Re: [AD] Clarifying and revising filesystem API

[Elias Pschernig <elias.pschernig@xxxxxxxxxx>]

Peter Wang wrote:
> Let me propose again a URI-like scheme, and strip the API down to
> stream operations only.
>
>
> ## API: al_register_stream_handler
>
>     bool al_register_stream_handler(const char *scheme,
> 	ALLEGRO_FILE_VTABLE *vtable);
>
> Register a handler for a particular scheme.  The scheme must be a
> non-empty string, not containing any colon characters, e.g. "file".
> vtable is a pointer to a table of function pointers.  The functions will
> be used to operate on the streams which were opened with the particular
> scheme.
>
> By default, the scheme "file" is registered and is handled using the C
> library stdio functions.

Could we have a function to set a per-thread default handler? The use cases I 
can think of would make things a lot simpler with that as I don't have to mess 
with prepending schemes to copies of my filename strings all the time. Something 
like:

al_set_default_stream_handler(ALLEGRO_FILE_VTABLE *vtable);

Passing NULL would reset to the builtin one.

> ## Writing to a memory block
>
>     /* You need to generate a unique token somehow.
>      * The addon could help you, or you could follow a convention
>      * like Java package names.
>      */
>     const char *token = "mem:b3457c95-db59-480d-91d0-9d0f61a6e7d1";

I find this really confusing... when doing it like this, what is the remaining 
advantage of using this over the _entry version?

--
Elias Pschernig <elias@xxxxxxxxxx>