This is an automated email from the ASF dual-hosted git repository.
morningman pushed a commit to branch branch-0.15
in repository https://gitbox.apache.org/repos/asf/incubator-doris.git
commit 36ae508076dee8daa75c0f548821c979a26bcdb6
Author: jiafeng.zhang <zhang...@gmail.com>
AuthorDate: Thu Oct 28 13:56:59 2021 +0800
Modify Chinese comment (#6951)
Modify Chinese comment
---
be/src/olap/olap_common.h | 19 +++++-----
be/src/olap/olap_cond.cpp | 38 ++++++++++----------
be/src/olap/out_stream.cpp | 35 +++++++++---------
be/src/olap/out_stream.h | 56 ++++++++++++++---------------
be/src/olap/schema_change.cpp | 84 +++++++++++++++++++++----------------------
5 files changed, 117 insertions(+), 115 deletions(-)
diff --git a/be/src/olap/olap_common.h b/be/src/olap/olap_common.h
index 0769870..7fe970e 100644
--- a/be/src/olap/olap_common.h
+++ b/be/src/olap/olap_common.h
@@ -146,10 +146,10 @@ enum FieldType {
OLAP_FIELD_TYPE_STRING = 26
};
-// 定义Field支持的所有聚集方法
-// 注意，实际中并非所有的类型都能使用以下所有的聚集方法
-// 例如对于string类型使用SUM就是毫无意义的(但不会导致程序崩溃)
-// Field类的实现并没有进行这类检查，应该在创建表的时候进行约束
+// Define all aggregation methods supported by Field
+// Note that in practice, not all types can use all the following aggregation
methods
+// For example, it is meaningless to use SUM for the string type (but it will
not cause the program to crash)
+// The implementation of the Field class does not perform such checks, and
should be constrained when creating the table
enum FieldAggregationMethod {
OLAP_FIELD_AGGREGATION_NONE = 0,
OLAP_FIELD_AGGREGATION_SUM = 1,
@@ -163,11 +163,14 @@ enum FieldAggregationMethod {
OLAP_FIELD_AGGREGATION_REPLACE_IF_NOT_NULL = 8,
};
-// 压缩算法类型
+// Compression algorithm type
enum OLAPCompressionType {
- OLAP_COMP_TRANSPORT = 1, // 用于网络传输的压缩算法，压缩率低，cpu开销低
- OLAP_COMP_STORAGE = 2, // 用于硬盘数据的压缩算法，压缩率高，cpu开销大
- OLAP_COMP_LZ4 = 3, // 用于储存的压缩算法，压缩率低，cpu开销低
+ // Compression algorithm used for network transmission, low compression
rate, low cpu overhead
+ OLAP_COMP_TRANSPORT = 1,
+ // Compression algorithm used for hard disk data, with high compression
rate and high CPU overhead
+ OLAP_COMP_STORAGE = 2,
+ // The compression algorithm used for storage, the compression rate is
low, and the cpu overhead is low
+ OLAP_COMP_LZ4 = 3,
};
enum PushType {
diff --git a/be/src/olap/olap_cond.cpp b/be/src/olap/olap_cond.cpp
index c9184c3..4bc6b51 100644
--- a/be/src/olap/olap_cond.cpp
+++ b/be/src/olap/olap_cond.cpp
@@ -35,25 +35,25 @@ using std::vector;
using doris::ColumnStatistics;
-//此文件主要用于对用户发送的查询条件和删除条件进行处理，逻辑上二者都可以分为三层
+//This file is mainly used to process query conditions and delete conditions
sent by users. Logically, both can be divided into three layers
//Condition->Condcolumn->Cond
-//Condition表示用户发的单个条件
-//Condcolumn表示一列上所有条件的集合。
-//Conds表示一列上的单个条件.
-//对于查询条件而言，各层级的条件之间都是逻辑与的关系
-//对于delete条件则有不同。Cond和Condcolumn之间是逻辑与的关系，而Condtion之间是逻辑或的关系。
-
-//具体到实现。
-//eval是用来过滤查询条件，包括堆row、block、version的过滤，具体使用哪一层看具体的调用地方。
-// 1. 没有单独过滤行的过滤条件，这部分在查询层进行。
-// 2. 过滤block在SegmentReader里面。
-// 3. 过滤version在Reader里面。调用delta_pruing_filter
+//Condition represents a single condition sent by the user
+//Condcolumn represents the collection of all conditions on a column.
