json path expressions in query are not working. select payload.push_id from githubEvents limit 10 throws unknown column name exception. But using json functions works. @User and @User - this is using json batch quickstart setup.

Thanks, will take a look.

I dont think that is tested well and can result in unexpected behavior when there is nested structure. Please use json functions if possible

ok, thanks

There are some examples in our test file that could help you decide: https://github.com/apache/pinot/blob/master/pinot-core/src/test/java/org/apache/pinot/queries/JsonPathQueriesTest.java

got it, then we can loose the identifier check for json type expression with dot

brackets are also valid

[

]

currently along with dot

.

In future, we hope to expand it to take a full json path expression

SELECT tablename.columnname.jsonpath FROM myTable

provided that

tablename.columnname

is. column of type

JSON

.

hmm, why

tablename.columnname

not just

columnname

?

From what I recall,

tablename.columnname

was already supported in queries before we expanded identifiers to JSON notation.

I feel those column information should be standardized before the check?

Yes, after that

JsonStatementOptimizer

does a check to ensure that the the dot notation is applying to a JSON column. I can add a fix for this if you like? I think we would need to add a couple of integration tests as well to make sure this doesn't break again. The change should have gotten caught by a test case 🙂 otherwise its easy to skip this again.

We should add some json queries into the integration test. That should be able to catch such regressions

Thanks