Hi,
We have a use case to perform soft delete over some record keys where we
nullify non-key fields and ignore any update for this record later on. We
thought of using a hudi meta field: "_hoodie_is_soft_deleted" as hudi hard
delete (_hoodie_is_deleted) does to make it simple to identify if the
p
Hi,
Hudi is complaining because '_hoodie_is_soft_deleted' is present in the
parquet file's schema but is not present in your incoming schema.
>From my experience, I would say it is a standard practice to add an extra
field which acts as a marker for soft deletion and needs to be persisted
with ev
Thanks for the response Pratyaksh!
We add this column to the Spark dataframe before calling the hudi upsert
and delete. And this should work like an extra nullable column in the
schema but it's not behaving like that, so wondering if we remove any
column with the prefix *'_hoodie' * in Hudi code.
Hi Aakash,
For the field to behave as a nullable extra field, you need to add default
value as null to the schema and make "null" as the first type in your union
schema for `_hoodie_is_soft_deleted`.Hope that helps.
On Fri, Jul 15, 2022 at 8:01 PM aakash aakash
wrote:
> Thanks for the response
yes, from the pasted schema, there is no default set for the newly added
column.
{
"name" : "*_hoodie_is_soft_deleted*",
"type" : [ "string", "null" ]
} ]
If you can fix that and give it a try, it should work.
On Sat, 16 Jul 2022 at 03:13, Pratyaksh Sharma
wrote:
> Hi Aakash,
>
> Fo