False duplicate key error

[Benjamin-Lee]

I was using some relatively simple code I wrote to generate FASTA files containing random sequences:

import random
import sys

number = int(sys.argv[1])
length = int(sys.argv[2])

for i in range(number):
    seq = ""
    for j in range(length):
        seq += random.choice("ATGC")
    with open(f"{number}-{length}bp-random-seqs.fa", "a+") as f:
        print(f"> seq {i}", file=f)
        print(seq, file=f)

The file ends up looking like this:

> seq 0
...
.
.
.
> seq n
...

However, I am getting the following error:

  File "/Users/BenjaminLee/.virtualenvs/squiggle/lib/python3.6/site-packages/pyfaidx/__init__.py", line 481, in read_fai
    raise ValueError('Duplicate key "%s"' % key)
ValueError: Duplicate key "seq"

It seems that it's only picking up on the seq in the description line, not on the identifier.

[mdshw5]

This is expected behavior, which was chosen to mimic the samtools faidx behavior of splitting deflines on whitespace. You can specify that the .fai index should not be used as a key name by passing the read_long_names argument:

>>> genes = Fasta('10-1000bp-random-seqs.fa', read_long_names=True)
>>> genes
Fasta("10-1000bp-random-seqs.fa")
>>> genes.keys()
odict_keys([' seq 0', ' seq 1', ' seq 2', ' seq 3', ' seq 4', ' seq 5', ' seq 6', ' seq 7', ' seq 8', ' seq 9'])

Since duplicate keys are detected during index reading in pyfaidx, the .fai will appear to contain duplicate sequences (which may be a problem for other tools) but will not contain whitespace, and so I think is "more correct" with respect to samtools behavior. See #111 for more information about how I arrived at this behavior.

The documentation should definitely be clearer about this, as well as other Fasta and Faidx arguments, so feel free to submit a PR if you want to add something to the README.