+//Conds represents a single condition on a column.
+//For query conditions, the conditions of each level are logical AND
relationships
+//There are different conditions for delete. The relationship between Cond and
Condcolumn is logical AND, and the relationship between Condtion is logical OR.
+
+//Specific to the realization.
+//eval is used to filter query conditions, including the filtering of heap
row, block, and version. Which layer is used depends on the specific calling
place.
+// 1. There is no filter condition to filter rows separately, this part is
carried out in the query layer.
+// 2. The filter block is in the SegmentReader.
+// 3. Filter version in Reader. Call delta_pruing_filter
//
-//del_eval用来过滤删除条件，包括堆block和version的过滤，但是这个过滤比eval多了一个状态，即部分过滤。
-// 1. 对行的过滤在DeleteHandler。
-// 这部分直接调用delete_condition_eval实现,内部调用eval函数，因为对row的过滤不涉及部分过滤这种状态。
-// 2. 过滤block是在SegmentReader里面,直接调用del_eval
-// 3. 过滤version实在Reader里面,调用rowset_pruning_filter
+//del_eval is used to filter deletion conditions, including the filtering of
heap block and version, but this filtering has one more state than eval, that
is, partial filtering.
+// 1. The filtering of rows is in DeleteHandler.
+// This part directly calls delete_condition_eval to achieve, and internally
calls the eval function, because the filtering of row does not involve partial
filtering.
+// 2. The filter block is in the SegmentReader, call del_eval directly
+// 3. The filter version is actually in Reader, call rowset_pruning_filter
namespace doris {
@@ -176,7 +176,7 @@ OLAPStatus Cond::init(const TCondition& tcond, const
TabletColumn& column) {
bool Cond::eval(const RowCursorCell& cell) const {
if (cell.is_null() && op != OP_IS) {
- //任何非OP_IS operand和NULL的运算都是false
+ //Any operation other than OP_IS operand and NULL is false
return false;
}
@@ -215,7 +215,7 @@ bool Cond::eval(const RowCursorCell& cell) const {
}
bool Cond::eval(const std::pair<WrapperField*, WrapperField*>& statistic)
const {
- //通过单列上的单个查询条件对version进行过滤
+ //A single query condition filtered by a single column
// When we apply column statistic, Field can be NULL when type is Varchar,
// we just ignore this cond
if (statistic.first == nullptr || statistic.second == nullptr) {
diff --git a/be/src/olap/out_stream.cpp b/be/src/olap/out_stream.cpp
index 8882085..b6900c6 100644
--- a/be/src/olap/out_stream.cpp
+++ b/be/src/olap/out_stream.cpp
@@ -186,22 +186,22 @@ OLAPStatus OutStream::_spill() {
return OLAP_SUCCESS;
}
- // 如果不压缩，直接读取current，注意output之后 current会被清空并设置为NULL
+ // If it is not compressed, read current directly. Note that current will
be cleared and set to NULL after output
if (_compressor == NULL) {
_current->flip();
_output_uncompress();
} else {
- // 如果需要压缩，
- // current移动到head后边的位置，留出head的空间
+ //If compression is required,
+ // The current moves to the position behind the head, leaving space
for the head
_current->set_limit(_current->position());
_current->set_position(sizeof(StreamHead));
- // 分配compress和overflow，这两个buffer大小其实是一样的
+ //Allocate compress and overflow, the two buffer sizes are actually
the same
if (OLAP_SUCCESS != (res = _make_sure_output_buffer())) {
return res;
}
- // 吧 current解压到compress和overflow
+ // Decompress current to compress and overflow
uint64_t head_pos = _compressed->position();
_compressed->set_position(head_pos + sizeof(StreamHead));
bool smaller = false;
@@ -213,8 +213,8 @@ OLAPStatus OutStream::_spill() {
}
if (smaller) {
- // 数据都压缩到_output和_overflow里, 重置_current
- // 注意这种情况下，current并没有被释放，因为实际上输出的compress
+ // Data are compressed into _output and _overflow, reset _current
