Querying variant data in Snowflake - snowflake-cloud-data-platform

Question

Here is the data variant source table I am using in my example. I want to make a query to parse this data into a table in snowflake from a variant src.
{
"col1": bool,
"col2": null,
"col3": "datetime",
"col4": int,
"col5": "string",
"col6": "string",
"array": [
{
"x": bool,
"y": null,
"v": "datetime",
"z": int,
"w": "string",
"q": "string",
"obj": {
"a": "bool",
"b": "float"
},
"col7": "datetime"
}
]
}
-- Here what I tried
SELECT
src:col1::string as col1,
src:col2::string as col2,
src:col3::string as col3,
src:col4::string as col4,
src:col5::string as col5,
src:col6::string as col6,
s.value:x::string as S_x,
s.value:y::string as s_y,
s.value:v::string as s_v,
s.value:z::string as s_z,
s.value:w::string as s_w,
s.value:q::string as s_q,
s.value:obj.value:a::string as s_obj_a,
s.value:obj.value:b::string as s_obj_b,
src:col7::string as col7
FROM tblvariant
, table(flatten(src:s)) s
;
Everything is working except that these two columns (a, b) are null while they should contain their data.
Any suggestion?
Many thanks!

Answer 1

Your sample JSON does not match with your SQL. Where are "stages" and "metadata"? Anyway, the problem seems about extra "value" keyword.
create or replace table tblvariant ( src variant )
as select parse_json ('
{
"col1": "bool",
"col2": null,
"col3": "datetime",
"col4": "int",
"col5": "string",
"col6": "string",
"stages": [
{
"x": "bool",
"y": null,
"v": "datetime",
"z": "int",
"w": "string",
"q": "string",
"obj": {
"a": "bool",
"b": "float"
},
"col7": "datetime"
}
]
}' );
As you see, I modified your sample JSON and renamed "array" to "stages" (according to your SQL). This SQL retrieves values of a and b:
SELECT
src:col1::string as col1,
src:col2::string as col2,
src:col3::string as col3,
src:col4::string as col4,
src:col5::string as col5,
src:col6::string as col6,
s.value:x::string as S_x,
s.value:y::string as s_y,
s.value:v::string as s_v,
s.value:z::string as s_z,
s.value:w::string as s_w,
s.value:q::string as s_q,
s.value:obj.a::string as s_obj_a,
s.value:obj.b::string as s_obj_b,
src:col7::string as col7
FROM tblvariant
, table(flatten(src:stages)) s
-- , table(flatten(s.value:metadata)) m
;

Answer 2

s.value:obj .value:a ::string as s_obj_a,
s.value:obj .value:b ::string as s_obj_b,
Accessing an object's keys can be done with the use of the dot (.) notation. You do not need to use the GET_PATH (:) operator to gain access to those fields:
s.value:metadata.a::string as s_m_a,
s.value:metadata.b::string as s_m_b,
You also do not need to run a second FLATTEN over the metadata object within your stages array, unless you truly need one exclusive row per metadata key, assuming metadata is an object type and not a nested array. If you just want to extract the values out into the same level as each array row, just using the above should suffice.