zero-based indexing is inconsistently applied in Fasta

[ihaque]

Setting one_based_attributes=False on a Fasta object does not propagate indexing operations properly down to the underlying Faidx or Sequence objects.

Given the following test data file:

>test_chrom
ACGTACGTACGTACGTACGTACGT

and using pyfaidx @ 6e54c9b with the following preamble:

>>> import pyfaidx
>>> f0 = pyfaidx.Fasta('tests/data/test.fa', one_based_attributes=False)
>>> f1 = pyfaidx.Fasta('tests/data/test.fa', one_based_attributes=True)

I get the following issues:

1. Sequences have inconsistent names depending on whether zero-based and one-based coordinates are used:

>>> f0['test_chrom'][:]     
>test_chrom
ACGTACGTACGTACGTACGTACGT
>>> f1['test_chrom'][:]
>test_chrom:1-24
ACGTACGTACGTACGTACGTACGT

1. Using the same coordinates between "zero-based" and "one-based" objects produces the exact same results, rather than the single-base shift expected:

>>> f1['test_chrom'][:][1:5]
>test_chrom:2-5
CGTA
>>> f0['test_chrom'][:][1:5]
>test_chrom
CGTA

[mdshw5]

Thanks so much for filing a reproducible issue! I have to explain that you have discovered one bug and encountered one confusing feature.

First, the Sequence.longname method should definitely be updated to behave as you expect.

The second issue you raise is actually expected behavior. The one_based_attributes argument only applies to the "friendly" coordinate representation but should have no effect on the slice notation. I should update the documentation to make this explicit. I believe that having a switch for the slicing notation would be quite confusing and allow subtle bugs to go unnoticed.

[mdshw5]

• Fix Sequence.longname behavior with one_based_attributes
• Add test case