-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/9413/#review16512
-----------------------------------------------------------



/proton/trunk/proton-c/bindings/python/proton.py
<https://reviews.apache.org/r/9413/#comment35008>

    I don't recall offhand if this notation works on python 2.4. If you haven't 
already, we should probably check just to be sure.



/proton/trunk/tests/python/proton_tests/interop.py
<https://reviews.apache.org/r/9413/#comment35009>

    I'm thinking we should actually check the data files into svn and load them 
from disk rather than generate them on the fly. I think this would be good for 
two reasons (1) the unit tests in other languages wouldn't need to depend on 
python, and (2) we would also catch compensating bugs, e.g. if I generate the 
test data on the fly all the time then I could introduce an encode bug along 
with a compensating decode bug and the tests wouldn't catch it, whereas if we 
check the files in, then we can verify that the current decode will work 
properly against historic versions of the encode as well.



/proton/trunk/tests/python/proton_tests/interop.py
<https://reviews.apache.org/r/9413/#comment35010>

    I'm wondering if we shouldn't just add methods to the Message class to 
directly encode/decode to/from binary data. This seems like it might be 
convenient here and its also plausible that end users would find it useful. It 
would also let us check interop on a few small aspects of the message 
encode/decode that might not be covered by just checking through the data 
interface here.



/proton/trunk/tests/python/proton_tests/interop_gen.py
<https://reviews.apache.org/r/9413/#comment35011>

    I'm wondering if we'll want to define a format such that we can add 
multiple values to check edge cases, e.g. binary with nulls in it, or a string 
with unicode characters, empty binary, string, and symbol values, etc.
    
    I'm not sure offhand what the best way to deal with that would be. I guess 
in some ways we'd need to update all the unit tests anyways in order to add to 
an extra assertion, so there might not be all that much point to a generic 
format.
    
    Either way I'd suggest starting out with as many edge cases as we can think 
of so when we port the python unit tests you have to other languages we get the 
additional coverage. Edgecases are a pretty common place where subtle mapping 
issues creep in. It's easy to confuse binary, string, and symbol and map them 
all into the same thing if you don't have sample values that exercise their 
differences.


- Rafael Schloming


On Feb. 12, 2013, 5:40 p.m., Alan Conway wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/9413/
> -----------------------------------------------------------
> 
> (Updated Feb. 12, 2013, 5:40 p.m.)
> 
> 
> Review request for qpid, Kenneth Giusti and Rafael Schloming.
> 
> 
> Description
> -------
> 
> PROTON-215: Add tests to verify AMQP type support for python bindings.
> 
> The test consists of
> - a generator program that generates AMQP fragments for all AMQP types
> - a set of unit tests to decode, verify and re-encode the fragments.
> 
> This will be extended intended to cover all the proton language bindings.  
> All bindings
> will re-use the generator, unit tests will be added for each language.
> 
> 
> Diffs
> -----
> 
>   /proton/trunk/proton-c/bindings/python/proton.py 1443924 
>   /proton/trunk/tests/python/proton_tests/__init__.py 1443924 
>   /proton/trunk/tests/python/proton_tests/interop.py PRE-CREATION 
>   /proton/trunk/tests/python/proton_tests/interop_gen.py PRE-CREATION 
> 
> Diff: https://reviews.apache.org/r/9413/diff/
> 
> 
> Testing
> -------
> 
> described_array test fails, skipping. I believe the tests have found their 
> first bug :)
> 
> 
> Thanks,
> 
> Alan Conway
> 
>

Reply via email to