG (not
quite everything yet, but getting there). I'm trying to sync up with
what Jan did for ST/X.

Nice!

I see a major issue though in terms of how we

name the package and class directories. For Xtreams I'm generating
following directory structure (sample):

...

Both are obviously valid choices, but the ultimate result is that
there's no way we can usefully manage these two structures as parallel
branches in git. So I'm wondering if we should agree on some additional
guidelines for cross-dialect compatibility? Thoughts?

Yes, that's a problem. I think we have to try to use it before writing
down a guideline. Xtreams could be a nice usecase.

Regarding this particular case - since in St/X package is first-class
object (first-class class to be precise), I can fix St/X Cypress
implementation and provide per-package hooks to supply directory/file
names for packages, classes and methods. This way, St/X would be able to
create same structure as STIG/FileTree (naming-wise). If you think it
would help, I can do it.

Regarding class directory names - I do think we should prepend namespace
if any. Just think about a package with two classes with
same name in different namespaces. Weird, indeed, but possible.
I do not insist on double underscore as delimiter, though. This is
merely a convention in St/X.

Namespace handling is another issue - each implementation handle them
differently, if at all. I don't think there will be any consensus in
foreseeable future. No idea how to easily solve that.

Jan