> > The only thing I would change, is I would drop it to three functions by
> > getting rid of apr_temp_get_unique. That function can be accounted
> > for by
> > passing NULL into apr_temp_get_filename for the directory. Also, the
> > apr_temp_get_unique really doesn't have anything to do with temporary
> > files. The PHP uniqid function just generates a unique string, APR
> > doesn't need to implement that feature, because APR-util already has
> > UUID
> > support, which is designed to generate unique strings.
>
> +1
>
> > The way I see thing working is that the user would do the following to
> > get
> > a filename in the system's temp directory:
> >
> > apr_temp_get_filename(apr_temp_get_directory(...), ".foo", pool);
>
> I think this should return a filehande, not a filename. Or make it
> two functions.
We already have the function that returns the filehandle. We are
specifically talking about adding the function that returns the filename.
> > To get a filename without a directory, they would do:
> >
> > apr_temp_get_filename(NULL, ".foo", pool);
>
> OK, but what's the point of that?
Sorry, this section of my message should have been deleted. My first
thought was that this is how you would get the apr_file_unique() function,
but as I said above, that support shouldn't be added to the library
anyway.
> > The ability to use a compiled in default path (PATH_TMP) is harder to
> > deal
> > with, because most platforms don't have that concept. I don't believe
> > that it is a good idea to add flags that only really make sense on some
> > platforms. We are better off IMNSHO, just ignoring that case for now,
> > and
> > force the app developer to handle it themselves by passing that macro
> > to
> > apr_temp_get_filename.
>
> Um, yuck. The whole point of apr_temp_get_directory() in David's API
> was to return that directory.
>
> You're talking about an equivalent to apr_temp_get_filename() which
> returns a directory instead of a file. That's two different functions,
> which I suggested previously.
Huh????? The API that we are discussing _has_ two functions for this.
One to get the filename and one to get the directory name. What we are
talking about here is how you deal with a default compiled-in path versus
the one from the environment variable. The solution that has been
proposed is a flag that is passed in. But that flag is only going to have
an affect on Unix systems. I would much rather just use logic like:
/* Add more if statemets for other platforms */
if (getenv("TEMP")) {
return getenv("TEMP")
}
if (getenv("TEMPDIR")) {
return getenv("TEMPDIR")
}
return PATH_TMP
This allows the system administrator to override the platform default.
What it removes is the ability to get the default value of PATH_TMP if the
administrator has overridden it.
Ryan