+ // Note that in this case, current is not released, because the
compress actually output
_current->set_position(sizeof(StreamHead));
_current->set_limit(_current->capacity());
@@ -228,15 +228,14 @@ OLAPStatus OutStream::_spill() {
_spilled_bytes += sizeof(StreamHead) + output_bytes;
} else {
- // 直接将_current输出
-
- // 如果之前还有_compress, 先输出m_compress
- // 注意此时一定没有_overflow
+ // directly output _current
+ // If there is _compress before, output m_compress first
+ // Note that there must be no _overflow at this time
_compressed->set_position(head_pos);
if (head_pos != 0) {
- // 之前_compressed里有数据, 这种情况下先输出compressed,
- // 此时_overflow一定是空的
+ // There was data in _compressed before, in this case, output
compressed first,
+ // At this time _overflow must be empty
_output_compressed();
}
@@ -253,11 +252,11 @@ OLAPStatus OutStream::write(const char* buffer, uint64_t
length) {
uint64_t remain = length;
while (remain > 0) {
- // 之所以扔进来，是因为在压缩的情况下，_current只会被创建一次
- // 之后一直在复用，输出的是compress
- // 而在未压缩的情况下，current会被放进列表，而无法复用，原因是
- // 如果复用的话，会修改之前的内容，因此需要重新分配。
- // 只分配一次那么第二块就会挂掉
+ // The reason why it was thrown in is because in the case of
compression, _current will only be created once
+ // It has been multiplexing since then, and the output is compress
+ // In the case of uncompressed, current will be put into the list and
cannot be reused. The reason is
+ // If it is reused, the previous content will be modified, so it
needs to be redistributed.
+ // Only allocate once and the second block will hang up
if (NULL == _current) {
res = _create_new_input_buffer();
if (OLAP_SUCCESS != res) {
diff --git a/be/src/olap/out_stream.h b/be/src/olap/out_stream.h
index 6473091..2376758 100644
--- a/be/src/olap/out_stream.h
+++ b/be/src/olap/out_stream.h
@@ -27,26 +27,26 @@
namespace doris {
class FileHandler;
-// 与OrcFile不同,我们底层没有HDFS无法保证存储数据的可靠性,所以必须写入
-// 校验值,在读取数据的时候检验这一校验值
-// 采用TLV类型的头部,有足够的扩展性
+// Unlike OrcFile, we cannot guarantee the reliability of stored data without
HDFS at the bottom, so we must write
+// Check value, check this check value when reading data
+// Adopt TLV type header, which has sufficient scalability
struct StreamHead {
enum StreamType { UNCOMPRESSED = 0, COMPRESSED = 1 };
- uint8_t type; // 256种类型, 应该足够以后的扩展了
- uint32_t length : 24; // 24位长度
- uint32_t checksum; // 32位校验值
+ uint8_t type; // 256 types, should be enough for future expansion
+ uint32_t length : 24; // 24-bit length
+ uint32_t checksum; // 32-bit check value
StreamHead() : type(COMPRESSED), length(0), checksum(0) {}
} __attribute__((packed));
-// 输出流,使用一组ByteBuffer缓存所有的数据
+// Output stream, use a set of ByteBuffer to buffer all data
class OutStream {
public:
- // 输出流支持压缩或者不压缩两种模式,如果启用压缩,给出压缩函数
+ // The output stream supports two modes: compressed or uncompressed. If
compression is enabled, the compression function is given
explicit OutStream(uint32_t buffer_size, Compressor compressor);
~OutStream();
- // 向流输出一个字节
+ // Output a byte to the stream
inline OLAPStatus write(char byte) {
OLAPStatus res = OLAP_SUCCESS;
if (_current == nullptr) {
@@ -71,26 +71,26 @@ public:
return _current->put(byte);
}
- // 向流输出一段数据
+ // Output a piece of data to the stream
OLAPStatus write(const char* buffer, uint64_t length);
- // 将流的当前位置记录在索引项中
+ // Record the current position of the stream in the index entry
void get_position(PositionEntryWriter* index_entry) const;
- // 返回流中所有数据的大小
+ // Returns the size of all data in the stream
uint64_t get_stream_length() const;
- // 返回已经分配的缓冲区大小
+ // Returns the size of the buffer that has been allocated
