Nearest range missing from the return of nearest_ranges when ranges are touching.

[brechtmann]

Thank you for this package!

However, I noticed some behavior that seems unexpected. In the example below, I initialize a PyRanges object with three ranges, and a second one with a single range. Since the range in the second object touches the middle range from the first object, I would expect the range [6-7) to be among the nearest ranges.

import pyranges as pr
f1 = pr.PyRanges(dict(Chromosome="chr1", Start=[1, 6, 20], End=[2, 7, 22], Strand=["+", "+", "+"]))
f1

Chromosome	Start	End	Strand
chr1	1	2	+
chr1	6	7	+
chr1	20	22	+

f2 = pr.PyRanges(dict(Chromosome="chr1", Start=[5], End=[6], Strand=["+"]))

Chromosome	Start	End	Strand
chr1	5	6	+

f1.nearest_ranges(f2)

Chromosome	Start	End	Strand	Chromosome_b	Start_b	End_b	Strand_b	Distance
chr1	1	2	+	chr1	5	6	+	4
chr1	20	22	+	chr1	5	6	+	15

When swapping the ranges the closest range is reported.

f2.nearest_ranges(f1)

Chromosome	Start	End	Strand	Chromosome_b	Start_b	End_b	Strand_b	Distance
chr1	5	6	+	chr1	6	7	+	1

pr.__version__

'1.1.4'

[endrebak]

Thanks for the report :) Will look into it.

[endrebak]

I've been ill but I'll look at this this weekend :)

[endrebak]

I thank you for reporting this :) There was a missing equals sign it seems. I'll push a fix.

[endrebak]

I have updated ruranges to fix this so if you install ruranges 0.0.14 it should work. I need to fix #110 to be able to update pyranges itself to use the library by default but I hope to be able to do so today

(Edit: ruranges is the underlying rust library that implements the nearest algorithm that pyranges uses)

[endrebak]

Should be fixed in latest :)

[endrebak]

Feel free to reopen if it does not work for you

[brechtmann]

Thanks for fixing this! I just confirmed that it works.

[brechtmann]

@endrebak I just installed pyranges1 1.3.2 (ruranges-0.1.1) and encountered the error from above again.

[brechtmann]

I made a simpler test case that you could add to your tests:

import pandas as pd
#import pyranges as pr
import pyranges1 as pr

def test_nearest_range_when_ranges_are_touching() -> None:
    r1 = pr.PyRanges(
        {
            "Chromosome": "chr1", 
            "Start": [1], 
            "End": [5], 
            "Strand": ["+"]
        }
    )
    r2 = pr.PyRanges(
        {
            "Chromosome": "chr1",
            "Start": [5],
            "End": [10],
            "Strand": ["+"]
        }
    )
    result = r2.nearest_ranges(r1)    

    _expected_result = {
        "Chromosome": {0: "chr1"},
        "Start": {0: 5},
        "End": {0: 10},
        "Strand": {0: "+"},
        "Chromosome_b": {0: "chr1"},
        "Start_b": {0: 1},
        "End_b": {0: 5},
        "Strand_b": {0: "+"},
        "Distance": {0: 1},
    }

    expected_result = pd.DataFrame(_expected_result)
    pd.testing.assert_frame_equal(result, expected_result)

[endrebak]

I will prioritize fixing this

[endrebak]

I will add a few tests in rust and ensure they are run as part of the precommit.

[endrebak]

I should add your test as a doctest to the pyranges code also :)

[endrebak]

Not commited yet: #137