Re: [PATCH v2 01/10] Define a structure for object IDs.

[Michael Haggerty <mhagger@alum.mit.edu>]

On 03/12/2015 01:26 AM, Junio C Hamano wrote:
> "brian m. carlson" <sandals@crustytoothpaste.net> writes:
> 
>> Michael Haggerty recommended that I call the structure element sha1
>> instead of oid in case we want to turn this into a union if we decide to
>> go the additional hash route.
> 
> I'd advise against it.
> 
> As I wrote in $gmane/265337 in response to Michael:
> 
>     > 4. We continue to support working with SHA-1s declared to be (unsigned
>     > char *) in some performance-critical code, even as we migrate most other
>     > code to using SHA-1s embedded within a (struct object_id). This will
>     > cost some duplication of code. To accept this approach, we would need an
>     > idea of *how much* code duplication would be needed. E.g., how many
>     > functions will need both (unsigned char *) versions and (struct
>     > object_id *) versions?
> 
>     ...
> 
>     I do not know what kind of code duplication you are worried about,
>     though.  If a callee needs "unsigned char *", the caller that has a
>     "struct object_id *o" should pass o->hash to the callee.
> 
> And that would break the abstraction effort if you start calling the
> field with a name that is specific to the underlying hash function.
> The caller has to change o->sha1 to o->sha256 instead of keeping
> that as o->oid and letting the callee handle the implementation
> details when calling
> 
>         if (!hashcmp(o1->oid, o2->oid))
>                 ; /* they are the same */
>         else
>                 ; /* they are different */
> [...]

Hmm, I guess you imagine that we might sometimes pack SHA-1s, sometimes
SHA-256s (or whatever) in the "oid" field, which would be dimensioned
large enough for either one (with, say, SHA-1s padded with zeros).

I was imagining that this would evolve into a union (or maybe struct) of
different hash types, like

    struct object_id {
        unsigned char hash_type;
        union {
            unsigned char sha1[GIT_SHA1_RAWSZ];
            unsigned char sha256[GIT_SHA256_RAWSZ];
        } hash;
    };

BTW in either case, any hopes of mapping object_id objects directly on
top of buffer memory would disappear.

Michael

-- 
Michael Haggerty
mhagger@alum.mit.edu