Daniel Becker created IMPALA-12926:
--------------------------------------
Summary: Remove AuxColumnType and add 'is_binary' field to
ColumnType
Key: IMPALA-12926
URL: https://issues.apache.org/jira/browse/IMPALA-12926
Project: IMPALA
Issue Type: Improvement
Components: Backend
Reporter: Daniel Becker
Assignee: Daniel Becker
Currently the STRING and BINARY types are not distinguished in most of the
backend. In contrast to the frontend, PrimitiveType::TYPE_BINARY is not used
there at all, TYPE_STRING being used instead. This is to ensure that everything
that works for STRING also works for BINARY. So far only file readers and
writers have had to handle them differently, and they have access to
ColumnDescriptors which contain AuxColumnType fields that differentiate these
two types.
However, only top-level columns have ColumnDescriptors. Adding support or
BINARYs within complex types (see IMPALA-11491 and IMPALA-12651) necessitates
adding type information about STRING vs BINARY to embedded fields as well.
Using PrimitiveType::TYPE_BINARY would probably be the cleanest solution but it
would affect huge parts of the code as TYPE_BINARY would have to be added to
hundreds of switch statements and this would be error prone.
Instead, we should introduce a new field in ColumnType: 'is_binary', which is
true if the type is a BINARY and false otherwise. We keep using TYPE_STRING as
the PrimitiveType of the ColumnType for BINARYs. This way full type information
is present in ColumnType but code that does not differentiate between STRING
and BINARY will continue to work for BINARY.
With this change, AuxColumnType is no longer needed and should be removed.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
