Biopython BioSQL error : AttributeError: 'DBSeq' object has no attribute '_data'

Question

Hi I am playing with Biopython and BioSQL,

I manage to create an MySQL database using Managing local biological databases with the BioSQL module tutorial:

and load 2909 fasta sequences in it through:

db = server["test_1"]

count = db.load(SeqIO.parse("Newtextfile-03.txt", "fasta"))
        print("Loaded %i records" % count)
        server.commit()

from this file: Newtextfile-03.txt

problem is trying just to print the database index list with:

for i in db:
    print(i)

works , while trying print element by element:

for i in db:
    print(i, db[i])

or with:

for key, record in db.items():
            print ("Key %r maps to a sequence record  %s" % (key, record))

gives an error:

2557 Traceback (most recent call last):
  File "./try_001.py", line 142, in <module>
    f()
  File "./try_001.py", line 120, in f
    print(i, db[i])
  File "......../lib/python3.8/site-packages/Bio/SeqRecord.py", line 652, in __str__
    lines.append(repr(self.seq))
  File "........//lib/python3.8/site-packages/Bio/Seq.py", line 109, in __repr__
    return f"{self.__class__.__name__}({self._data!r})"

AttributeError: 'DBSeq' object has no attribute '_data'

the error occurs for three records namely : [2164, 2522, 2557]

inspecting the Newtextfile-03.txt file results in finding fasta sequences of less than 100 aminoacids.

Is there any way to resolve the error ? It sounds strange to me that I can load the records into the database but not retrieving them. Is anything I am missing about the record load ? or similar ?

I am working with the database using:

 server = BioSeqDatabase.open_database(
    driver='MySQLdb' , 
    user=password.userz,
    password=password.pazzword,
    host="XXXXXXXX",
    db="bio-data-first")

db = server.new_database("test_1", description="Just for testing")
        server.commit()

using driver="MySQLdb", wasn't able to have driver='mysql.connector' to work,

biopython is version 1.78

Answer 1

It's a bug in Biopython 1.78.

What those three sequences have in common is that they're less than 61 characters long. When you print the record it implicitly calls __str__ for the record, which calls __repr__ for the Seq. DBSeq gets its method from Seq, which looks like:

    def __repr__(self):
        """Return (truncated) representation of the sequence for debugging."""
        if len(self) > 60:
            # Shows the last three letters as it is often useful to see if
            # there is a stop codon at the end of a sequence.
            # Note total length is 54+3+3=60
            return f"{self.__class__.__name__}('{str(self[:54])}...{str(self[-3:])}')"
        else:
            return f"{self.__class__.__name__}({self._data!r})"

...and then it only tries to access the nonexistent _data for the short sequences.

So there's a mismatch between DBSeq's lack of a _data attribute and its use of Seq.__repr__. I don't see a DBSeq class at all in the latest Biopython, so maybe this is fixed already. But there's no tag for 1.79 yet that I can see and I can't get the very latest code to import successfully so I'm not sure. In the meantime if you avoid letting repr(record) happen you should be able to work around it.