Re: SQL/JSON: functions - Mailing list pgsql-hackers

On 4/8/22 08:02, Justin Pryzby wrote:
> On Thu, Mar 31, 2022 at 04:25:58PM -0400, Andrew Dunstan wrote:
>> No code chunks left, only a documentation patch which should land
> Documentation review for a6baa4bad.
>
>> Construct a JSON the provided strings:
> a JSON what ?
> *from* the provided strings ?
>
>> Construct a JSON from the provided values various types:
> should say "a JSON scalar" ?
> *of* various types ?
>
>> Construct a JSON object from the provided key/value pairs of various types:
> For comparison, that one looks ok.
>
> +      <function>JSON_EXISTS</function> function checks whether the provided
> +   <function>JSON_VALUE</function> function extracts a value from the provided
> +   <function>JSON_QUERY</function> function extracts an <acronym>SQL/JSON</acronym>
> +      <function>JSON_TABLE</function> function queries <acronym>JSON</acronym> data
> +     <function>JSON_TABLE</function> uses the
> +      <function>JSON_SERIALIZE</function> function transforms a SQL/JSON value
>
> I think all these should all begin with "THE >...< function ...", like the
> others do.
>
> +To use other types, you must create the <literal>CAST</literal> from <type>json</type> for this type.
> => create a cast from json to this type.
>
> +Values can be null, but not keys.
> I think it's clearer to say "..but keys cannot."
>
> +          For any scalar other than a number or a Boolean the text
>
> Boolean COMMA the text
>
> +     The path name must be unique and cannot coincide with column names.
> Maybe say "name must be unique and distinct from the column names."
>
> +      ... If you specify a <command>GROUP BY</command>
> +      or an <command>ORDER BY</command> clause, this function returns a separate JSON object
> +      for each table row.
>
> "for each table row" sounds inaccurate or imprecise.  The SELECT docs say this:
> | GROUP BY will condense into a single row all selected rows that share the same values for the grouped expressions
>
> BTW, the documentation references look a little like OIDs...
> Does someone already have an SNMP-based doc browser ?
> | For details, see Section 9.16.3.4.2.



Many thanks, useful.

I already had a couple of these items on my list but I ran out of time
before tiredness overcame me last night.

I'm planning on removing some of that stuff that generates the last
complaint if I can do it without too much violence.


cheers


andrew

--
Andrew Dunstan
EDB: https://www.enterprisedb.com