[
https://issues.apache.org/jira/browse/FLINK-23933?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17419655#comment-17419655
]
Francesco Guardiani commented on FLINK-23933:
---------------------------------------------
> Can't this issue also just be solved by making sure that the JSON format can
> cast to the expected schema?
This should already work, e.g.:
{code:java}
my ROW(nested ROW(data ROW<(structure TIMESTAMP(3))))
{code}
But this is very cumbersome to work with, and chances are that the user will
end up creating either a VIEW or computed columns to translate this
multidimensional data schema into a plain representation made of scalar
columns.
I think your point about "implicit" record is interesting. Without discussing
about new syntax, let's pick this example:
{code:java}
CREATE TABLE input(
a AS my.nested.data.structure
)
WITH (
'format' = 'json',
'json.timestamp-format.standard' = 'ISO-8601'
);
{code}
>From a very high point of view, the definition of a implicitly defines that
>the record will contain the nested objects "my.nested.data" and then inside
>that last object, it will have a field called "structure". What our syntax
>misses here is a way to tell that the "structure" attribute is of type
>TIMESTAMP(3). Implementation wise, walking through the data structure should
>already be possible, as stated in the first part of the comment, so what it
>would be nice to have is something like this implicit definition to reduce
>friction when working with nested data structures.
> flink-json formats support a option to parse nested json as a String
> --------------------------------------------------------------------
>
> Key: FLINK-23933
> URL: https://issues.apache.org/jira/browse/FLINK-23933
> Project: Flink
> Issue Type: Improvement
> Components: Formats (JSON, Avro, Parquet, ORC, SequenceFile)
> Affects Versions: 1.13.2
> Reporter: Shengnan YU
> Priority: Major
>
> When use Flink SQL to read message with nested json, eg. \{"a":{"b":111,
> "c":"xxx"}, "d":12345}，The nested field must be specified ROW or MAP in the
> table schema.
> {code:sql}
> CREATE TABLE input(
> a ROW(`b`: BIGINT, `c` VARCHAR),
> d BIGINT
> )
> WITH (
> 'format' = 'json',
> // ...
> );
> {code}
> I'd like to create a option to allow we read the json-object type value as a
> String.
> {code:sql}
> CREATE TABLE input (
> a VARCHAR,
> d VARCHAR
> )
> WITH (
> 'format' = 'json',
> 'json.forceStringFields' = 'a,d'
> // ...
> );
> {code}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
