Re: [PATCH v2 16/21] Make sha1_array_append take a struct object_id *

From: Junio C Hamano <gitster@pobox.com>
Date: Mar 29, 2017, 15:14 UTC
Message-ID: <xmqqvaqsazec.fsf@gitster.mtv.corp.google.com>
In-Reply-To: <20170329000648.vb2hdqfizjqjyryr@genre.crustytoothpaste.net>

"brian m. carlson" <sandals@crustytoothpaste.net> writes:

Show 18 quoted lines

> On Tue, Mar 28, 2017 at 10:27:41AM -0700, Junio C Hamano wrote:
>> ... 
>> After all, the original written by a human said E2[E3].hash (or
>> array->sha1[i]) because to the human's mind, E2 is a series of
>> things that can be indexed with an int E3, and even though 
>> 
>>     *(E2 + E3)
>>     E2[E3]
>>     E3[E2]
>> 
>> all mean the same thing, the human decided that E2[E3] is the most
>> natural way to express this particular reference to an item in the
>> array.  &E2[E3] would keep that intention by the original author
>> better than E2 + E3.
>
> I'm happy to make that change.  I'm an experienced C programmer, so a
> pointer addition seems very readable and natural to me, but if you think
> it's better or more readable as &E2[E3], I can certainly reroll.

The obvious three possible variants I listed above are equivalent to the language lawyers and the compiler ;-).

This change is about dropping the need for ".hash", and I think a faithful, boring and mechanical conversion that tries to preserve the intent of the original author would be more appropriate. It is entirely possible that some places where the original said E2[E3] were easier to understand if it were *(E2 + E3), thus we may want to further rewrite such a place to (E2 + E3) instead of &E2[E3] after the mechanical conversion.

Thanks.