Document go time nanosecond precision on linux x86_64

[rdlaitila]

This caused a bit of confusion when tests started failing when moving unit tests into a linux container from my mac workstation. Not the fault of edgedb-go but may be a good edge case to document for others.

It appears time.Now() on certain platforms can generate more than 6 decimal places of nanosecond precision. ie: 2022-04-18 22:13:26.9245872 +0000 UTC on linux x86_64 as one example

However edgedb only stores up to 6 digits of nanosecond precision (postgres thing?)

edgedb> select <datetime>'2022-04-18T17:54:23.1876103-04:00';
{<datetime>'2022-04-18T21:54:23.187610Z'}

An example test that always passes on my intel mac workstation, but will intermittently fail inside a linux container:

func Test_time_precision(t *testing.T) {
	now := time.Now().UTC()
	t.Logf("now: %s", now)
	var result time.Time
	query := "select <datetime>$now"
	args := map[string]any{"now": now}
	err := TestClient.QuerySingle(TestContext, query, &result, args)
	assert.NoError(t, err)
	t.Logf("result: %s", result)
	t.Logf("same: %v", now.Equal(result))
}

Consistent pass on mac:

=== RUN   Test_time_precision
    test.go:168: now: 2022-04-18 22:21:02.126656 +0000 UTC
    test.go:174: result: 2022-04-18 22:21:02.126656 +0000 UTC
    est.go:175: same: true

Intermittent failure in linux container:

=== RUN   Test_time_precision
    test.go:168: now: 2022-04-18 22:13:26.9245872 +0000 UTC
    test.go:174: result: 2022-04-18 22:13:26.924587 +0000 UTC
    test.go:175: same: false

one naive option is to round both values before comparing:

assert.Equal(t, now.Round(1000*time.Nanosecond), result.Round(1000*time.Nanosecond))

[fmoor]

Thanks for pointing this out! I wonder if this should go in the docs for the datetime type and other cal::* types in the edgedb/edgedb repo :thinking:

cc @1st1 @colinhacks