On 02/10/2012 01:14 PM, Peter Eisentraut wrote:
The auto_explain module appears to create invalid JSON (in all versions
since 9.0). For example, with the settings
auto_explain.log_format = 'json'
auto_explain.log_min_duration = 0
the query
select * from pg_type;
produces this in the log:
LOG: duration: 529.808 ms plan:
[
"Query Text": "select * from pg_type;",
"Plan": {
"Node Type": "Seq Scan",
"Relation Name": "pg_type",
"Alias": "pg_type",
"Startup Cost": 0.00,
"Total Cost": 9.87,
"Plan Rows": 287,
"Plan Width": 611
}
]
Note that at the top level, it uses the array delimiters [ ] for what is
actually an object (key/value). Running this through a JSON parser will
fail.
By contrast, EXPLAIN (FORMAT JSON) on the command line produces:
QUERY PLAN
-----------------------------------
[
{
"Plan": {
"Node Type": "Seq Scan",
"Relation Name": "pg_type",
"Alias": "pg_type",
"Startup Cost": 0.00,
"Total Cost": 9.87,
"Plan Rows": 287,
"Plan Width": 611
}
}
]
(1 row)
Yeah, looks like this dates back to when we first got JSON output.
Auto-explain does this:
ExplainBeginOutput(&es);
ExplainQueryText(&es, queryDesc);
ExplainPrintPlan(&es, queryDesc);
ExplainEndOutput(&es);
But ExplainBeginOutput says:
case EXPLAIN_FORMAT_JSON:
/* top-level structure is an array of plans */
appendStringInfoChar(es->str, '[');
Now that's not true in the auto-explain case, which prints one query +
one plan.
Since this is an exposed API, I don't think we can just change it. We
probably need a new API that does the right thing for beginning and
ending auto_explain output. (ExplainBeginLabeledOutput?)
cheers
andrew
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
