[tarantool-patches] Re: [PATCH v1 4/4] sql: got rid of redundant bitmask helpers

Vladislav Shpilevoy v.shpilevoy at tarantool.org
Fri Feb 22 20:52:45 MSK 2019


During verbal discussion we decided, that there are no
places, where non-smart 64 bit mask is needed. But this
patch proves the contrary. Bitmask in most places of
its usage is not smart, and can not be smart, because
is used for only limited number of optimizations relying
on some limited resources like cursors.

So now for me it is obvious, that in such places it makes
no sense to use 64 bit mask - anyway smartness is not
available.

I think, you should keep Bitmask 32 bit. What is more,
32bit mask, introduced in the first commit, is never used
as a smart one. It means, that you all in all should finally
make any 32 bit mask as not smart, replace Bitmask with
uint32_t, and refactor its usages with new methods from the
first commit.

Personally, I still think, that we should introduce bitmask_t
and smartmask_t to explicitly distinguish between them and
never mix. And to be able to change their size with ease.
Because your assumption that we never need non-smart masks
appeared to be wrong.

Of course, you can consult the server team chat, Kostja, or
Kirill, if you do not want to do it.




More information about the Tarantool-patches mailing list