[
https://issues.apache.org/jira/browse/FLINK-40580?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ramin Gharib updated FLINK-40580:
---------------------------------
Description:
Casting a VARIANT to CHAR/VARCHAR/STRING renders a stored scalar the way a
regular SQL cast of that kind would (a boolean becomes TRUE, a timestamp uses
the SQL format, a binary value is read as UTF-8). A VARIANT that holds an
object or an array has no scalar form, so today the cast fails:
{code:java}
Cannot cast a VARIANT OBJECT value to a character string. Use the JSON_STRING
function to obtain its JSON representation. {code}
Redirecting users to JSON_STRING for the container case is surprising: the
value already has a well-defined textual form, its JSON representation, which
both JSON_STRING and the result-print path already produce.
Proposal: make CAST(v AS STRING) render an object or array as its JSON
representation (via Variant#toJson()) instead of failing. Scalar rendering is
unchanged, so a top-level scalar string still casts unquoted while a string
nested inside a rendered container stays quoted, matching JSON_STRING and the
print path.
{code:java}
CAST(PARSE_JSON('["a","b"]') AS STRING) -- ["a","b"] (was: runtime error)
CAST(PARSE_JSON('\{"a":1}') AS STRING) -- \{"a":1} (was: runtime error)
CAST(PARSE_JSON('"foo"') AS STRING) -- foo (unchanged)
{code}
A bounded CHAR/VARCHAR target trims the rendered JSON like any other
over-length value. TRY_CAST returns the same JSON rather than NULL for a
container. A VARIANT storing a JSON null still casts to SQL NULL.
Scope: only OBJECT and ARRAY change here. Scalar kinds whose string rendering
is still unsupported (TIME and nanosecond timestamps, tracked in FLINK-40492)
keep failing.
was:
Casting a VARIANT to CHAR/VARCHAR/STRING renders a stored scalar the way a
regular SQL cast of that kind would (a boolean becomes TRUE, a timestamp uses
the SQL format, a binary value is read as UTF-8). A VARIANT that holds an
object or an array has no scalar form, so today the cast fails:
{code}Cannot cast a VARIANT OBJECT value to a character string. Use the
JSON_STRING function to obtain its JSON representation.\{code}
Redirecting users to JSON_STRING for the container case is surprising: the
value already has a well-defined textual form, its JSON representation, which
both JSON_STRING and the result-print path already produce.
Proposal: make CAST(v AS STRING) render an object or array as its JSON
representation (via Variant#toJson()) instead of failing. Scalar rendering is
unchanged, so a top-level scalar string still casts unquoted while a string
nested inside a rendered container stays quoted, matching JSON_STRING and the
print path.
{code:sql}
CAST(PARSE_JSON('["a","b"]') AS STRING) -- ["a","b"] (was: runtime error)
CAST(PARSE_JSON('\{"a":1}') AS STRING) -- \{"a":1} (was: runtime error)
CAST(PARSE_JSON('"foo"') AS STRING) -- foo (unchanged)
{code}
A bounded CHAR(n)/VARCHAR(n) target trims the rendered JSON like any other
over-length value. TRY_CAST returns the same JSON rather than NULL for a
container. A VARIANT storing a JSON null still casts to SQL NULL.
Scope: only OBJECT and ARRAY change here. Scalar kinds whose string rendering
is still unsupported (TIME and nanosecond timestamps, tracked in FLINK-40492)
keep failing.
> Cast from VARIANT to a character string should render an object or array as
> JSON
> --------------------------------------------------------------------------------
>
> Key: FLINK-40580
> URL: https://issues.apache.org/jira/browse/FLINK-40580
> Project: Flink
> Issue Type: Improvement
> Reporter: Ramin Gharib
> Assignee: Ramin Gharib
> Priority: Major
>
> Casting a VARIANT to CHAR/VARCHAR/STRING renders a stored scalar the way a
> regular SQL cast of that kind would (a boolean becomes TRUE, a timestamp uses
> the SQL format, a binary value is read as UTF-8). A VARIANT that holds an
> object or an array has no scalar form, so today the cast fails:
> {code:java}
> Cannot cast a VARIANT OBJECT value to a character string. Use the JSON_STRING
> function to obtain its JSON representation. {code}
> Redirecting users to JSON_STRING for the container case is surprising: the
> value already has a well-defined textual form, its JSON representation, which
> both JSON_STRING and the result-print path already produce.
>
> Proposal: make CAST(v AS STRING) render an object or array as its JSON
> representation (via Variant#toJson()) instead of failing. Scalar rendering is
> unchanged, so a top-level scalar string still casts unquoted while a string
> nested inside a rendered container stays quoted, matching JSON_STRING and the
> print path.
> {code:java}
> CAST(PARSE_JSON('["a","b"]') AS STRING) -- ["a","b"] (was: runtime error)
> CAST(PARSE_JSON('\{"a":1}') AS STRING) -- \{"a":1} (was: runtime error)
> CAST(PARSE_JSON('"foo"') AS STRING) -- foo (unchanged)
> {code}
> A bounded CHAR/VARCHAR target trims the rendered JSON like any other
> over-length value. TRY_CAST returns the same JSON rather than NULL for a
> container. A VARIANT storing a JSON null still casts to SQL NULL.
> Scope: only OBJECT and ARRAY change here. Scalar kinds whose string rendering
> is still unsupported (TIME and nanosecond timestamps, tracked in FLINK-40492)
> keep failing.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
