database/mariadb
1
0
Fork 0
mirror of https://github.com/MariaDB/server.git synced 2025-08-01 03:47:19 +03:00
Code Activity

MDEV-11371 - column compression

Browse Source 
Storage engine independent support for column compression.

TINYBLOB, BLOB, MEDIUMBLOB, LONGBLOB, TINYTEXT, TEXT, MEDIUMTEXT, LONGTEXT,
VARCHAR and VARBINARY columns can be compressed.

New COMPRESSED column attribute added:
COMPRESSED[=<compression_method>]

System variables added:
column_compression_threshold
column_compression_zlib_level
column_compression_zlib_strategy
column_compression_zlib_wrap

Status variables added:
Column_compressions
Column_decompressions

Limitations:
- the only supported method currently is zlib
- CSV storage engine stores data uncompressed on-disk even if COMPRESSED
  attribute is present
- it is not possible to create indexes over compressed columns.
This commit is contained in:
Sergey Vojtovich
2017-04-24 17:54:18 +04:00
parent dd4e9cdded
commit fdc4779235
32 changed files with 2648 additions and 151 deletions
Download Patch File Download Diff File Expand all files Collapse all files

42
sql/rpl_utility.cc
View File

@ -128,6 +128,8 @@ max_display_length_for_field(enum_field_types sql_type, unsigned int metadata)
case MYSQL_TYPE_VAR_STRING:
case MYSQL_TYPE_VARCHAR:
return metadata;
case MYSQL_TYPE_VARCHAR_COMPRESSED:
return metadata - 1;
/*
The actual length for these types does not really matter since
@ -145,6 +147,7 @@ max_display_length_for_field(enum_field_types sql_type, unsigned int metadata)
return my_set_bits(3 * 8);
case MYSQL_TYPE_BLOB:
case MYSQL_TYPE_BLOB_COMPRESSED:
/*
For the blob type, Field::real_type() lies and say that all
blobs are of type MYSQL_TYPE_BLOB. In that case, we have to look
@ -294,6 +297,7 @@ uint32 table_def::calc_field_size(uint col, uchar *master_data) const
break;
}
case MYSQL_TYPE_VARCHAR:
case MYSQL_TYPE_VARCHAR_COMPRESSED:
{
length= m_field_metadata[col] > 255 ? 2 : 1; // c&p of Field_varstring::data_length()
length+= length == 1 ? (uint32) *master_data : uint2korr(master_data);
@ -303,6 +307,7 @@ uint32 table_def::calc_field_size(uint col, uchar *master_data) const
case MYSQL_TYPE_MEDIUM_BLOB:
case MYSQL_TYPE_LONG_BLOB:
case MYSQL_TYPE_BLOB:
case MYSQL_TYPE_BLOB_COMPRESSED:
case MYSQL_TYPE_GEOMETRY:
{
/*
@ -406,11 +411,14 @@ void show_sql_type(enum_field_types type, uint16 metadata, String *str, CHARSET_
case MYSQL_TYPE_VAR_STRING:
case MYSQL_TYPE_VARCHAR:
case MYSQL_TYPE_VARCHAR_COMPRESSED:
{
CHARSET_INFO *cs= str->charset();
uint32 length=
cs->cset->snprintf(cs, (char*) str->ptr(), str->alloced_length(),
"varchar(%u)", metadata);
"varchar(%u)%s", metadata,
type == MYSQL_TYPE_VARCHAR_COMPRESSED ? " compressed"
: "");
str->length(length);
}
break;
@ -455,6 +463,7 @@ void show_sql_type(enum_field_types type, uint16 metadata, String *str, CHARSET_
break;
case MYSQL_TYPE_BLOB:
case MYSQL_TYPE_BLOB_COMPRESSED:
/*
Field::real_type() lies regarding the actual type of a BLOB, so
it is necessary to check the pack length to figure out what kind
@ -482,6 +491,9 @@ void show_sql_type(enum_field_types type, uint16 metadata, String *str, CHARSET_
DBUG_ASSERT(0);
break;
}
if (type == MYSQL_TYPE_BLOB_COMPRESSED)
str->append(STRING_WITH_LEN(" compressed"));
break;
case MYSQL_TYPE_STRING:
@ -583,6 +595,7 @@ can_convert_field_to(Field *field,
int *order_var)
{
DBUG_ENTER("can_convert_field_to");
bool same_type;
#ifndef DBUG_OFF
char field_type_buf[MAX_FIELD_WIDTH];
String field_type(field_type_buf, sizeof(field_type_buf), &my_charset_latin1);
@ -590,11 +603,30 @@ can_convert_field_to(Field *field,
DBUG_PRINT("enter", ("field_type: %s, target_type: %d, source_type: %d, source_metadata: 0x%x",
field_type.c_ptr_safe(), field->real_type(), source_type, metadata));
#endif
/**
@todo
Implement Field_varstring_cmopressed::real_type() and
Field_blob_compressed::real_type() properly. All occurencies
of Field::real_type() have to be inspected and adjusted if needed.
Until it is not ready we have to compare source_type against
binlog_type() when replicating from or to compressed data types.
@sa Comment for Field::binlog_type()
*/
if (source_type == MYSQL_TYPE_VARCHAR_COMPRESSED ||
source_type == MYSQL_TYPE_BLOB_COMPRESSED ||
field->binlog_type() == MYSQL_TYPE_VARCHAR_COMPRESSED ||
field->binlog_type() == MYSQL_TYPE_BLOB_COMPRESSED)
same_type= field->binlog_type() == source_type;
else
same_type= field->real_type() == source_type;
/*
If the real type is the same, we need to check the metadata to
decide if conversions are allowed.
*/
if (field->real_type() == source_type)
if (same_type)
{
if (metadata == 0) // Metadata can only be zero if no metadata was provided
{
@ -731,18 +763,22 @@ can_convert_field_to(Field *field,
case MYSQL_TYPE_MEDIUM_BLOB:
case MYSQL_TYPE_LONG_BLOB:
case MYSQL_TYPE_BLOB:
case MYSQL_TYPE_BLOB_COMPRESSED:
case MYSQL_TYPE_STRING:
case MYSQL_TYPE_VAR_STRING:
case MYSQL_TYPE_VARCHAR:
case MYSQL_TYPE_VARCHAR_COMPRESSED:
switch (field->real_type())
{
case MYSQL_TYPE_TINY_BLOB:
case MYSQL_TYPE_MEDIUM_BLOB:
case MYSQL_TYPE_LONG_BLOB:
case MYSQL_TYPE_BLOB:
case MYSQL_TYPE_BLOB_COMPRESSED:
case MYSQL_TYPE_STRING:
case MYSQL_TYPE_VAR_STRING:
case MYSQL_TYPE_VARCHAR:
case MYSQL_TYPE_VARCHAR_COMPRESSED:
*order_var= compare_lengths(field, source_type, metadata);
/*
Here we know that the types are different, so if the order
@ -1036,6 +1072,7 @@ table_def::table_def(unsigned char *types, ulong size,
switch (binlog_type(i)) {
case MYSQL_TYPE_TINY_BLOB:
case MYSQL_TYPE_BLOB:
case MYSQL_TYPE_BLOB_COMPRESSED:
case MYSQL_TYPE_MEDIUM_BLOB:
case MYSQL_TYPE_LONG_BLOB:
case MYSQL_TYPE_DOUBLE:
@ -1066,6 +1103,7 @@ table_def::table_def(unsigned char *types, ulong size,
break;
}
case MYSQL_TYPE_VARCHAR:
case MYSQL_TYPE_VARCHAR_COMPRESSED:
{
/*
These types store two bytes.