[datatable-help] Playing nice(r) with S4?

[Steve Lianoglou]

Almost forgot ... I also added a minimal *.Rd file for the S4-class definition:

https://github.com/lianos/datatable/blob/master/pkg/man/data.table-class.Rd

-steve



On Tue, Nov 30, 2010 at 12:03 AM, Steve Lianoglou
<mailinglist.honeypot at gmail.com> wrote:
> Hi,
>
> I think running the tests by invoking the test.data.table from inside
> an "Examples" section of an .Rd file are causing some pain.
>
> There are complaints of locked environments:
>
> Error in assign(mname, def, where) :
>  cannot add bindings to a locked environment
>
> when defining S4 classes and methods inside the test.data.table function.
>
> Defining the class outside of the function works and lets the tests
> pass. Defining functions out there doesn't help, so those tests are
> commented out (see lines 552 - 556):
>
> https://github.com/lianos/datatable/blob/master/pkg/R/test.data.table.R
>
> I'm not sure how you feel about this approach, but these work.
>
> IMHO it would be better to use something like testthat or RUnit and
> run the test harness for the package with those. I have a feeling we'd
> avoid these problems that way. If the devs are willing to consider
> that, I'd volunteer to port the tests over.
>
> At any rate, let me know if you're OK with these S4 modifications and
> I can commit back to the svn repo.
>
> In summary, I've:
>
> * Added an AllS4.R file to add the setOldClass stuff to include
> data.table's in S4 "slots" and defined as(...) methods to convert
> data.table to data.frame, and vice versa.
>
> * Modified the namespace to export the data.table class for other
> packages to use it.
>
> * Bumped version number in DESCRIPTION to 1.5.2, and added an Imports:
> methods to it.
>
> * Made changes to the NEWS file
>
> * Added some tests to test.data.table() function.
>
> I've used this updated package as an import to another package I'm
> dev'ing and things seem to work as expected ...
>
> Thanks,
> -steve
>
>
> On Wed, Nov 24, 2010 at 6:01 PM, Matthew Dowle <mdowle at mdowle.plus.com> wrote:
>>
>> For the setClass call itself that would be ok, and then use the class
>> somehow, to test it worked?
>>
>> setClass("DTS4", representation(info="data.table"))
>> test(228, is.data.table(new("DTS4")))
>>
>>
>> On Wed, 2010-11-24 at 14:26 -0500, Steve Lianoglou wrote:
>>> Hi,
>>>
>>> On Wed, Nov 24, 2010 at 2:13 PM, Matthew Dowle <mdowle at mdowle.plus.com> wrote:
>>> > Try
>>> >
>>> >  test(228, setClass("DTS4", representation(info="data.table")), "DTS4")
>>>
>>> I tried that at first, but you get an error from (I think) defining a
>>> new class in the wrong place, specifically:
>>>
>>> Test 228 Error in assign(mname, def, where) :
>>>   cannot add bindings to a locked environment
>>>
>>> > In test.data.table we only use try() for when the correct result is an
>>> > error and the test is that we get the right error message with a grep in
>>> > another test just afterwards. Otherwise we can leave test() to try().
>>> >
>>> > Code that shouldn't fail can just be written (and would cause
>>> > test.data.table to stop then and there on error) or inside a test().
>>>
>>> So, in light of the above problem, would you rather I just write the
>>> code that defines the new S4 class and not test that "it worked" since
>>> we assume implicitly it works if no error stops test.data.table() form
>>> completing?
>>>
>>> -steve
>>>
>>
>>
>>
>
>
>
> --
> Steve Lianoglou
> Graduate Student: Computational Systems Biology
>  | Memorial Sloan-Kettering Cancer Center
>  | Weill Medical College of Cornell University
> Contact Info: http://cbio.mskcc.org/~lianos/contact
>



-- 
Steve Lianoglou
Graduate Student: Computational Systems Biology
 | Memorial Sloan-Kettering Cancer Center
 | Weill Medical College of Cornell University
Contact Info: http://cbio.mskcc.org/~lianos/contact