builtin: logic op not compatibility with float

[jingyugao]

Signed-off-by: jingyugao [email protected]

What problem does this PR solve?

#11552
logic operate not compatibility with float.

What is changed and how it works?

Make logic op cast args to real instead of int.
convert 0.1 to int will be 0, "0 or null" is null.
convert 0.1 to float will be 0.1, "0.1 or null" is 1.

Check List

Tests

• Unit test
  add unit test

Code changes
changed in function body

Side effects
no
Related changes
no

[sre-bot]

Hi contributor, thanks for your PR.

This patch needs to be approved by someone of admins. They should reply with "/ok-to-test" to accept this PR for running test automatically.

[CLAassistant]

All committers have signed the CLA.

[jingyugao]

According to my test,mysql will convert args of logic operato to float64.(not read the mysql source code)
For example:
select '1e-323' or nul; // 1
select '1e-324' or null // NULL

convert '1e-324' to float64 is 0. "0 or null" is null.

[codecov]

Codecov Report

Merging #11587 into master will not change coverage.
The diff coverage is n/a.

@@             Coverage Diff             @@
##             master     #11587   +/-   ##
===========================================
  Coverage   81.4299%   81.4299%           
===========================================
  Files           452        452           
  Lines         97194      97194           
===========================================
  Hits          79145      79145           
  Misses        12394      12394           
  Partials       5655       5655

[qw4990]

@jingyugao Thanks for your contribution, and we will review this PR soon.

[jingyugao]

gsed  -i 's/func (\(.\) \*\(.*\)) evalInt/func (\1 \*\2) evalReal(row chunk.Row) (float64, bool, error) { \narg, isNull, err := \1.evalInt(row) \nreturn float64(arg), isNull, err \n}\n\n func (\1 \*\2) evalInt/' expression/*.go

At last I use sed to add evalReal method for builtin*, it seems not bad.
Maybe evalBool is better than evalReal,but it seems not easy to hand.

[jingyugao]

By the way, this change fix a bug about unary not.

select not 0.1 // 0 in mysql
select not 0.1 // 1 in tidb befroe

because tidb convert 0.1 to 0.

[zz-jason]

@jingyugao Could you merge the master branch and resolve conflicts?

[qw4990]

Do we have to remove ast.And and ast.Or from canFuncBePushed since the implementation of these two functions are not consistent with TiKV now?
What's your opinion @XuHuaiyu ？

[jingyugao]

I refer to this pr, #10498 . ShouId keep the builtinLogicAndSig signature compatible to tiKV?

[qw4990]

@jingyugao Sorry for my late reply, this issue seems fixed by this PR/12173.

[jingyugao]

@jingyugao Sorry for my late reply, this issue seems fixed by this PR/12173.

sorry, too. busy these days.
Wish commit into TiDB next year.