Re: JSON Function Bike Shedding

On Fri, Feb 15, 2013 at 11:25 AM, Robert Haas <robertmhaas@gmail.com> wrote:
>> Note that I have given json_get() and json_get_path() the same names, as it seems to me that the former is the same
asthe latter, with only one parameter. Same for json_get_as_text() and json_get_path_as_text().
 
>
> I realize I'm in the minority here, but -1 from me on all of this.
> Should we also rename xml_is_well_formed() to just is_well_formed()?
> string_agg() to agg()?  Eventually we will have more data types, and
> some of them will have functions that could also be called rows() or
> get_values(), but it's unlikely that they'll have exactly the same
> behavior, which will start to make things confusing.

It's a little late, but I'd like to rebut this point:
> string_agg() to agg()?

This not germane to the discussion. string_agg means you are
aggregating *to* a string, not from one, which is a completely
different thing.  This also applies to to_char, to_date, etc.  If you
wanted to do just 'agg()', you'd have to supply output type somehow --
the only way to do that in postgres is to use hstore null::foo trick
(which is not an improvement obviously).

>  xml_is_well_formed() to just is_well_formed()?

Again, this is not the same thing. It does not work on the xml type,
but text, so you'd have to supply a hint to specific behaviors if you
wanted to abstract type out of the function.  Because the returned
type is unambiguously boolean though, you can get away with:

validate(format text, data text);

select validate('json', <json string>);
select validate('xml', <xml string>);
etc.

if you wanted to.  And yes, I absolutely think this is superior to
cluttering the public namespace with xml specific verbage, and could
be extended to other formats.  Look at the other way: we currently
have encode(format text, stuff bytea).  Would we be better off with
hex_encode(bytea), escape_encode(bytea)... .etc?

The argument for removing json_ prefix is that when function behaviors
are unambiguously controlled by the arguments, decorating the function
name to match the input argument is unnecessary verbosity.

merlin