Static tables

This section provides an introduction to static tables, and their types and schema.
In addition to static tables, the YTsaurus system supports dynamic tables. For more information about the types and features of dynamic tables, see Dynamic tables.

General description

The YTsaurus system can store various objects: files, documents, and tables.

There are two types of tables: static and dynamic.
Static tables are suitable for storing rarely changed data and enable you to write new data to the end of tables.
Dynamic tables are suitable for frequently changed data.

Static tables are a classic table type in the YTsaurus system.
Such tables are physically divided into parts (chunks) and each part contains a fragment of table entries.

You can change the static table data only when:

• Merging.
• Deleting.
• Adding entries to the end of the table.

Static tables can be sorted and unsorted.

Sorted tables

For a sorted table, a set of immutable key columns is known.
The table entries appear to be physically or logically ordered by key. Therefore, sorted tables enable you to effectively search for data by key.

A sorted table has at least one key column, the sorted attribute of the table is true. The attribute is read-only.

There are key (sorted) columns in the sorted table schema. They are marked with the sort_order field in the schema.

A list of all key columns is available in the key_columns and sorted_by attributes of the table. The sequence of columns is defined in the table schema. The attributes are read-only.

Unsorted tables

For unsorted tables, the concept of a key is not defined, so searching for data by key is ineffective. However, you can access it by row numbers.

Attributes

Any static table has the attributes represented in the table:

Name	Type	Description	Mandatory
sorted	bool	Whether the table is sorted.	No
key_columns	array<string>	Key column names.	Yes
dynamic	bool	Whether the table is dynamic.	No
schema	TableSchema	Table schema.	No
data_weight	integer	A "logical" amount of uncompressed data written to a table. Depends only on the values in the table cells and the number of rows. Calculated as row_count + sum(data_weight(value)) for all values in the table cells. data_weight for a value depends on the physical type of the value: for int64, uint64, and double — 8 bytes; for bool and null — 1 byte; for string — string length; for any — length of the value serialized in binary YSON.	Yes

Besides that, all tables are chunk owners. As a result, they get the corresponding attributes represented in the table.

Static table schema

A static table schema is a list of column descriptions. For a detailed description of the schema format, see Data schema.

Limitations

A number of size and content type limitations are imposed on table rows and schema:

• The number of columns in a static table cannot exceed 32,768. We do not recommend using more than a thousand columns.

• The column name must contain from 1 to 256 characters. The column name can be a random sequence of bytes. Cannot start with the reserved system prefix @.

• The maximum length of string values in a static table is limited by the maximum string weight.

• The maximum string weight is 128 megabytes. The string weight is the sum of the lengths of all values in the given string in bytes. The lengths of the values are counted depending on the type:

  • int64, uint64, and double: 8 bytes.
  • boolean: 1 byte.
  • string: String length.
  • any: The length of the structure serialized in binary yson, in bytes.
  • null: 0 bytes.

• The maximum key weight in a sorted table is 256 kilobytes. The default limitation is 16 kilobytes. The key weight is counted similarly to the string weight.

  Attention!

  We strongly advise against raising the limit on the key weight, because there is a risk of overflowing the master server memory. Changing the max_key_weight setting is only allowed as a hotfix.