types: match MySQL behavior with datetime delimiters

[smola]

What problem does this PR solve?

Issue Number: close #17277

Problem Summary: select date '2020-02--1' fails

What is changed and how it works?

• types.ParseDateFormat changed datetime splitting to match MySQL 5.7.
• Any ASCII punctuation character (i.e. printable non-alphanumeric) is a valid delimiter.
• Additionally, (space) and T are valid delimiters if they appear between time and date.
• Multiple consecutive delimiters are accepted, and handled as a single delimiter.
• Consecutive delimiters can be mixed (e.g. -./ or T/TT).
• This PR does not apply to types.ParseDuration, since it has additional validation.

Check List

• Unit tests

Backwards compatibility

• In previous TiDB versions, T and (space) were valid delimiters anywhere in a datetime string. After this change, using T or (space) as a delimiter is an error except between date and time.

Release note

• Datetime parsing now matches MySQL 5.7 behavior for delimiters.

[CLAassistant]

All committers have signed the CLA.

[Reminiscent]

The unit-test is failed. Please fix them. Thanks!

[2020-05-25T07:44:53.516Z] FAIL: integration_test.go:1174: testIntegrationSuite2.TestTimeBuiltin

[2020-05-25T07:44:53.516Z] 

[2020-05-25T07:44:53.516Z] integration_test.go:1446:

[2020-05-25T07:44:53.516Z]     result = tk.MustQuery("select addtime(a, b), addtime(cast(a as date), b), addtime(b,a), addtime(a,c), addtime(b," +

[2020-05-25T07:44:53.516Z]         "c), addtime(c,a), addtime(c,b)" +

[2020-05-25T07:44:53.516Z]         " from t;")

[2020-05-25T07:44:53.516Z] /home/jenkins/agent/workspace/tidb_ghpr_check_2/go/src/github.com/pingcap/tidb/util/testkit/testkit.go:288:

[2020-05-25T07:44:53.516Z]     tk.c.Check(err, check.IsNil, comment)

[2020-05-25T07:44:53.516Z] ... value *errors.withStack = [types:1292]Incorrect time value: '{0 0 0 1 1 1 0}' ("[types:1292]Incorrect time value: '{0 0 0 1 1 1 0}'")

[2020-05-25T07:44:53.516Z] ... sql:select addtime(a, b), addtime(cast(a as date), b), addtime(b,a), addtime(a,c), addtime(b,c), addtime(c,a), addtime(c,b) from t;, args:[]

[2020-05-25T07:44:53.516Z] 

[2020-05-25T07:44:53.516Z] integration_test.go:1449:

[2020-05-25T07:44:53.516Z]     result.Check(testkit.Rows("<nil> <nil> <nil> 2017-01-01 13:31:32 2017-01-01 13:31:32 <nil> <nil>"))

[2020-05-25T07:44:53.516Z] /home/jenkins/agent/workspace/tidb_ghpr_check_2/go/src/github.com/pingcap/tidb/util/testkit/testkit.go:62:

[2020-05-25T07:44:53.516Z]     res.c.Assert(resBuff.String(), check.Equals, needBuff.String(), res.comment)

[2020-05-25T07:44:53.516Z] ... obtained string = ""

[2020-05-25T07:44:53.516Z] ... expected string = "[<nil> <nil> <nil> 2017-01-01 13:31:32 2017-01-01 13:31:32 <nil> <nil>]\n"

[2020-05-25T07:44:53.516Z] ... sql:select addtime(a, b), addtime(cast(a as date), b), addtime(b,a), addtime(a,c), addtime(b,c), addtime(c,a), addtime(c,b) from t;, args:[]

[smola]

@Reminiscent Oops. I missed that one. I have fixed the integration test, it used 2017 01-01 12:30:31 as a date, which produces an error in MySQL (see the initial notes about backwards compatibility). So I changed the test to 2017-01-01 12:30:31.

Just for the record, I tested on MySQL 8 with the following:

create database if not exists test;
use test;
drop table if exists t;
create table t(a datetime, b timestamp, c time);
insert into t values("2017 01-01 12:30:31", "2017 01-01 12:30:31", "01:01:01");

This produces:

Error Code: 1292. Incorrect datetime value: '2017 01-01 12:30:31' for column 'a' at row 1

I still have 4 unit tests failing locally, but they fail on master too. They are probably related to #17013

[codecov]

Codecov Report

Merging #17376 into master will not change coverage.
The diff coverage is n/a.

@@             Coverage Diff             @@
##             master     #17376   +/-   ##
===========================================
  Coverage   79.9031%   79.9031%           
===========================================
  Files           520        520           
  Lines        140385     140385           
===========================================
  Hits         112172     112172           
  Misses        19248      19248           
  Partials       8965       8965           

[wjhuang2016]

LGTM @Reminiscent PTAL

[Reminiscent]

LGTM. Great work!

[sre-bot]

/run-all-tests

[sre-bot]

@smola merge failed.

[Reminiscent]

/merge

[sre-bot]

/run-all-tests

[sre-bot]

cherry pick to release-3.0 in PR #17499

[sre-bot]

cherry pick to release-3.1 in PR #17500

[sre-bot]

cherry pick to release-4.0 in PR #17501