thevar1able/ClickHouse
1
0
Fork 0
mirror of https://github.com/ClickHouse/ClickHouse.git synced 2024-12-01 12:01:58 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix alter column rename

Browse Source
This commit is contained in:
Anton Popov 2022-07-28 13:10:43 +00:00
parent 56039c8780
commit 04fd72cdef
5 changed files with 66 additions and 35 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
src/Storages/MergeTree/IMergeTreeDataPart.h
View File

@ -524,6 +524,7 @@ private:
/// Map from name of column to its serialization info.
SerializationInfoByName serialization_infos;
/// Serializations for every columns and subcolumns by their names.
SerializationByName serializations;
/// Columns description for more convenient access

2
src/Storages/MergeTree/IMergeTreeReader.cpp
View File

@ -40,6 +40,8 @@ IMergeTreeReader::IMergeTreeReader(
, metadata_snapshot(metadata_snapshot_)
, all_mark_ranges(all_mark_ranges_)
, alter_conversions(storage.getAlterConversionsForPart(data_part))
/// For wide parts convert plain arrays of Nested to subcolumns
/// to allow to use shared offset column from cache.
, requested_columns(isWidePart(data_part) ? Nested::convertToSubcolumns(columns_) : columns_)
, part_columns(isWidePart(data_part) ? Nested::collect(data_part->getColumns()) : data_part->getColumns())
{

3
src/Storages/MergeTree/IMergeTreeReader.h
View File

@ -75,7 +75,10 @@ protected:
/// Stores states for IDataType::deserializeBinaryBulk
DeserializeBinaryBulkStateMap deserialize_binary_bulk_state_map;
/// Actual column names and types of columns in part,
/// which may differ from table metadata.
NamesAndTypes columns_to_read;
/// Actual serialization of columns in part.
Serializations serializations;
UncompressedCache * uncompressed_cache;

2
src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp
View File

@ -583,7 +583,7 @@ void MergeTreeDataPartWriterWide::finishDataSerialization(bool sync)
{
if (column.type->isValueRepresentedByNumber()
&& !column.type->haveSubtypes()
&& data_part->getSerialization(columnn.name)->getKind() == ISerialization::Kind::DEFAULT)
&& data_part->getSerialization(column.name)->getKind() == ISerialization::Kind::DEFAULT)
{
validateColumnOfFixedSize(column);
}

61
src/Storages/MergeTree/MutateTask.cpp
View File

@ -459,8 +459,21 @@ static NameToNameVector collectFilesForRenames(
const MutationCommands & commands_for_removes,
const String & mrk_extension)
{
/// Collect counts for shared streams of different columns. As an example, Nested columns have shared stream with array sizes.
std::unordered_map<String, size_t> stream_counts;
for (const auto & column : source_part->getColumns())
{
if (auto serialization = source_part->tryGetSerialization(column.name))
{
serialization->enumerateStreams(
[&](const ISerialization::SubstreamPath & substream_path)
{
++stream_counts[ISerialization::getFileNameForStream(column, substream_path)];
});
}
}
NameToNameVector rename_vector;
NameSet renamed_streams;
/// Remove old data
for (const auto & command : commands_for_removes)
@ -483,6 +496,22 @@ static NameToNameVector collectFilesForRenames(
if (source_part->checksums.has(command.column_name + ".proj"))
rename_vector.emplace_back(command.column_name + ".proj", "");
}
else if (command.type == MutationCommand::Type::DROP_COLUMN)
{
ISerialization::StreamCallback callback = [&](const ISerialization::SubstreamPath & substream_path)
{
String stream_name = ISerialization::getFileNameForStream({command.column_name, command.data_type}, substream_path);
/// Delete files if they are no longer shared with another column.
if (--stream_counts[stream_name] == 0)
{
rename_vector.emplace_back(stream_name + ".bin", "");
rename_vector.emplace_back(stream_name + mrk_extension, "");
}
};
if (auto serialization = source_part->tryGetSerialization(command.column_name))
serialization->enumerateStreams(callback);
}
else if (command.type == MutationCommand::Type::RENAME_COLUMN)
{
String escaped_name_from = escapeForFileName(command.column_name);
@ -495,7 +524,6 @@ static NameToNameVector collectFilesForRenames(
if (stream_from != stream_to)
{
renamed_streams.insert(stream_from);
rename_vector.emplace_back(stream_from + ".bin", stream_to + ".bin");
rename_vector.emplace_back(stream_from + mrk_extension, stream_to + mrk_extension);
}
@ -504,42 +532,39 @@ static NameToNameVector collectFilesForRenames(
if (auto serialization = source_part->tryGetSerialization(command.column_name))
serialization->enumerateStreams(callback);
}
}
else if (command.type == MutationCommand::Type::READ_COLUMN)
{
/// Remove files for streams that exist in source_part,
/// but were removed in new_part by MODIFY COLUMN from
/// type with higher number of streams (e.g. LowCardinality -> String).
auto collect_all_stream_names = [&](const auto & data_part)
auto collect_stream_names = [&](const auto & data_part)
{
NameSet res;
for (const auto & column : data_part->getColumns())
{
if (auto serialization = data_part->tryGetSerialization(column.name))
if (auto serialization = data_part->tryGetSerialization(command.column_name))
{
serialization->enumerateStreams(
[&](const ISerialization::SubstreamPath & substream_path)
{
res.insert(ISerialization::getFileNameForStream(column.name, substream_path));
res.insert(ISerialization::getFileNameForStream(command.column_name, substream_path));
});
}
}
return res;
};
/// Remove files for streams that exists in source part,
/// but were removed in new_part by DROP COLUMN
/// or MODIFY COLUMN from type with higher number of streams
/// (e.g. LowCardinality -> String).
auto old_streams = collect_all_stream_names(source_part);
auto new_streams = collect_all_stream_names(new_part);
auto old_streams = collect_stream_names(source_part);
auto new_streams = collect_stream_names(new_part);
for (const auto & old_stream : old_streams)
{
if (!new_streams.contains(old_stream) && !renamed_streams.contains(old_stream))
if (!new_streams.contains(old_stream))
{
rename_vector.emplace_back(old_stream + ".bin", "");
rename_vector.emplace_back(old_stream + mrk_extension, "");
}
}
}
}
return rename_vector;
}