Re: [QUESTION] how to diff one blob with nothing

From: ZheNing Hu <adlternative@gmail.com>
Date: Jul 26, 2023, 18:00 UTC
Message-ID: <CAOLTT8R84Zrtpd=j6o2V=Y_uD54XAS5EA7NWHsdfL+XCkD5cqA@mail.gmail.com>
In-Reply-To: <xmqqo7k7c1yw.fsf@gitster.g>

Junio C Hamano <gitster@pobox.com> 于2023年7月20日周四 00:15写道：

Show 31 quoted lines

>
> ZheNing Hu <adlternative@gmail.com> writes:
>
> > I want to diff two blobs right now, and one of them
> > may be empty, so I tried using
> > 0000000000000000000000000000000000000000 or
> > e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 (empty blobID)
> > to test its effect, and the result I found was:
> >
> > git diff 00750edc07d6415dcc07ae0351e9397b0222b7ba
> > 0000000000000000000000000000000000000000
> > fatal: bad object 0000000000000000000000000000000000000000
>
> As the object name for an empty blob is not all-0, this is expected.
>
> > git diff 00750edc07d6415dcc07ae0351e9397b0222b7ba
> > e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
> > fatal: bad object e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
> >
> > Since the "empty object" has not been created, the
> > git diff operation fails.
>
> If you haven't created one, of course it would fail.  It should help
> to do
>
>     $ git hash-object -w --stdin </dev/null
>
> before running
>
>     $ git diff 00750edc e69de29bb
>

This is a viable solution, but it's a bit ugly since a read-only "diff" requires ”write“ an empty blob.

> Long time ago, with 346245a1 (hard-code the empty tree object,
> 2008-02-13) we taught git what an empty-tree looks like, but it
> seems that we never did the same for an empty blob, perhaps?
>

It would be great to have that capability.

> Interesting.  I am not sure if it is a good idea to teach empty_blob
> to find_cached_object() and leaning negative but I haven't thought
> things through on this.
>

I currently don't have time to delve into implementation details, so I went with the approach that requires writing an empty blob, but I'll take a closer look at it later.

>
>

Thanks, ZheNing Hu