[Biopython] Translation of partial codons

[Peter Cock]

On Thu, Mar 21, 2013 at 5:56 PM, Fields, Christopher J
<cjfields at illinois.edu> wrote:
> On Mar 21, 2013, at 12:19 PM, Peter Cock <p.j.a.cock at googlemail.com>
>  wrote:
>
>> On Thu, Mar 21, 2013 at 5:10 PM, Iddo Friedberg <idoerg at gmail.com> wrote:
>>> Suggstions so far:
>>> 1. Raise an exception. This may cause  code running on existing data to
>>> change behavior. I.e. it ran before well on bad length sequences, but as of
>>> the new code installtion, things will break.
>>
>> Yes, but in most cases this will be a good thing. The minority of people
>> knowingly dealing with partial sequences can make this explicit by first
>> ensuring their sequence is a multiple of three in length (by padding or
>> cropping as most appropriate to their use case).
>>
>>> 2. Add a default length_check=True to the translate method. Again, this may
>>> cause exiting code to behave differently wiht the same data once user
>>> upgrades. Unless the user explicitly changes the call to
>>> myseq.translate(length_check=False)
>>
>> A sensible approach to making likely errors explicit, with an easy work-around
>> for the old implicit truncation. The downside is yet another argument to the
>> translate functions/methods, which are already pretty complicated. I prefer (1).
>>
>>> 3. My suggestion: use length_check=False as default. Code behaves the same
>>> as before, so no data-induced breakages. If the user wants to check length,
>>> the explicitly pass a True value. So we give the option of checking length,
>>> and retaining code-behavior legacy.
>>>
>>> length_check, being an argument, does not need to be passed explicitly.
>>
>> I don't like this, even though it is backwards compatible for the corner
>> case. I think the old behaviour is a bug.
>>
>> Regards,
>>
>> Peter
>
> That's basically the approach we took on the bioperl end, e.g. the
> old behavior was an unintended bug (it was a little more complex
> than that in reality, but in essence it boils down to that).  It was
> too magic, and the old behavior can be regained with a parameter
> setting.  I don't think we throw an exception, but maybe we should...
>
> Anyway, I would think going with something that would be following
> the tenant of least surprise would be very python-esque :)
>
> chris

I started work on making this an exception, and from our test suite
realised that simple ORF finding is an example where this change
is likely to be noticed. I have therefore for now just added a new
warning if translating partial codons, which can be upgraded to a
full exception in future (or removed depending on how people react).

https://github.com/biopython/biopython/commit/c0112a7b79a61eabe0adea78bb70d572f1950cde

Peter