uint64_t get_total_buffer_size() const;
- // 将缓存的数据流输出到文件
+ // Output the cached data stream to a file
OLAPStatus write_to_file(FileHandler* file_handle, uint32_t
write_mbytes_per_sec) const;
bool is_suppressed() const { return _is_suppressed; }
void suppress() { _is_suppressed = true; }
- // 将数据输出到output_buffers
+ // Output data to output_buffers
OLAPStatus flush();
- // 计算输出数据的crc32值
+ // Calculate the crc32 value of the output data
uint32_t crc32(uint32_t checksum) const;
const std::vector<StorageByteBuffer*>& output_buffers() { return
_output_buffers; }
@@ -115,33 +115,33 @@ private:
void _output_compressed();
OLAPStatus _make_sure_output_buffer();
- uint32_t _buffer_size; // 压缩块大小
- Compressor _compressor; // 压缩函数,如果为NULL表示不压缩
- std::vector<StorageByteBuffer*> _output_buffers; // 缓冲所有的输出
- bool _is_suppressed; // 流是否被终止
- StorageByteBuffer* _current; // 缓存未压缩的数据
- StorageByteBuffer* _compressed; // 即将输出到output_buffers中的字节
- StorageByteBuffer* _overflow; // _output中放不下的字节
- uint64_t _spilled_bytes; // 已经输出到output的字节数
+ uint32_t _buffer_size; // Compressed block size
+ Compressor _compressor; // Compression function,
if NULL means no compression
+ std::vector<StorageByteBuffer*> _output_buffers; // Buffer all output
+ bool _is_suppressed; // Whether the stream is
terminated
+ StorageByteBuffer* _current; // Cache uncompressed data
+ StorageByteBuffer* _compressed; // Bytes to be output to
output_buffers
+ StorageByteBuffer* _overflow; // Bytes that can't fit
in _output
+ uint64_t _spilled_bytes; // The number of bytes
that have been output to output
DISALLOW_COPY_AND_ASSIGN(OutStream);
};
-// 定义输出流的工厂方法
-// 将所有的输出流托管,同时封装了诸如压缩算法,是否启用Index,block大小等信息
+// Define the factory method of the output stream
+// Host all output streams, and encapsulate information such as compression
algorithm, whether to enable Index, block size, etc.
class OutStreamFactory {
public:
explicit OutStreamFactory(CompressKind compress_kind, uint32_t
stream_buffer_size);
~OutStreamFactory();
- // 创建后的stream的生命期依旧由OutStreamFactory管理
+ //The lifetime of the stream after creation is still managed by
OutStreamFactory
OutStream* create_stream(uint32_t column_unique_id,
StreamInfoMessage::Kind kind);
const std::map<StreamName, OutStream*>& streams() const { return _streams;
}
private:
- std::map<StreamName, OutStream*> _streams; // 所有创建过的流
+ std::map<StreamName, OutStream*> _streams; // All created streams
CompressKind _compress_kind;
Compressor _compressor;
uint32_t _stream_buffer_size;
diff --git a/be/src/olap/schema_change.cpp b/be/src/olap/schema_change.cpp
index 49a7f96..8a1eb34 100644
--- a/be/src/olap/schema_change.cpp
+++ b/be/src/olap/schema_change.cpp
@@ -466,13 +466,13 @@ OLAPStatus RowBlockChanger::change_row_block(const
RowBlock* ref_block, int32_t
return OLAP_ERR_NOT_INITED;
}
- // a.1 先判断数据是否需要过滤，最终只有标记为1的才是留下需要的
- // 对于没有filter的来说，相当于全部设置为1后留下
+ // a.1 First determine whether the data needs to be filtered, and finally
only those marked as 1 are left as needed
+ // For those without filter, it is equivalent to leave after setting all
to 1
const uint32_t row_num = ref_block->row_block_info().row_num;
- // (0表示过滤掉不要，1表示要,过程中2表示此row要切后续不需要再比较其他列)
+ // (0 means no need to filter out, 1 means yes, during the process 2 means
that this row needs to be cut and there is no need to compare other columns
later)
std::vector<int8_t> is_data_left_vec(row_num, 1);
- // 一行一行地进行比较
+ // Compare each row
for (size_t row_index = 0; row_index < row_num; ++row_index) {
ref_block->get_row(row_index, &read_helper);
@@ -486,14 +486,14 @@ OLAPStatus RowBlockChanger::change_row_block(const
