stats: Ensure that events are aggregated only once

[psaiz]

❤️ Thank you for your contribution!

Description

Ensures that a second aggregation does not reprocess any aggregation already performed and that has not changed,

It addresses https://github.com/zenodo/rdm-project/issues/523

• closes https://github.com/zenodo/rdm-project/issues/523

[ntarocco]

@psaiz this is not now a string, but down there it is compared like a number/date with <. Are you sure that this is correct?
Are the tests catching this?

[psaiz]

Thanks for looking into this. It is a string, isn't it? Note that the comparison is with strings.

I've added these prints to make sure:

                 ).replace(**dict.fromkeys(INTERVAL_ROUNDING[self.interval], 0))
+                print("Comparing")
+                print(aggregation["last_update"]["value_as_string"])
+                print(previous)
                 if aggregation["last_update"]["value_as_string"] and previous:
                     if aggregation["last_update"]["value_as_string"] < previous:
+                        print("SKIPPING EVENT")
                         continue

And the test seems to be doing it properly:

Comparing
2017-06-02T11:00:00.000Z
None
Comparing
2017-06-02T11:00:00.000Z
None
Comparing
2017-07-01T05:00:00.000Z
2017-06-02T11:59:00
Comparing
2017-07-01T05:00:00.000Z
2017-06-02T11:59:00
Comparing
2017-07-01T05:00:00.000Z
2017-07-01T05:59:00
SKIPPING EVENT

[ntarocco]

Sorry, I wanted to write it is now a string, but I wrote it is not a string, I am sorry :)
Do we really want to compare strings in this way?
https://docs.python.org/3/tutorial/datastructures.html#comparing-sequences-and-other-types

I would feel much safer if we convert everything first with datetime.fromisoformat(), and then we use the Python date comparison. WDYT?

[psaiz]

Ok, I have no strong opinions about comparing strings or dates. The added value of comparing the dates is that there is no need to modify the incoming object.

Do you prefer this approach?