io.rdrecord() fails when sampto is 'end'

[atpage]

Calling rdsamp() or rdrecord() with sampto='end' replaces sampto with sig_len from the header, which may be None, causing check_read_inputs() to fail. I have patched this locally in the case where sig_len is None by replacing sig_len with filesize*8/bits_per_sample/n_sig, where bits_per_sample is looked up in this dict keyed on record.fmt:

bits_per_sample = {
    '8':    8,
    '16':  16,
    '24':  24,
    '32':  32,
    '61':  16,
    '80':   8,
    '160': 16,
    '212': 12,
    '310': 32.0/3,
    '311': 32.0/3,
}

[atpage]

Also, I saw the recently closed issues related to this, but I am already on 2.1.1.

[cx1111]

Do you have an example of a record without a signal length?

[atpage]

I don't know if there are any examples on Physionet, but it is not a required field in the header. [1]

I do have files (which I cannot share) that have been exported to MIT format by an ECG device manufacturer, and the first line of the header is

<name> <channels> <sample rate>

i.e. signal length is omitted. That's how I encountered the problem.

[1] https://www.physionet.org/physiotools/wag/header-5.htm (number of samples per signal)