RowBlock* ref_block, int32_t
}
}
- // a.2 计算留下的row num
+ // a.2 Calculate the left row num
uint32_t new_row_num = row_num - *filtered_rows;
const bool need_filter_data = (new_row_num != row_num);
const bool filter_all = (new_row_num == 0);
MemPool* mem_pool = mutable_block->mem_pool();
- // b. 根据前面的过滤信息，只对还标记为1的处理
+ // b. According to the previous filtering information, only processes that
are also marked as 1
for (size_t i = 0, len = mutable_block->tablet_schema().num_columns();
!filter_all && i < len;
++i) {
int32_t ref_column = _schema_mapping[i].ref_column;
@@ -537,15 +537,15 @@ OLAPStatus RowBlockChanger::change_row_block(const
RowBlock* ref_block, int32_t
FieldType reftype =
ref_block->tablet_schema().column(ref_column).type();
FieldType newtype =
mutable_block->tablet_schema().column(i).type();
if (newtype == reftype) {
- // 效率低下，也可以直接计算变长域拷贝，但仍然会破坏封装
+ // Low efficiency, you can also directly calculate the
variable length domain copy, but it will still destroy the package
for (size_t row_index = 0, new_row_index = 0;
row_index < ref_block->row_block_info().row_num;
++row_index) {
- // 不需要的row，每次处理到这个row时就跳过
+ // Unneeded row, skip every time this row is processed
if (need_filter_data && is_data_left_vec[row_index] == 0) {
continue;
}
- // 指定新的要写入的row index（不同于读的row_index）
+ // Specify the new row index to be written (different from
the read row_index)
mutable_block->get_row(new_row_index++, &write_helper);
ref_block->get_row(row_index, &read_helper);
@@ -597,10 +597,10 @@ OLAPStatus RowBlockChanger::change_row_block(const
RowBlock* ref_block, int32_t
}
}
}
- // 从ref_column 写入 i列。
+ // Write column i from ref_column.
} else {
// copy and alter the field
- // 此处可以暂时不动，新类型暂时不涉及类型转换
+ // You can stay here for the time being, the new type does not
involve type conversion for the time being
switch (reftype) {
case OLAP_FIELD_TYPE_TINYINT:
CONVERT_FROM_TYPE(int8_t);
@@ -637,10 +637,10 @@ OLAPStatus RowBlockChanger::change_row_block(const
RowBlock* ref_block, int32_t
}
}
} else {
- // 新增列，写入默认值
+ // New column, write default value
for (size_t row_index = 0, new_row_index = 0;
row_index < ref_block->row_block_info().row_num; ++row_index)
{
- // 不需要的row，每次处理到这个row时就跳过
+ // Unneeded row, skip every time this row is processed
if (need_filter_data && is_data_left_vec[row_index] == 0) {
continue;
}
@@ -658,9 +658,9 @@ OLAPStatus RowBlockChanger::change_row_block(const
RowBlock* ref_block, int32_t
}
}
- // NOTE 当前mutable_block的内存row_num还是和ref一样多
- // （其实在init时就可以重新init成少的，filter留下的new_row_num）
- // 在split_table时，可能会出现因为过滤导致没有数据
+ // NOTE The current row_num of mutable_block is still as much as ref
+ // (Actually, you can re-init into less when init, the new_row_num left by
the filter)
+ // In split_table, there may be no data due to filtering
mutable_block->finalize(new_row_num);
return OLAP_SUCCESS;
}
@@ -766,7 +766,7 @@ OLAPStatus RowBlockAllocator::allocate(RowBlock**
row_block, size_t num_rows, bo
return OLAP_SUCCESS;
}
- // TODO(lijiao) : 为什么舍弃原有的m_row_block_buffer
+ // TODO(lijiao) : Why abandon the original m_row_block_buffer
*row_block = new (nothrow) RowBlock(&_tablet_schema);
if (*row_block == nullptr) {
@@ -1041,7 +1041,7 @@ OLAPStatus
SchemaChangeDirectly::process(RowsetReaderSharedPtr rowset_reader,
RETURN_NOT_OK(reserve_block(&new_row_block,
ref_row_block->row_block_info().row_num,
_row_block_allocator));
- // 将ref改为new。这一步按道理来说确实需要等大的块，但理论上和writer无关。
+ // Change ref to new. This step is reasonable to say that it does need
to wait for a large block, but theoretically it has nothing to do with the
writer.
