Correct sampto behaviour using rdsamp and rdann?

[kevindlau]

When reading the signal with rdsamp and the corresponding annotations with rdann using the same sampto value results in an IndexError when accessing the signal array it includes an annotation outside the array.

In the example below, to read the first 100 entries, sampto must be 100. If using the same value for the annotations up to and including 100 are read, rather than stopping at 99 (i.e., the 100th entry in a 0-indexed array).

Is this the expected behaviour?

Using: python 3.9.18 with wfdb 4.1.2

import numpy as np
import wfdb

# create test data
n = 250
test_sig = np.zeros((n, 1))
test_sig[:, 0] = np.linspace(0, 1, n)
print(f'original shape: {test_sig.shape}')

# write to file
wfdb.wrsamp('test', fs = 250, units=['mV'], sig_name=['test'], p_signal=test_sig, fmt=['16'])

# create annotation and write
wfdb.Annotation(
    record_name='test', 
    extension='atr',
    sample=np.array([0, 99, 100, 249]), 
    symbol=['N', 'N', 'N', 'N']
).wrann()

samp_from = 0
samp_to = 100

# read values
values, _ = wfdb.rdsamp('test', sampfrom=samp_from, sampto=samp_to)
print(f'sampto shape: {values.shape}')


atr_1 = wfdb.rdann('test', extension='atr', sampfrom=samp_from, sampto=samp_to, shift_samps=True)
print(f'samples: {atr_1.sample}')
# IndexError
# print(f'values at samples: {values[atr_1.sample, :]}')

samp_to_minus_1 = samp_to - 1
atr_2 = wfdb.rdann('test', extension='atr', sampfrom=samp_from, sampto=samp_to_minus_1, shift_samps=True)
print(f'samples: {atr_2.sample}')
# No IndexError
# print(f'values at samples: {values[atr_2.sample, :]}')

[kevindlau]

Should the 2nd condition be < rather than <=?

wfdb-python/wfdb/io/annotation.py

Line 242 in 8d81359

np.where(self.sample >= sampfrom), np.where(self.sample <= sampto)

[bemoody]

I agree, I think rdann should be consistent with rdsamp and I think sampfrom/sampto should define a half-open interval.

This would be an API-breaking change, though. I've kind of wanted to change that API in any case, perhaps renaming the args to from/to or start/end or something, so maybe we should do that at the same time.