read_bedlike about pyranges HOT 2 OPEN

Thanks for developing this excellent resource.
On this point, I'm happy with the current solution posted on readthedocs: pr.PyRanges(pandas.read_table(f)) , however, in terms of being 'too strict', using this appreach did give me the following error message:

The dataframe does not have all the columns Chromosome, Start and End

Which surprised me, until I realized that the columns were "chrom", "start", and "end" (lower-case), and didn't match. The pr.read_bed function was much more forgiving for variations on these names.

Of course I can manually change these to match the needed "Chromosome", "Start"/"End"(capitalized), but "chrom/start/end" are the standard output from bedtools functions like unionbedg, etc.

With the current solution based on pr.PyRanges(pandas.read_table(f)) is there a way to accommodate column labels like this? In the long run, it would be ideal if the read_bedlike (or whatever) could allow for these column labels so that it could work smoothly together with bedtools (which is pretty important for anyone working with bed-like files).

Thanks again for this project, it looks really promising.

from pyranges.

DataFrames can only have one name per column. Therefore I cannot allow "chrom" and "Chromosome" to refer to the same column.

If I allowed pyranges to have aliases for column names (start and Start) code that used the dataframes underlying the pyranges would break since no aliasing-code is implemented in pandas.

One possibility is to write a function to read bedtools-like data:

def read_bedtools(f):
    df = pd.read_table(f)
    df = df.rename(columns={"chrom": "Chromosome", "start": "Start", "end": "End", "strand": "Strand"})
    return pr.PyRanges(df)

# the opposite for write_bedtools

Thanks for the question and the kind words btw.

from pyranges.