github-actions[bot] commented on code in PR #65329:
URL: https://github.com/apache/doris/pull/65329#discussion_r3601023103
##########
fe/fe-core/src/main/java/org/apache/doris/nereids/trees/plans/commands/info/ColumnDefinition.java:
##########
@@ -207,19 +217,27 @@ public String getComment(boolean escapeQuota) {
* toSql
*/
public String toSql() {
+ return toSql("`" + name + "`");
+ }
+
+ /**
+ * Convert this column definition to SQL with a caller-provided column
name.
+ */
+ public String toSql(String columnNameSql) {
Review Comment:
[P2] Round-trip nested literal metadata with SQL-mode-aware quoting
`AddColumnOp.toSql()` and `ModifyColumnOp.toSql()` now route path-aware
definitions through this overload, but it still renders decoded DEFAULT/COMMENT
values with `SqlUtils.escapeQuota`. In ordinary mode a value such as `C:\\tmp`
is emitted with a single backslash and `parseStringLiteral` turns `\\t` into a
tab; under `NO_BACKSLASH_ESCAPES`, an embedded `"` is emitted as `\\"` and
reparses with an extra backslash. Replaying the rendered ALTER can therefore
commit a different Iceberg initial default or doc. Please use the
SQL-mode-aware inverse of `parseStringLiteral` for both literal defaults and
comments, and add ADD/MODIFY round-trip cases in both modes.
##########
be/src/format_v2/table/iceberg_reader.cpp:
##########
@@ -91,31 +91,14 @@ static Status build_missing_equality_delete_key_expr(const
format::ColumnDefinit
return Status::OK();
}
- Field initial_default;
- if (table_field.initial_default_value_is_base64 ||
- table_field.type->get_primitive_type() == TYPE_VARBINARY) {
- // New FE versions mark every Iceberg UUID/BINARY/FIXED default as
Base64 regardless of its
- // Doris mapping. Keep the VARBINARY fallback for scan descriptors
produced before that
- // marker existed. Decode before parsing so STRING/CHAR and VARBINARY
all compare against
- // the raw bytes stored in equality-delete files.
- std::string decoded_default;
- if (!base64_decode(*table_field.initial_default_value,
&decoded_default)) {
- return Status::InvalidArgument("Invalid Base64 Iceberg initial
default for field {}",
- table_field.name);
- }
- if (table_field.type->get_primitive_type() == TYPE_VARBINARY) {
- initial_default =
Field::create_field<TYPE_VARBINARY>(StringView(decoded_default));
- } else {
-
DORIS_CHECK(is_string_type(table_field.type->get_primitive_type()));
- initial_default =
Field::create_field<TYPE_STRING>(decoded_default);
- }
- } else {
- // An added field's initial default is its logical value in every
older data file that lacks
- // the physical column. FE normalizes the string for the current Doris
table type.
- RETURN_IF_ERROR(table_field.type->get_serde()->from_fe_string(
- *table_field.initial_default_value, initial_default));
- }
+ ColumnPtr initial_default_column;
+ RETURN_IF_ERROR(doris::iceberg::parse_initial_default(
Review Comment:
[P1] Resolve nested equality-delete keys before applying defaults
This helper can now parse the propagated initial default, but both
equality-delete readers still resolve `equality_ids` only among top-level
fields: v2 scans the top-level delete/data schemas and `root_field.fields`, and
v1 has the same root-only current-schema lookup. The Iceberg specification
permits primitive equality keys nested through structs, including optional
parents. For an old data file missing newly added `s.c DEFAULT 7`, a valid
equality delete on field ID `s.c` therefore fails with `Can not find field id`
before this helper is reached instead of comparing against 7. Please resolve
field IDs recursively and retain the nested projection path for v1/v2, with an
old-file equality-delete regression.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
