Neo4j time tests

neo4j_datetime_tests.txt

I started with the following node creaton loop

for i in range(2):
    tx.run(f"MERGE (:Node {{date: datetime('2018-06-01T{i:02}:00:00Z')}})")
    tx.run(f"MERGE (:Node {{date: datetime('2018-06-01T{i:02}:00:00-01:00')}})")
    
Here is a TZ reference table
| UTC | -01:00 |
|-----|--------|
|   0 |     23 |
|   1 |      0 |
|   2 |      1 |
|   3 |      2 |
|-----|--------|

Partial strings don't match
n.date = '2018'
[]
n.date > '2018'
[]
n.date < '2018'
[]

Full strings don't match either
n.date < '2019-01-01T00:00:00'
[]
n.date < '2019-01-01T00:00:00Z'
[]

TZ unaware datetimes seem to be casted to UTC
When using a < > operator, it takes both TZ
n.date < datetime('2019-01-01T00:00:00')
[[neotime.DateTime(2018, 6, 1, 0, 0, 0.0, tzinfo=<UTC>)], [neotime.DateTime(2018, 6, 1, 0, 0, 0.0, tzinfo=pytz.FixedOffset(-60))], [neotime.DateTime(2018, 6, 1, 1, 0, 0.0, tzinfo=<UTC>)], [neotime.DateTime(2018, 6, 1, 1, 0, 0.0, tzinfo=pytz.FixedOffset(-60))]]
n.date < datetime('2019-01-01T00:00:00Z')
[[neotime.DateTime(2018, 6, 1, 0, 0, 0.0, tzinfo=<UTC>)], [neotime.DateTime(2018, 6, 1, 0, 0, 0.0, tzinfo=pytz.FixedOffset(-60))], [neotime.DateTime(2018, 6, 1, 1, 0, 0.0, tzinfo=<UTC>)], [neotime.DateTime(2018, 6, 1, 1, 0, 0.0, tzinfo=pytz.FixedOffset(-60))]]

When using an = operator, it matches on datetime and TZ
n.date = datetime('2018-06-01T00:00:00')
[[neotime.DateTime(2018, 6, 1, 0, 0, 0.0, tzinfo=<UTC>)]]
n.date = datetime('2018-06-01T01:00:00')
[[neotime.DateTime(2018, 6, 1, 1, 0, 0.0, tzinfo=<UTC>)]]
n.date = datetime('2018-06-01T02:00:00')
[]
n.date = datetime('2018-06-01T00:00:00Z')
[[neotime.DateTime(2018, 6, 1, 0, 0, 0.0, tzinfo=<UTC>)]]
n.date = datetime('2018-06-01T01:00:00Z')
[[neotime.DateTime(2018, 6, 1, 1, 0, 0.0, tzinfo=<UTC>)]]
n.date = datetime('2018-06-01T02:00:00Z')
[]

So TZ equivalence doesn't work in equality
n.date = datetime('2018-05-31T23:00:00-01:00')
[]
n.date = datetime('2018-06-01T00:00:00Z')
[[neotime.DateTime(2018, 6, 1, 0, 0, 0.0, tzinfo=<UTC>)]]

If we use >= 1 second before, we have both TZ
n.date >= datetime('2018-06-01T00:59:59Z')
[[neotime.DateTime(2018, 6, 1, 0, 0, 0.0, tzinfo=pytz.FixedOffset(-60))], [neotime.DateTime(2018, 6, 1, 1, 0, 0.0, tzinfo=<UTC>)], [neotime.DateTime(2018, 6, 1, 1, 0, 0.0, tzinfo=pytz.FixedOffset(-60))]]

If we use >= right on time, we lose one valid -01:00 entry
n.date >= datetime('2018-06-01T01:00:00Z')
[[neotime.DateTime(2018, 6, 1, 1, 0, 0.0, tzinfo=<UTC>)], [neotime.DateTime(2018, 6, 1, 1, 0, 0.0, tzinfo=pytz.FixedOffset(-60))]]

If we use a combination of >= and <= with the 1 second off trick, we get both
n.date >= datetime('2018-06-01T00:59:59Z') AND n.date <= datetime('2018-06-01T01:00:00Z')
[[neotime.DateTime(2018, 6, 1, 0, 0, 0.0, tzinfo=pytz.FixedOffset(-60))], [neotime.DateTime(2018, 6, 1, 1, 0, 0.0, tzinfo=<UTC>)]]

On the other hand
RETURN datetime('2018-06-01T01:00:00Z').epochMillis = datetime('2018-06-01T00:00:00-01:00').epochMillis
[[True]]

So maybe that when we query and care about a time point/range but don't care about the TZ, we ought to use epochMillis.