Re: Variable substitution in psql backtick expansion

On Sun, Apr 2, 2017 at 12:29 PM, Fabien COELHO <coelho(at)cri(dot)ensmp(dot)fr> wrote:

>
> Hello Corey,
>
> I'm anxious to help with these patches, but they seem a bit of a moving
>> target. Happy to jump in and review as soon as we've settled on what
>> should
>> be done.
>>
>
> The "v3" I sent basically adds both client & server version numbers in
> client-side variables, basically same code as suggested by Pavel for the
> server version, and some documentation.
>

patch applies via patch -p1

Works as advertised.
# \echo SERVER_VERSION_NAME
SERVER_VERSION_NAME
# \echo :SERVER_VERSION_NAME
10.0
# \echo :SERVER_VERSION_NUM
100000
# \echo :VERSION_NUM
100000

The new documentation is clear, and accurately reflects current name style.

Looking at #define STRINGIFY(), I got curious where else STRINGIFY was used:

$ git grep STRINGIFY
src/bin/psql/startup.c:#define STRINGIFY2(symbol) #symbol
src/bin/psql/startup.c:#define STRINGIFY(symbol) STRINGIFY2(symbol)
src/bin/psql/startup.c: SetVariable(pset.vars, "VERSION_NUM",
STRINGIFY(PG_VERSION_NUM));
src/tools/msvc/Solution.pm:s{PG_VERSION_STR "[^"]+"}{__STRINGIFY(x)
#x\n#define __STRINGIFY2(z) __STRINGIFY(z)\n#define PG_VERSION_STR
"PostgreSQL $self->{strver}$extraver, compiled by Visual C++ build "
__STRINGIFY2(_MSC_VER) ", $bits-bit"};

Without digging too deep, it seems like the redefinition wouldn't be
harmful, but it might make sense to not use the name STRINGIFY() if only to
avoid confusion with Solution.pm.

> The questions are:
>
> - which version should be provided (10.0 100000 ...)
>

A fixed length string without decimals seems best for the multitude of
tools that will want to manipulate that data.

> - how should they be named?
>
> In v3 there is VERSION{,_NAME,_NUM} for client and
> SERVER_VERSION_{NUM,NAME} or SVERSION_NUM suggested
> by Pavel for server.
>

SERVER_VERSION_* is good.
VERSION_* is ok. Would CLIENT_VERSION_* or PSQL_VERSION_* be better?

> - how desirable/useful is it to have this in 10?
>

Extensions and extension-ish packages will love the _NUM vars. The sooner
the better.

There's a lesser need for the _NAME vars.