Compression suggestion

[filonenko-mikhail]

Motivation

There is compression feature in Tarantool EE version. Such feature works per fields. The motivation is to show this possibility for User.

(OBSOLETE, see notion https://www.notion.so/dd15fd804ef94d69b91205b35d5ec57a ) Propolsal

• Make a separated lua module

• Make a lua function which mayBeCompressed(space name, field name) -> float

• Select random 10000 tuples from space

  • Make iteration with yields every 1000

• Make a temporary space with the same format as space name

  • Make field name is_compressed = true

• Insert random 10000 tuples into temporary space

• Compare original space bsize and temporary space bsize

• Return compression ratio

• Make a iteration over spaces

• Choose only string fields

• Filter fields, choose only fields which is not in indexes

• Make a list with result [{name: aaa field:}, {}]

Corner cases

Handle cases:

• Skip empty spaces (space:len == 0)
• Skip system spaces ( id < 512 or name starts with _ (underscore)
• When iterate check that space exists. It can be dropped.
• The only function can be called only once at same time.
• It is possible that tuple has more fields than space:format
• It is possible that tuple has less fields than space:format