Re: Using defines for protocol characters

On Mon, Aug 07, 2023 at 02:24:58PM -0400, Tom Lane wrote:
> Robert Haas <robertmhaas(at)gmail(dot)com> writes:
>> IMHO, the correspondence between the names in the patch and the
>> traditional names in the documentation could be stronger. For example,
>> the documentation mentions EmptyQueryResponse and
>> NegotiateProtocolVersion, but in this patch those become
>> PQMSG_RESP_NEGOTIATE_PROTOCOL and PQMSG_RESP_EMPTY_QUERY. I think we
>> could consider directly using the names from the documentation, right
>> down to capitalization, perhaps with a prefix, but I'm also totally
>> fine with this use of uppercase letters and underscores. But why not
>> do a strict translation, like EmptyQueryResponse ->
>> PQMSG_EMPTY_QUERY_RESPONSE, NegotiateProtocolVersion ->
>> PQMSG_NEGOTIATE_PROTOCOL_VERSION? To me at least, the current patch is
>> inventing new and slightly different names for things that already
>> have names...
>
> +1. For ease of greppability, maybe even PQMSG_EmptyQueryResponse
> and so on? Then one grep would find both uses of the constants and
> code/docs references. Not sure if the prefix should be all-caps or
> not if we go this way.

+1 for Tom's suggestion. I have no strong opinion about the prefix, but I
do think easing greppability is worthwhile.

As mentioned earlier [0], I think we should also consider putting the
definitions in pqcomm.h.

[0] https://postgr.es/m/20230803185356.GA1144430%40nathanxps13

--
Nathan Bossart
Amazon Web Services: https://aws.amazon.com