# Should associate trigger automatic finalization? **URL:** \ **Category:** Help\ **Created:** [November 23, 2021, 1:33pm UTC](https://fortran-lang.discourse.group/t/should-associate-trigger-automatic-finalization/2324 "2021-11-23T13:33:18Z")\ **Posts on this page:** 11\ **Page:** 1
**Author:** ![ivanpribec](https://yyz2.discourse-cdn.com/free1/user_avatar/fortran-lang.discourse.group/ivanpribec/32/3290_2.png) [@ivanpribec](https://fortran-lang.discourse.group/u/ivanpribec)\ **Post date:** [November 23, 2021, 1:33pm UTC](https://fortran-lang.discourse.group/t/should-associate-trigger-automatic-finalization/2324/1 "2021-11-23T13:33:18Z")
Consider the following program: ```fortran module associate_example implicit none type :: sparse integer :: colidx integer, allocatable :: rows(:) real, allocatable :: values(:) contains final :: finalize end type contains subroutine finalize(this) type(sparse) :: this print *, "Finalizing column ", this%colidx end subroutine end module associate_example program associate_test use associate_example implicit none associate(s => sparse(3,[1,2,3],[5.,3.,8.])) print *, "In associate:" print *, s%colidx print *, s%rows print *, s%values end associate block type(sparse) :: s print *, "In block:" s = sparse(3,[1,2,3],[5.,3.,8.]) print *, s%colidx print *, s%rows print *, s%values end block end program ``` Both gfortran (v10.3) and ifort (v2021.4) compile the program without warnings. The output of gfortran is: ```nohighlight $ gfortran -Wall associate_test.f90 $ ./a.out munmap_chunk(): invalid pointer Program received signal SIGABRT: Process abort signal. ``` The output of Intel Fortran is: ```nohighlight $ ifort -warn all associate_test.f90 $ ./a.out In associate: 3 1 2 3 5.000000 3.000000 8.000000 In block: Finalizing column 0 3 1 2 3 5.000000 3.000000 8.000000 Finalizing column 3 ``` I would expect finalization to occur at the end of the associate construct. Is this mandated by the standard, or should we avoid associating with an expression that results in a derived type? In the result from ifort, does the output line finalizing column 0 come from the finalization of the left-hand side in the assignment? ---
**Author:** ![ivanpribec](https://yyz2.discourse-cdn.com/free1/user_avatar/fortran-lang.discourse.group/ivanpribec/32/3290_2.png) [@ivanpribec](https://fortran-lang.discourse.group/u/ivanpribec)\ **Post date:** [November 23, 2021, 1:58pm UTC](https://fortran-lang.discourse.group/t/should-associate-trigger-automatic-finalization/2324/2 "2021-11-23T13:58:12Z")
If I overload the structure constructor by adding a custom constructor function: ```fortran interface sparse module procedure new_sparse end interface ``` ```fortran function new_sparse(idx,rows,values) result(new) integer, intent(in) :: idx integer, intent(in) :: rows(:) real, intent(in) :: values(:) type(sparse) :: new new%colidx = idx allocate(new%rows,source=rows) allocate(new%values,source=values) end function ``` I get the behavior I expected: ```nohighlight $ ifort -warn all associate_test.f90 (base) ipribec@ipribec-T530:~/fortran$ ./a.out In associate: 3 1 2 3 5.000000 3.000000 8.000000 Finalizing column 3 In block: Finalizing column 0 Finalizing column 3 3 1 2 3 5.000000 3.000000 8.000000 Finalizing column 3 ``` ---
**Author:** ![everythingfunctional](https://yyz2.discourse-cdn.com/free1/user_avatar/fortran-lang.discourse.group/everythingfunctional/32/176_2.png) [@everythingfunctional](https://fortran-lang.discourse.group/u/everythingfunctional)\ **Post date:** [November 23, 2021, 2:50pm UTC](https://fortran-lang.discourse.group/t/should-associate-trigger-automatic-finalization/2324/3 "2021-11-23T14:50:41Z")
In the case of gfortran, I have submitted several bug reports related to memory errors in the use of `associate`. In the case of Intel, without the overloaded constructor, it looks as though some “optimizations” are being made, that may or may not be standards conforming. Perhaps someone with a more thorough understanding of this part of the standard could comment on whether this is allowed. ---