uint64_t filtered_rows = 0;
res = _row_block_changer.change_row_block(ref_row_block,
rowset_reader->version().second,
new_row_block.get(),
&filtered_rows);
@@ -1096,10 +1096,10 @@ SchemaChangeWithSorting::SchemaChangeWithSorting(const
RowBlockChanger& row_bloc
_row_block_changer(row_block_changer),
_memory_limitation(memory_limitation),
_row_block_allocator(nullptr) {
- // 每次SchemaChange做外排的时候，会写一些临时版本（比如999,1000,1001），为避免Cache冲突，临时
- // 版本进行2个处理：
- // 1. 随机值作为VersionHash
- // 2. 版本号取一个BIG NUMBER加上当前正在进行SchemaChange的版本号
+ // Every time SchemaChange is used for external rowing, some temporary
versions (such as 999, 1000, 1001) will be written, in order to avoid Cache
conflicts, temporary
+ // The version performs 2 processes:
+ // 1. Random value as VersionHash
+ // 2. The version number takes a BIG NUMBER plus the version number of the
current SchemaChange
_temp_delta_versions.first = (1 << 28);
_temp_delta_versions.second = (1 << 28);
// TODO(zyh): remove the magic number
@@ -1666,8 +1666,8 @@ OLAPStatus
SchemaChangeHandler::schema_version_convert(TabletSharedPtr base_tabl
<< "base_tablet=" << base_tablet->full_name()
<< ", new_tablet=" << new_tablet->full_name();
- // a. 解析Alter请求，转换成内部的表示形式
- // 不使用DELETE_DATA命令指定的删除条件
+ // a. Parse the Alter request and convert it into an internal
representation
+ // Do not use the delete condition specified by the DELETE_DATA command
RowBlockChanger rb_changer(new_tablet->tablet_schema());
bool sc_sorting = false;
bool sc_directly = false;
@@ -1679,9 +1679,9 @@ OLAPStatus
SchemaChangeHandler::schema_version_convert(TabletSharedPtr base_tabl
return res;
}
- // NOTE split_table如果使用row_block，会导致原block变小
- // 但由于历史数据在后续base/cumulative后还是会变成正常，故用directly也可以
- // b. 生成历史数据转换器
+ // NOTE split_table if row_block is used, the original block will become
smaller
+ // But since the historical data will become normal after the subsequent
base/cumulative, it is also possible to use directly
+ // b. Generate historical data converter
SchemaChange* sc_procedure = nullptr;
if (sc_sorting) {
size_t memory_limitation =
config::memory_limitation_per_thread_for_schema_change;
@@ -1702,7 +1702,7 @@ OLAPStatus
SchemaChangeHandler::schema_version_convert(TabletSharedPtr base_tabl
return OLAP_ERR_MALLOC_ERROR;
}
- // c. 转换数据
+ // c. Convert data
DeleteHandler delete_handler;
std::vector<ColumnId> return_columns;
size_t num_cols = base_tablet->tablet_schema().num_columns();
@@ -1817,15 +1817,15 @@ OLAPStatus
SchemaChangeHandler::_convert_historical_rowsets(const SchemaChangePa
}
}
- // change中增加了filter信息，在_parse_request中会设置filter的column信息
- // 并在每次row block的change时，过滤一些数据
+ // Add filter information in change, and filter column information will be
set in _parse_request
+ // And filter some data every time the row block changes
RowBlockChanger rb_changer(sc_params.new_tablet->tablet_schema(),
sc_params.delete_handler);
bool sc_sorting = false;
bool sc_directly = false;
SchemaChange* sc_procedure = nullptr;
- // a. 解析Alter请求，转换成内部的表示形式
+ // a.Parse the Alter request and convert it into an internal representation
OLAPStatus res = _parse_request(sc_params.base_tablet,
sc_params.new_tablet, &rb_changer,
&sc_sorting, &sc_directly,
sc_params.materialized_params_map);
if (res != OLAP_SUCCESS) {
@@ -1833,7 +1833,7 @@ OLAPStatus
SchemaChangeHandler::_convert_historical_rowsets(const SchemaChangePa
goto PROCESS_ALTER_EXIT;
}
- // b. 生成历史数据转换器
+ // b. Generate historical data converter
