Fix counting column size in wide part for Dynamid and JSON types

This commit is contained in:
avogar 2024-11-06 15:15:29 +00:00
parent ea806b05a0
commit 020b69647a
4 changed files with 25 additions and 2 deletions

View File

@ -108,7 +108,7 @@ ColumnSize MergeTreeDataPartWide::getColumnSizeImpl(
auto mrk_checksum = checksums.files.find(*stream_name + getMarksFileExtension()); auto mrk_checksum = checksums.files.find(*stream_name + getMarksFileExtension());
if (mrk_checksum != checksums.files.end()) if (mrk_checksum != checksums.files.end())
size.marks += mrk_checksum->second.file_size; size.marks += mrk_checksum->second.file_size;
}); }, column.type, getColumnSample(column));
return size; return size;
} }

View File

@ -172,7 +172,7 @@ size_t MergeTreeReaderWide::readRows(
throw; throw;
} }
if (column->empty()) if (column->empty() && max_rows_to_read > 0)
res_columns[pos] = nullptr; res_columns[pos] = nullptr;
} }

View File

@ -0,0 +1 @@
test 10.00 million 352.87 MiB 39.43 MiB 39.45 MiB

View File

@ -0,0 +1,22 @@
-- Tags: no-random-settings
set allow_experimental_dynamic_type = 1;
set allow_experimental_json_type = 1;
drop table if exists test;
create table test (d Dynamic, json JSON) engine=MergeTree order by tuple() settings min_rows_for_wide_part=0, min_bytes_for_wide_part=1;
insert into test select number, '{"a" : 42, "b" : "Hello, World"}' from numbers(10000000);
SELECT
`table`,
formatReadableQuantity(sum(rows)) AS rows,
formatReadableSize(sum(data_uncompressed_bytes)) AS data_size_uncompressed,
formatReadableSize(sum(data_compressed_bytes)) AS data_size_compressed,
formatReadableSize(sum(bytes_on_disk)) AS total_size_on_disk
FROM system.parts
WHERE active AND (database = currentDatabase()) AND (`table` = 'test')
GROUP BY `table`
ORDER BY `table` ASC;
drop table test;