Re: [HACKERS] SQL/JSON in PostgreSQL

2018-01-02 3:04 GMT+01:00 Nikita Glukhov <n(dot)gluhov(at)postgrespro(dot)ru>:

> On 29.11.2017 05:24, Michael Paquier wrote:
>
> On Wed, Nov 15, 2017 at 10:17 AM, Nikita Glukhov
>> <n(dot)gluhov(at)postgrespro(dot)ru> wrote:
>>
>>> Attached the new version of the patches where displaying of SQL/JSON
>>> constructor nodes was fixed. I decided not to invent new nodes but to
>>> extend
>>> existing FuncExpr, Aggref, WindowFunc nodes with new formatting fields
>>> that
>>> give us ability to override default displaying in ruleutils.c. Also new
>>> invisible CoercionForm was added for hiding casts in which FORMAT and
>>> RETUNING
>>> clauses are transformed.
>>>
>> Okay, I can see that the patch is still in the same situation two
>> weeks after I looked at it first, so I am marking it as returned with
>> feedback. This needs to be broken down, and get documentation. At this
>> point getting a review out of this patch is not something I'd
>> recommend until it is put in a shape that makes it easier. Please also
>> help in reviewing other's patches, yours here is very large.
>>
>
> Attached new version of patches:
>
> 1. Added some comments for the jsonpath code and some documentation for
> jsonpath operators and opclasses. Sorry that complete documentation for
> jsonpath itself is still missing.
>
> 2. Added support for custom user-defined item methods and functions in
> jsonpath.
> This feature allows us to move our non-standard methods (map, reduce,
> fold, min,
> max) to the extension contrib/jsnopathx. Now user can implement all
> standard
> JavaScript array methods by himself.
>
> 3. Added variable jsonpath specifications in SQL/JSON functions.
>
> 4. Added subtransactions inside PG_TRY/PG_CATCH block that is used for ON
> ERROR
> clause support in JSON_EXISTS, JSON_VALUE, JSON_QUERY (see ExecEvalJson()
> in
> src/backend/executor/execExpeInterp.c).
>
>
> The last addition is the most questionable. By using standard ON ERROR
> сlause
> we can specify default value for expression when an error occurs during
> casting
> JSON data to the target SQL type. Cast expressions can contain arbitrary
> user
> functions, so they should be executed inside own subtransaction like it is
> done
> in plpgsql (see exec_stmt_block()). Subtransaction start obviously
> introduces
> significant performance overhead (except the case when ERROR ON ERROR is
> used and error handling is unnecessary):
>
> =# EXPLAIN (ANALYZE)
> SELECT JSON_VALUE(jsonb '1', '$' RETURNING int ERROR ON ERROR)
> FROM generate_series(1, 1000000);
> ...
> Execution time: 395.238 ms
>
> =# EXPLAIN (ANALYZE)
> SELECT JSON_VALUE(jsonb '1', '$' RETURNING int)
> FROM generate_series(1, 1000000);
> ...
> Execution time: 914.850 ms
>

Subtransactions are pretty expensive - so using subtransaction for every
output row doesn't look like good solution.

In this case the exception is created in your code, so it should not be
PostgreSQL exception necessary.

Without ERROR ON ERROR clause it ignore errors by default?

>
> To partially eliminate this overhead, I'm trying to examine cast-expression
> volatility:
> * mutable -- subtransaction is started
> * stable -- subtransaction is not started, only new ResourceOwner is
> created
> * immutable -- ResourceOwner is not even created
> But don't know if it is correct to rely on volatility here. And also I
> doubt
> that we can start multiple subtransactions (for each SQL/JSON function
> evaluation) within a single SELECT statement.
>
>
> --
> Nikita Glukhov
> Postgres Professional: http://www.postgrespro.com
> The Russian Postgres Company
>