alandefreitas / pareto Goto Github PK

Hello,

I am trying to understand what the number on the right side of the = sign represent in the examples:

import pyfront

pf = pyfront.front(['minimization', 'maximization'])

pf[0.68322, 0.545438] = 17
pf[-0.204484, 0.819538] = 8

In this case, I assume in pf[0.68322, 0.545438] = 17, 0.68322 represent the value of objective 1 and 0.545438 represents the value of objective 2. Is that so? What does 17 represent? Could you clarify what the different numbers stand for?

Also could verify if there's a way to install the library using a package manager (like pip) or does it have to be compiled from source?

Would it be possible to provide python bindings with the 36m suffix? The current bindings is pareto.cpython-38-x86_64-linux-gnu.so and the 38 is not supported on my linux machine, while the pyfront.cpython-36m-x86_64-linux-gnu.so was supported.

Would appreciate a 36m version of the bindings.

Feature category

• enhancement - indicators
• enhancement - data structures
• enhancement - bindings
• enhancement - build system
• enhancement - documentation

The problem
Hello Alan. This is a great library. Thank you for releasing it.
Can you please make the default_capacity of 1000 as a compile time
define?

static constexpr size_t default_capacity =
number_of_compile_dimensions == 0 ? 1000
: number_of_compile_dimensions <= 10
? static_cast(50)
<< (number_of_compile_dimensions - 1)
: 100000;

I ran into an issue where I exceeded the 1000 limit and end up asserting when _size != total_front_size().
It took some tracing to see why there was an assert as I didn't know of the 1000 limit.

The solution I'd like
#ifndef DEFAULT_CAPACITY
#define DEFAULT_CAPACITY 1000
#endif

static constexpr size_t default_capacity =
number_of_compile_dimensions == 0 ? DEFAULT_CAPACITY
: number_of_compile_dimensions <= 10
? static_cast(50)
<< (number_of_compile_dimensions - 1)
: 100000;

Alternatives I've considered
None

Additional context

Hi Alan, I'm making a call to pareto::front.hypervolume() and found that this call leaks memory. When running through valgrind, this is the error that I'm getting:

==77189== Mismatched free() / delete / delete []
==77189==    at 0x4C32D3B: free (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==77189==    by 0x1A6082: pareto::fpli_hv(double*, int, int, double const*) (hypervolume.h:820)
==77189==    by 0x1ACCDA: pareto::front<long, 3ul, unsigned int, pareto::r_tree<long, 3ul, unsigned int, std::less<long>, std::allocator<std::pair<pareto::point<long, 3ul, void> const, unsigned int> > > >::hypervolume(pareto::point<long, 3ul, void>) const (front.h:930)

And the final valgrind ouput looks like this:

==77189== 
==77189== HEAP SUMMARY:
==77189==     in use at exit: 391,049 bytes in 4,115 blocks
==77189==   total heap usage: 1,901,304 allocs, 1,897,189 frees, 471,587,726 bytes allocated
==77189== 
==77189== LEAK SUMMARY:
==77189==    definitely lost: 8,008 bytes in 32 blocks
==77189==    indirectly lost: 265,176 bytes in 2,133 blocks
==77189==      possibly lost: 0 bytes in 0 blocks
==77189==    still reachable: 117,865 bytes in 1,950 blocks
==77189==         suppressed: 0 bytes in 0 blocks
==77189== Rerun with --leak-check=full to see details of leaked memory
==77189== 
==77189== For counts of detected and suppressed errors, rerun with: -v
==77189== ERROR SUMMARY: 32487 errors from 7 contexts (suppressed: 0 from 0)

Do you have any ideas for fixing this problem? Thanks!