[asdf-devel] SOURCE-FILE-TYPE

[Faré]

On Sat, Mar 22, 2014 at 5:25 PM, Robert P. Goldman <rpgoldman at sift.info> wrote:
> I have a note from you requesting that I mark SOURCE-FILE-TYPE as
> deprecated in code and documentation.
>
> I'm not sure I fully understand.
>
> The code in component.lisp still uses this, and it's not obvious how to
> remove it (nor what it means that it can return :DIRECTORY).
>
> I *believe* that what you mean is that SOURCE-FILE-TYPE will stay, but
> that outsiders should no longer make methods on it: instead they should
> use the :FILE-TYPE slot on FILE-COMPONENT.  But maybe not; see my point
> 3 below.
>
Yes, that's what I meant: the functionality remains for backward
compatibility until all clients have been upgraded for months/years,
but no one should define specialized methods on it anymore in new
code, and the functionality will go away eventually.

> Some follow up:
>
> 1.  This comment: "
> ;; SOURCE-FILE-TYPE below is strictly for backward-compatibility with ASDF1.
> ;; We ought to be able to extract this from the component alone with
> COMPONENT-TYPE.
>
> That's out of date, right?  Should refer to extract this from
> FILE-COMPONENTs along with FILE-TYPE, correct?
>
Yes.

> 2.  Why is FILE-TYPE an ACCESSOR instead of a READER?  Should anyone be
> setting this, except as an INITARG?
>
Reader is probably what it should be indeed. I don't know why I made
it accessor.

> 3.  What is the return type of SOURCE-FILE-TYPE, and in particular why
> does it not return an error when applied to a PARENT-COMPONENT?  I think
> it's because it's not being used only as a FILE type, but is being used
> as a TYPE argument for ASDF's pathname functions, which have a more
> expansive notion of :type.  I suppose maybe this means that
> SOURCE-FILE-type is a misnomer, and that's one of the reasons it should die.
>
I suppose the return type is (or string null (eql :directory));
yes, it's a misnomer, a misdesign, and it should die.

> I'm working my way through the ASDF manual, end-to-end.  I will probably
> commit a marked up PDF to the repo pretty soon, so anyone who is
> interested can see where I'm going.
>
Oops, I just made a few small changes to the manual. I hope they don't conflict.

PS: I also removed old merged branches from the repository, and
created new ones.
I invite you to merge in the renamed-bundle-op branch in which I
rename bundle-operation as promised, and also the build-op branch that
implements the strings as operation designator as well as renamed the
build-system function to build (but I think the consensus is to
further rename it to make). The standard-syntax branch is not ready
and contains previous experiments for syntax control, that I hope we
can complete and merge after the current release.

—♯ƒ • François-René ÐVB Rideau •Reflection&Cybernethics• http://fare.tunes.org
The weak terrorist who predictably loses, causing death for those
he claims to defend is infinitely more evil than the strong one who wins.