Subject: Re: [geometry] Proposal: grouping tests according to the directory/logical structure
From: Barend Gehrels (barend_at_[hidden])
Date: 2015-05-27 15:57:00
Hi Adam,
Adam Wulkiewicz schreef op 27-5-2015 om 19:39:
>
> We have many tests at Geometry in a single matrix:
> http://www.boost.org/development/tests/develop/developer/geometry.html
> For me it is sometimes hard to quickly find the test I'm looking for.
>
> I propose to group the tests the same way how they're grouped in the
> directories structure by naming them accordingly. Each test name would
> contain:
> 1. a name (or names) of a logical part of the library taken from the
> containing directory (algorithms, strategies, core, etc.)
> 2. a name of the tested algorithm, class, concept (within, box, etc.)
> 3. a name of test variant (e.g. areal_areal or multi)
>
> Furthermore I propose to keep the file and directories structure short
> (i.e. in many cases the same as they are now) and to specify the names
> in Jamfiles.
> This can be done by setting the target-name in
> run/compile/compile-fail rules, as mentioned here:
> http://www.boost.org/build/doc/html/bbv2/builtins/testing.html
>
> The target-name would be created from the directory name,
> algorithm/concept name and test variant name separated by underscore,
> e.g.:
> algorithms/area_multi.cpp -> algorithms_area_multi
> algorithms/buffer/buffer_point.cpp -> algorithms_buffer_point
> algorithms/distance/distance.cpp -> algorithms_distance
> algorithms/relational_operations/relate/relate_areal_areal.cpp ->
> algorithms_relate_areal_areal
> core/access.cpp -> core_access
> geometries/box.cpp -> geometries_box
> strategies/andoyer.cpp -> strategies_andoyer
> util/range.cpp -> util_range
> etc.
>
> I also propose to rename some of the files to reflect the new naming
> convention, i.e. to always keep test variant identifier at the end (as
> on the list above), e.g.:
> multi_area.cpp -> area_multi.cpp
> point_buffer.cpp -> buffer_point.cpp
> etc.
Thanks for the proposal. I agree with this, both the target-names and
the renaming of some of the files.
If there are no objections, I can take care for the buffer part (as it
needs also the same renaming in my non-committed project files)
>
> Alternatively, sometimes the name of an algorithm could also be
> ommited in a file name if it was redundant and clear enough, e.g.:
> buffer/point_buffer.cpp -> buffer/point.cpp
> however personally I'd prefer keeping the redundant algorithm name. At
> least for now. And if we agreed that it should be done this way, do it
> in the future (as the next step) consistently for all of the tests.
OK for me.
Regards, Barend