On Wed, Jun 12, 2013 at 9:02 AM, Andrew Tipton <andrew@kiwidrew.com> wrote:
> Hi all,
>
> I recently wanted to declare a CHECK constraint to ensure that a JSON value
> was an object with keys that were scalars (not nested objects or arrays).
> This proved to be more difficult than I had expected. In the end, I had to
> write a json_typeof() function in pl/pgsql. It's a simple function but uses
> a rather horrifying technique: it traps the exceptions raised when trying
> to apply json_array_length() to a non-array and json_each() to a non-object!
> I couldn't figure out a robust technique to further determine the type of a
> scalar value (int, float, text, or boolean) so this function simply returns
> one of 'array', 'object', or 'scalar'.
>
> CREATE FUNCTION json_typeof(json_value JSON)
> RETURNS TEXT AS $$
> BEGIN
> BEGIN
> PERFORM json_array_length(json_value);
> RETURN 'array';
> EXCEPTION WHEN invalid_parameter_value THEN NULL;
> END;
> BEGIN
> PERFORM json_each(json_value);
> RETURN 'object';
> EXCEPTION WHEN invalid_parameter_value THEN NULL;
> END;
> RETURN 'scalar';
> END;
> $$ LANGUAGE plpgsql STRICT IMMUTABLE;
>
> With this function, it's fairly simple to define a domain which only permits
> an array or an object as the top-level element:
>
> CREATE DOMAIN json_document CHECK (json_typeof(VALUE) IN ('array',
> 'object'));
>
> I feel that the json_typeof function is worthy of being a built-in function.
> Thoughts?
no argument against json_typeof, but why can you just peek the first
non-whitespace character? json is famously easy to parse (see here:
http://www.json.org/)
create or replace function json_typeof(_json json) returns text as
$$
select case substring(ltrim($1::text), 1, 1)
when '[' then 'array'
when '{' then 'object'
end;
$$ language sql immutable;
you could expand this mechanic fairly easy to cover all json types.
note exception handlers are very heavy for this type of operation.
merlin
