Re: [PATCH v2 3/3] rev-parse: match ^{<type>} case-insensitively

From: Junio C Hamano <gitster@pobox.com>
Date: Mar 27, 2017, 00:36 UTC
Message-ID: <xmqq60iv4ku2.fsf@gitster.mtv.corp.google.com>
In-Reply-To: <20170326121654.22035-4-avarab@gmail.com>

Ævar Arnfjörð Bjarmason  <avarab@gmail.com> writes:

Show 14 quoted lines

> Change the revision parsing logic to match ^{commit}, ^{tree}, ^{blob}
> etc. case-insensitively.
>
> Before this change supplying anything except the lower-case forms
> emits an "unknown revision or path not in the working tree"
> error. This change makes upper-case & mixed-case versions equivalent
> to the lower-case versions.
>
> The rationale for this change is the same as for making @{upstream}
> and related suffixes case-insensitive in "rev-parse: match
> @{upstream}, @{u} and @{push} case-insensitively", but unlike those
> suffixes this change introduces the potential confusion of accepting
> TREE or BLOB here, but not as an argument to e.g. "cat-file -t <type>"
> or "hash-object -t <type>".

It's not "potential confusion". This closes the door for us to introduce "TREE" as a separate object type in the future.

If we agree to make a declaration that all typenames are officially spelled in lowercase [*1*] and at the UI level we accept typenames spelled in any case, then we can adopt this change (and then we need to update "cat-file -t" etc. to match it).

I do not at all mind to see if the list concensus is to make such a declaration and permanently close the door for typenames that are not lowercase, and after seeing such a concensus I'd gladly appreciate this patch, but I do not want to see a change like this that decides the future of the system, pretending as an innocuous change, sneaked in without making sure that everybody is aware of its implications.

[Footnote]

*1* "officially spelled in lowercase" is necessary to avoid
    confusion, because we are not making typenames case insensitive,
    allowing an object whose raw-bytes in its loose object
    representation before deflating begins with "COMMIT" and take it
    as an object of <commit> type.  Instead, such a declaration will
    make such an object an invalid one (as opposed to "object of an
    unknown type", as it currently is).