Sort before interpolating

[zihaoxu98]

In np.interp it is presumed that xs is sorted. This PR ensures that the xs is sorted before apply interpolation.

This solves the randomness issue in XENONnT/fuse#230. Here is the explanation:

• If in straxen.URLconfig, '&sort=xxxx' is used, then in the MongoDB interface of rframe, it sends $sort: xxxx to MongoDB. However, if xxxx is not sufficient enough to determine the order due to duplicated values, the output of MongoDB will be random, meaning that xxxx is sorted but within each value of xxxx it's random. See the sort consistency section in the document.
• This randomness can finally propagate to the interpolation. For example, if we require '&sort=xxxx', the time can be randomly unsorted. But this conflicts with the presumption of np.interp, which finally causes randomness of the interpolated output.

---

An MWE is

import xedocs
import datetime

accessor = xedocs.databases.straxen_db()['pmt_area_to_pes']
datasource = accessor.schema.default_datasource()
labels = {'detector': 'tpc', 'run_id': '046000', 'version': 'v10'}
query = accessor.schema.compile_query(accessor.storage, **labels)
index = query.index.indexes[2]

x = [{'version': 'v10',
      'created_date': datetime.datetime(2023, 10, 24, 15, 40, 47, 945000),
      'comments': '',
      'time': datetime.datetime(2022, 7, 8, 8, 46, 56),
      'detector': 'tpc',
      'pmt': 0,
      'value': 0.0079474222},
     {'version': 'v10',
      'created_date': datetime.datetime(2023, 10, 24, 15, 40, 47, 945000),
      'comments': '',
      'time': datetime.datetime(2022, 7, 19, 15, 48, 16),
      'detector': 'tpc',
      'pmt': 0,
      'value': 0.0079188448}]

y = [{'version': 'v10',
      'created_date': datetime.datetime(2023, 10, 24, 15, 40, 47, 945000),
      'comments': '',
      'time': datetime.datetime(2022, 7, 19, 15, 48, 16),
      'detector': 'tpc',
      'pmt': 0,
      'value': 0.0079188448},
     {'version': 'v10',
      'created_date': datetime.datetime(2023, 10, 24, 15, 40, 47, 945000),
      'comments': '',
      'time': datetime.datetime(2022, 7, 8, 8, 46, 56),
      'detector': 'tpc',
      'pmt': 0,
      'value': 0.0079474222}]

label = datetime.datetime(2022, 7, 16, 7, 18, 14, 927000)

print(index.reduce(x, label)[0]['value'])

print(index.reduce(y, label)[0]['value'])

Before the PR it outputs:

0.007927332994063166
0.0079474222

After the PR it outputs the same values:

0.007927332994063166
0.007927332994063166

[jmosbacher]

Nice catch!