Address issue #36118, allow "24:00" to be parsed.

[rohany]

Addressed #36118.

Achieved by special casing when the input string represents the time 24:00, and adjusting some maximum values for timeofday objects.

[cockroach-teamcity]

This change is 

[solongordon]

Nice job getting this up so quickly!

Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @rohany and @solongordon)

---

pkg/sql/sem/tree/datum.go, line 1823 at r1 (raw file):

	// special case on 24:00 and 24:00:00 as the parser
	// does not handle these correctly.
	if s == "24:00" || s == "24:00:00" {

We're getting into edge cases here but what about "24:00:00.000"? Might need a regular expression to catch all the cases unfortunately.

---

pkg/sql/sem/tree/datum.go, line 1824 at r1 (raw file):

	// does not handle these correctly.
	if s == "24:00" || s == "24:00:00" {
		return MakeDTime(timeofday.New(24, 0, 0, 0)), nil

timeofday.Max should work here

---

pkg/util/timeofday/time_of_day.go, line 72 at r1 (raw file):

	// 24:00 is a valid time, so differentiate it from 00:00
	if i == microsecondsPerDay {
		return TimeOfDay(i)

Nit: Could just return Max here.

---

pkg/util/timeofday/time_of_day.go, line 119 at r1 (raw file):

// Hour returns the hour specified by t, in the range [0, 24].
func (t TimeOfDay) Hour() int {
	if int64(t) == microsecondsPerDay {

Alternatively: if t == Max

---

pkg/util/timeofday/time_of_day.go, line 120 at r1 (raw file):

func (t TimeOfDay) Hour() int {
	if int64(t) == microsecondsPerDay {
		return int(int64(t) / microsecondsPerHour)

How about just return 24? No need to do the math. (Or maybe return hoursPerDay to avoid the magic number.)

[solongordon]

Great, getting close! I added a few more thoughts. Also, would you please change the commit and PR messages to match our standard? The subject should indicate the top level package affected, like sql: allow 24:00 to be parsed. The body should be wrapped at 72 characters. And rather than referencing the issue number in the subject, you can put Fixes #37898 as a separate line at the end.

More details (maybe too many) here: https://wiki.crdb.io/wiki/spaces/CRDB/pages/73072807/Git+Commit+Messages

Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @rohany)

---

pkg/sql/logictest/testdata/logic_test/time, line 32 at r4 (raw file):

statement error could not parse
SELECT ('24:00:01'::TIME);

Nit: Unnecessary parens

---

pkg/sql/logictest/testdata/logic_test/time, line 35 at r4 (raw file):

statement error could not parse
SELECT ('24:00:00.001'::TIME);

Nit: Unnecessary parens

---

pkg/sql/sem/tree/datum.go, line 1823 at r1 (raw file):

Previously, solongordon (Solon) wrote…

We're getting into edge cases here but what about "24:00:00.000"? Might need a regular expression to catch all the cases unfortunately.

Thanks for the regex. Couple more requests here:

• I think we should support any number of trailing zeroes, not just up to 6.
• Should probably add a ^ anchor at the beginning. (Seems like 124:00 would get parsed to 24:00 as it stands.)
• For efficiency we should probably compile the regex once up front as we discussed.
• Could you add a few tests in datum_test.go to make sure we're parsing the various forms correctly?

---

pkg/util/timeofday/time_of_day.go, line 56 at r4 (raw file):

	micros := time.Duration(micro) * time.Microsecond
	sum := int64((hours + minutes + seconds + micros) / time.Microsecond)
	return FromInt(sum)

Nit: Looks like this changed for no reason.

[solongordon]

One nit about the tests which you can fix if you want.

By the way, it's helpful if when you address review comments you reply to the comments with "Done." Otherwise the reviewer doesn't get notified that you fixed them.

Reviewable status: complete! 1 of 0 LGTMs obtained (waiting on @rohany)

---

pkg/sql/sem/tree/datum_test.go, line 115 at r6 (raw file):

			`'00:00:00'`, `'23:59:59.999999'`},
		{`'24:00:00.000000':::time`, `'23:59:59.999999'`, `'00:00:00.000001'`,
			`'00:00:00'`, `'23:59:59.999999'`},

Nit: I think it would make more sense to put these tests in TestParseDTime rather than TestDatumOrdering. Or maybe keep the first one here to verify the ordering and put the rest in the parse test.

[solongordon]

Oh, also I would suggest squashing your six commits into one before merging.

Reviewable status: complete! 1 of 0 LGTMs obtained (waiting on @rohany)

[rohany]

bors r+

[craig]

Build succeeded

• GitHub CI (Cockroach)