if (sc_sorting) {
size_t memory_limitation =
config::memory_limitation_per_thread_for_schema_change;
LOG(INFO) << "doing schema change with sorting for base_tablet "
@@ -1857,14 +1857,14 @@ OLAPStatus
SchemaChangeHandler::_convert_historical_rowsets(const SchemaChangePa
goto PROCESS_ALTER_EXIT;
}
- // c. 转换历史数据
+ // c.Convert historical data
for (auto& rs_reader : sc_params.ref_rowset_readers) {
VLOG_TRACE << "begin to convert a history rowset. version=" <<
rs_reader->version().first
<< "-" << rs_reader->version().second;
// set status for monitor
- // 只要有一个new_table为running，ref table就设置为running
- // NOTE 如果第一个sub_table先fail，这里会继续按正常走
+ // As long as there is a new_table as running, ref table is set as
running
+ // NOTE If the first sub_table fails first, it will continue to go as
normal here
TabletSharedPtr new_tablet = sc_params.new_tablet;
RowsetWriterContext writer_context;
@@ -1906,8 +1906,8 @@ OLAPStatus
SchemaChangeHandler::_convert_historical_rowsets(const SchemaChangePa
}
new_tablet->data_dir()->remove_pending_ids(ROWSET_ID_PREFIX +
rowset_writer->rowset_id().to_string());
- // 将新版本的数据加入header
- // 为了防止死锁的出现，一定要先锁住旧表，再锁住新表
+ // Add the new version of the data to the header
+ // In order to prevent the occurrence of deadlock, we must first lock
the old table, and then lock the new table
sc_params.new_tablet->obtain_push_lock();
RowsetSharedPtr new_rowset = rowset_writer->build();
if (new_rowset == nullptr) {
@@ -1941,7 +1941,7 @@ OLAPStatus
SchemaChangeHandler::_convert_historical_rowsets(const SchemaChangePa
<< " version=" << rs_reader->version().first << "-"
<< rs_reader->version().second;
}
- // XXX: 此时应该不取消SchemaChange状态，因为新Delta还要转换成新旧Schema的版本
+ // XXX:The SchemaChange state should not be canceled at this time, because
the new Delta has to be converted to the old and new Schema version
PROCESS_ALTER_EXIT : {
// save tablet meta here because rowset meta is not saved during add rowset
WriteLock new_wlock(sc_params.new_tablet->get_header_lock_ptr());
@@ -1960,7 +1960,7 @@ PROCESS_ALTER_EXIT : {
}
// @static
-// 分析column的mapping以及filter key的mapping
+// Analyze the mapping of the column and the mapping of the filter key
OLAPStatus SchemaChangeHandler::_parse_request(
TabletSharedPtr base_tablet, TabletSharedPtr new_tablet,
RowBlockChanger* rb_changer,
bool* sc_sorting, bool* sc_directly,
@@ -2014,7 +2014,7 @@ OLAPStatus SchemaChangeHandler::_parse_request(
continue;
}
- // 新加列走这里
+ // Newly added column go here
//if (new_column_schema.is_allow_null ||
new_column_schema.has_default_value) {
{
column_mapping->ref_column = -1;
@@ -2034,7 +2034,7 @@ OLAPStatus SchemaChangeHandler::_parse_request(
continue;
}
- // XXX: 只有DROP COLUMN时，遇到新Schema转旧Schema时会进入这里。
+ // XXX: Only when DROP COLUMN, you will enter here when you encounter
a new Schema to an old Schema。
column_mapping->ref_column = -1;
if (OLAP_SUCCESS != (res = _init_column_mapping(column_mapping,
new_column, ""))) {
@@ -2048,7 +2048,7 @@ OLAPStatus SchemaChangeHandler::_parse_request(
// Check if re-aggregation is needed.
*sc_sorting = false;
- // 若Key列的引用序列出现乱序，则需要重排序
+ // If the reference sequence of the Key column is out of order, it needs
to be reordered
int num_default_value = 0;
for (int i = 0, new_schema_size = new_tablet->num_key_columns(); i <
new_schema_size; ++i) {
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@doris.apache.org
For additional commands, e-mail: commits-h...@doris.apache.org
