Mark Evenson evenson@panix.com writes:

On 11/7/10 5:02 AM, Pascal J. Bourguignon wrote:

...

CLHS says:

 19.2.2.1.2.2 Common Case in Pathname Components

 For the functions in Figure 19-2, a value of :common for the :case
 argument that these functions should receive and yield strings in
 component values according to the following conventions:

 * All uppercase means to use a file system's customary case.
 * All lowercase means to use the opposite of the customary case.
 * Mixed case represents itself.

 Note that these conventions have been chosen in such a way that
 translation from :local to :common and back to :local is
 information-preserving.

What an odd corner of the CLHS, littered with the bones of extinct filesystems! This behavior that seems like it would produce unpleasantries much more than it would help anyone, violating the principle of least surprise. But we claim that ABCL will be ANSI first and foremost, so I guess we have to pay attention here…

I should mention that I use routinely logical pathnames, and they work very well, on all implementations that conformantly implement them.

That means that all my logical pathnames are always upcase letters, and no special symbols (apart from an occasional dash).

I understand that using implementation dependant syntax (special characters, lower-case, mixed-case) in logical pathnames can lead to surprises, but I avoid them.

Now, it's not a question of extinct file systems. You still have to deal with at least two different major syntaxes: POSIX and MS-Windows/MS-DOS.

I just cannot use in programs that will have to run on both Linux or MacOSX and MS-Windows physical pathname literals, because they don't use the same syntax.

On the other hand, #P"HOME:SRC;LISP;EXAMPLE.LISP" can be translated to either #P"C:\Users Files\Pascal\src\lisp\example.lisp" or #P"/home/pjb/src/lisp/example.lisp" portably on these targets.

The point for an implementation is to know what it's targets are, and to implement proper behavior when translating and merging logical pathnames. You must be careful when merging logical pathnames with physical pathnames, notably when it's done in make-pathname: the :case parameter must have effect only on the new strings, not on the :defaults parameter.

I suppose there is no definition of "customary case", huh?

Indeed.

Pascal suggests a notion of filesystem statistics ("99.999% of files are lowercase") but for a developer dealing with lots of Java files this isn't very releavant.

Does Java provide its own file system, or does it map to the host file system?

Notice that logical pathnames don't have to be able to represent all the physical pathnames, without an explicit translation.

If you have a file system that is case sensitive, then you may write:

(setf (logical-pathname-translations "APP") (list (list #P"APP:EXAMPLE;FILE.LISP" #P"/MyApp/Example/File.lisp") (list #P"APP:EXAMPLE;ANOTHER-FILE.LISP" #P"/MyApp/Example/AnotherFile.lisp") (list #P"APP:EXAMPLE;*.DATA" #P"/MyApp/Example/*.appdata") (list #P"APP:EXAMPLE;SOME-MODULE.*" #P"/MyApp/Example/SomeBigModule.*") (list #P"APP:APP.CONFIGURATION" #P"/MyApp/.config") (list #P"APP:CUSTOMARY;**;*" #P"/MyApp;**;*") (list #P"APP:CUSTOMARY;**;*.*" #P"/MyApp;**;*.*")))

to be able to access to files with strange physical names.

The question indeed is to decide what is the customary case. You can choose uppercase or you can choose lowercase. This matters only for file systems that are case sensitive, but even file systems that are not case sensitive, such as MS-DOS, have a prefered way to display the names.

Lacking a more precise definition, I would then advocate that we declare lowercase to be the customary case across UNIX, OSX, and Windows for ABCL, patching our behavior accordingly.

That would be good.

I thought briefly about advocating Windows to have "uppercase" as its customary case, but that is only really true in a DOS world which "modern" Windows systems really aren't any more.

Yes.

And although we theoretically run on JVMs on say VMS, if we run into such needs, we might devote a user-accessible special variable to control the implementation.

This is also a good option.

Pending further comments, I'll open a ticket.

Thank you.