[Feature] Support sort key be able to nullable. #15641
Conversation
|
clang-tidy review says "All clean, LGTM! 👍" |
|
Kudos, SonarCloud Quality Gate passed!
|
| buff.clear(); | ||
| for (int j = 0; j < ncol; j++) { | ||
| if (cols[j]->is_null(i)) { | ||
| buff.push_back(SORT_KEY_NULL_FIRST_MARKER); |
There was a problem hiding this comment.
If the data is also started as 0x00, how to handle it?
There was a problem hiding this comment.
If the data is also started as 0x00, how to handle it?
If we're a middle component of a composite key, we need to add a \x00 at the end in order to separate this component from the next one. However, if we just did that, we'd have issues where a key that actually has \x00 in it would compare wrong, so we have to instead add \x00\x00, and encode \x00 as \x00\x01.
| @@ -238,9 +241,6 @@ inline Status decode_slice(Slice* src, std::string* dest, bool is_last) { | |||
| } | |||
|
|
|||
| bool PrimaryKeyEncoder::is_supported(const VectorizedField& f) { | |||
| if (f.is_nullable()) { | |||
| return false; | |||
| } | |||
There was a problem hiding this comment.
Will this affect Primary Key's restriction? so primary key still don't support nulllable yet?
There was a problem hiding this comment.
Will this affect Primary Key's restriction? so primary key still don't support nulllable yet?
Yes, the keys of PrimaryKey table should not be nullable.
|
run starrocks_admit_test |
Support sort key be able to nullable. Signed-off-by: zhangqiang <qiangzh95@gmail.com>
Support sort key be able to nullable. Signed-off-by: zhangqiang <qiangzh95@gmail.com>
What type of PR is this：
Which issues of this PR fixes ：
Fixes #15057
close #14289
Problem Summary(Required) ：
Sort key should can be nullable, this PR support it. In primary key encoder, when encode a nullable column, if the datum is null, fill a null marker in the front of it, if the datum is not null, just fill a normal marker in the front of it.
Checklist:
Bugfix cherry-pick branch check: