Bug: endaq.ide.get_doc() returns data outside requested start/end when the window is out-of-bounds

[XanthanGum]

Summary
When calling endaq.ide.get_doc() with timezone-aware datetime.datetime values for start and end that do not overlap the recording interval of the IDE file, the function still returns a document with non-empty channels and data outside the requested window.
Expected behavior is an object where each channel contains 0 samples (i.e., the request yields empty data when the requested window is outside the recording interval).

Steps to Reproduce
Parse a filename to obtain timezone-aware start/end datetimes (AWST, UTC+08):

from datetime import datetime
from zoneinfo import ZoneInfo
import endaq

filename = "20251018_RC63_LH_SAM4_BATCH_1_202510180930_202510181830"
tz = ZoneInfo("Australia/Perth")

start_time = datetime.strptime(filename.split("_")[-2], "%Y%m%d%H%M").replace(tzinfo=tz)
end_time   = datetime.strptime(filename.split("_")[-1], "%Y%m%d%H%M").replace(tzinfo=tz)

print(start_time)  # 2025-10-18 09:30:00+08:00
print(end_time)    # 2025-10-18 18:30:00+08:00

Request a trimmed document using start/end that do not overlap the IDE’s actual data (the file below appears to contain data on Oct 19 UTC, not Oct 18 AWST):

doc = endaq.ide.get_doc(
    r"C:\...\20251018_RC63_LH_SAM4_BATCH_2_202510180930_202510181830.IDE",
    start=start_time,
    end=end_time,
)

Inspect the channel table:

endaq.ide.get_channel_table(doc)

Output includes non-empty channels and a warning:

...\endaq\ide\info.py:201: RuntimeWarning: divide by zero encountered in scalar divide
  rate = samples / (duration / 10 ** 6)

Example (excerpt):

channel	name	type	units	start	end	duration	samples	rate
8.0	X (2000g)	Acceleration	g	00:00.0008	00:00.0552	00:00.0543	2720	5000.78 Hz
8.1	Y (2000g)	Acceleration	g	00:00.0008	00:00.0552	00:00.0543	2720	5000.78 Hz
8.2	Z (2000g)	Acceleration	g	00:00.0008	00:00.0552	00:00.0543	2720	5000.78 Hz
80.0	X (40g)	Acceleration	g	00:00.0015	00:00.0506	00:00.0491	248	504.87 Hz
80.1	Y (40g)	Acceleration	g	00:00.0015	00:00.0506	00:00.0491	248	504.87 Hz
80.2	Z (40g)	Acceleration	g	00:00.0015	00:00.0506	00:00.0491	248	504.87 Hz
20.0	Internal Pressure	Pressure	Pa	00:00.0039	00:06.0112	00:06.0073	62	10.21 Hz
20.1	Internal Temperature	Temperature	°C	00:00.0039	00:06.0112	00:06.0073	62	10.21 Hz
65.0	X	Quaternion	q	00:00.0111	00:01.0122	00:01.0010	102	100.94 Hz
65.1	Y	Quaternion	q	00:00.0111	00:01.0122	00:01.0010	102	100.94 Hz
65.2	Z	Quaternion	q	00:00.0111	00:01.0122	00:01.0010	102	100.94 Hz
65.3	W	Quaternion	q	00:00.0111	00:01.0122	00:01.0010	102	100.94 Hz
65.4	Acc	Quaternion	q	00:00.0111	00:01.0122	00:01.0010	102	100.94 Hz
88.0	Latitude	Location	Degrees	01:33:55.0979	01:34:27.0322	00:31.0342	32	1.02 Hz
88.1	Longitude	Location	Degrees	01:33:55.0979	01:34:27.0322	00:31.0342	32	1.02 Hz
88.2	Time	Unix Epoch	s	01:33:55.0979	01:34:27.0322	00:31.0342	32	1.02 Hz
88.3	Ground Speed	GNSS Speed	m/s	01:33:55.0979	01:34:27.0322	00:31.0342	32	1.02 Hz
102.0	GNSS Time:00	Unix Epoch Reference	s	01:33:59.0322	01:33:59.0322	00:00.0000	1	inf Hz

Retrieve primary sensor data and check timestamps:

df = endaq.ide.get_primary_sensor_data(doc=doc)
print(df.head())
print(len(df))

Example timestamps are UTC on Oct 19, not within the requested Oct 18 AWST window:

                                    X (2000g)  Y (2000g)  Z (2000g)
timestamp
2025-10-19 09:57:15.008941+00:00      3.853685  -1.895579  -1.323955
2025-10-19 09:57:15.009141042+00:00   3.853685  -1.998095  -0.949977
...
2720

Manually filter with the same start_time/end_time:

df_trimmed = df.loc[(df.index > start_time) & (df.index < end_time)]
print(df_trimmed.head())
print(len(df_trimmed))

This yields an empty DataFrame (as expected when the window is out-of-bounds):

Empty DataFrame
Columns: [X (2000g), Y (2000g), Z (2000g)]
Index: []
0

Expected Behavior
If start/end are outside the recording interval of the IDE file, endaq.ide.get_doc() should return a document whose channels each have 0 samples and no data, reflecting the requested empty interval.

Actual Behavior
endaq.ide.get_doc() returns a document containing non-empty channels and data outside the requested start/end window when the window does not overlap the recording interval.
The resulting get_primary_sensor_data(doc) returns data with timestamps outside the requested range.
A RuntimeWarning: divide by zero encountered in scalar divide appears in endaq\ide\info.py:201 (likely due to a channel with zero duration and one sample resulting in inf Hz), which may be related to edge handling but is not the primary issue.

[StokesMIDE]

@XanthanGum Good catch. Thanks for the detailed report. I've coped this to our internal issue/task tracking system. I have some urgent work to complete, but I'll investigate it soon.