Suggest: show filename, number of loci and number of individuals/samples in `read_bed` error message

[richelbilderbeek]

Dear genio maintainers,

Thanks for genio and its many tests!

The .bed format is not readable by humans. To read a .bed file, one needs to specify the number of loci and the number of individuals. However, when reading a .bed file, these numbers may have been incorrect. In that case, genio gives a clear error message:

<Rcpp::exception/C++Error/error/condition>
Error in `read_bed_cpp(file, m_loci, n_ind)`: Input BED file continued after all requested rows were read!  Either the specified the number of loci was too low or the input file is corrupt!

My suggestion is to add the .bed filename, number of loci and number of individuals in the error message:

<Rcpp::exception/C++Error/error/condition>
Error in `read_bed_cpp(file, m_loci, n_ind)`: Input BED file continued after all requested rows were read!  Either the specified the number of loci was too low or the input file is corrupt!
.bed filename: ~/my_file.bed
Number of loci: 42
Number of individuals: 314

This would definitely help my debugging sessions.

I would volunteer to add this, would GitHub Actions be active. I do volunteer to add GitHub Actions :-)

[alexviiia]

Sorry I'm behind the times regarding GitHub actions, that sounds nice!

Regarding the numbers you're asking, I can report the numbers that were part of the input, but you already know those, so is that just for ease? And also, if you don't already know the true dimensions (from the accompanying BIM and FAM tables), I wouldn't guess them because there are usually too many solutions. The true dimensions are unknown just from an isolated BED file, not just because a file with 100 entries could be 10*10 or 5*20, for example, but also because of byte padding (for example, data with fixed number of loci and either 5, 6, 7, or 8 individuals all have the same output file sizes because each byte holds 4 individuals). Maybe you knew this already but just wanted to make sure we're on the same page.

[richelbilderbeek]

Hi @alexviiia

Regarding the numbers you're asking, I can report the numbers that were part of the input, but you already know those, so is that just for ease?

Yes, for read_bed you are completely right. However, for read_plink, those numbers are determined internally. In my use case, I use read_plink on a daily basis. As read_bed indeed knows its input arguments, it would be easy to add those to the error message. Alternatively, read_plink could catch (tryCatch) the error from read_bed, then add extra info. Either option I would enjoy!

Thanks for sharing the hassle of reading .bed files. Luckily I can use the genio R package :-)