Interval.__setitem__?

[daler]

Interval.__setitem__?

[brentp]

self._bed.fields[key] = string(value)

then the key still has to be an integer. for the key as a string maybe (untested)

if isinstance(key, basestring):
    setattr(self, key)

[daler]

OK, trying to setitem sort of works:

    def __setitem__(self, object key, object value):
        if isinstance(key, (int,long)):
            nfields = self._bed.fields.size()
            if key >= nfields:
                raise IndexError('field index out of range')
            elif key < 0: key = nfields + key
            self._bed.fields[key] = string(value)
        elif isinstance(key, (basestring)):
            setattr(self, key, value)

it looks like the chrom is getting changed (verified by printing the fields) but this is not recognized when i access Interval.chrom . . . see the test at:
https://github.com/daler/pybedtools/blob/exp/pybedtools/test/test_cbedtools.py#L117

the failed test shows this:

FAIL: testSetItem (pybedtools.test.test_cbedtools.IntervalTest)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/ryan/proj/pybedtools/pybedtools/test/test_cbedtools.py", line 123, in testSetItem
    self.assertEqual(iv['chrom'], 'fake')
AssertionError: 'chr21' != 'fake'
-------------------- >> begin captured stdout << ---------------------
['fake', '9719768', '9721892', 'ALR/Alpha', '1004', '+']

--------------------- >> end captured stdout << ----------------------

my guess is that a similar issue causes testAppend to fail as well:
https://github.com/daler/pybedtools/blob/exp/pybedtools/test/test_cbedtools.py#L126

[brentp]

ah. it's because sometimes it's setting/getting the .chrom property in c++ (with the string) and others it's setting/getting the field (with the integer key).
i'll think about how to fix.

[daler]

i ran into all sorts of issues trying to get this to work 1) with GFF files and 2) BED files that get something appended to them (and therefore getting nonempty otherFields vector).

i made some changes to getitem, which pays attention to the current length of fields and otherFields.

ba0ce3b

However, it looks like a GFF Interval, when created, gets 3 things put in its otherFields. I tried to account for this in the new getitem, but this seems like the wrong way to do it.

Can you explain how fields and otherFields should work for GFF files and for BED files of various lengths (say, BED3 vs BED9)?

[daler]

OK, found the answer to my question in bedtool.h -- GFF parser puts "source", "fname", and "group" in the otherFields vector, while the rest of the GFF line goes into the standard BED6 fields. BED7+ has 7th and beyond placed into otherFields as expected.

[brentp]

i am also working on this a bit. trying to make it much simpler in the c and cython parts.

[brentp]

we'll have to check that all the set and normal property setters also update fields.
(note to self) make sure the field is in LOOKUPS and follow the set for chrom/ or start/end.

[brentp]

the .name setting is not working for GFF/GTF. needs something like:

self._bed.fields[8] = string(";".join("%s=%s" % (k, v) for k,v in attrs.iteritems()))

but also account for GTF

[brentp]

this is fixed as of your: d391fae yes?