**Author:** ![FortranFan](https://avatars.discourse-cdn.com/v4/letter/f/96bed5/32.png) [@FortranFan](https://fortran-lang.discourse.group/u/FortranFan)\ **Post date:** [November 23, 2021, 3:34pm UTC](https://fortran-lang.discourse.group/t/should-associate-trigger-automatic-finalization/2324/4 "2021-11-23T15:34:49Z")
> [@ivanpribec](#): > > … I would expect finalization to occur at the end of the associate construct. > > Is this mandated by the standard, or should we avoid associating with an expression that results in a derived type? … Per my understanding, the standard is silent on the question of the `ASSOCIATE` construct when the selector is an expression of a finalizable entity. Thus the chances are zero any of the compiler implementations will perform finalization in the first case shown here. If finalization is important, then as a matter of good practice it does make sense to “avoid associating with an expression that results in a derived type.” But instead to make the selector an object that will be finalized as per the semantics stated in the standard. ---
**Author:** ![FortranFan](https://avatars.discourse-cdn.com/v4/letter/f/96bed5/32.png) [@FortranFan](https://fortran-lang.discourse.group/u/FortranFan)\ **Post date:** [November 23, 2021, 3:36pm UTC](https://fortran-lang.discourse.group/t/should-associate-trigger-automatic-finalization/2324/5 "2021-11-23T15:36:54Z")
> [@ivanpribec](#): > > … In the result from ifort, does the output line finalizing column 0 come from the finalization of the left-hand side in the assignment? Re: “the output line finalizing column 0,” it is the LHS in the assignment `s = sparse(3, ..)` in the `BLOCK` construct getting finalized as per the standard. ---
**Author:** ![Reinhold\_Bader](https://avatars.discourse-cdn.com/v4/letter/r/e95f7d/32.png) [@Reinhold\_Bader](https://fortran-lang.discourse.group/u/Reinhold_Bader)\ **Post date:** [November 23, 2021, 3:39pm UTC](https://fortran-lang.discourse.group/t/should-associate-trigger-automatic-finalization/2324/6 "2021-11-23T15:39:37Z")
For the case where the structure constructor is used: no finalization should happen here at all, because use of the structure constructor is not one of the trigger cases. For the case where a function overloads the structure constructor, the following text from the standard applies, I think: F2018 / 7.5.6.3 when finalization occurs … If an executable construct references a nonpointer function, the result is finalized after execution of the innermost executable construct containing the reference. … which I take to mean that the finalizer must be executed once the ASSOCIATE block completes. Regards Reinhold ---
**Author:** ![FortranFan](https://avatars.discourse-cdn.com/v4/letter/f/96bed5/32.png) [@FortranFan](https://fortran-lang.discourse.group/u/FortranFan)\ **Post date:** [November 23, 2021, 3:53pm UTC](https://fortran-lang.discourse.group/t/should-associate-trigger-automatic-finalization/2324/7 "2021-11-23T15:53:15Z")
> [@Reinhold\_Bader](#): > > For the case where a function overloads the structure constructor, the following text from the standard applies, I think: > > F2018 / 7.5.6.3 when finalization occurs > … > If an executable construct references a nonpointer function, the result is finalized after execution of the innermost executable construct containing the reference. > … > which I take to mean that the finalizer must be executed once the ASSOCIATE block completes. @Reinhold_Bader , welcome to the forum. My hunch is compiler implementations will find the `associate-stmt` as not an executable construct and not do the finalization, that the instructions in the block part of the construct, the stuff between the `associate-stmt` and the `end-associate-stmt` is what they consider an executable-construct here. ---
**Author:** ![Reinhold\_Bader](https://avatars.discourse-cdn.com/v4/letter/r/e95f7d/32.png) [@Reinhold\_Bader](https://fortran-lang.discourse.group/u/Reinhold_Bader)\ **Post date:** [November 23, 2021, 4:10pm UTC](https://fortran-lang.discourse.group/t/should-associate-trigger-automatic-finalization/2324/8 "2021-11-23T16:10:43Z")
Maybe some people do, but the standard text does consider the associate-stmt as part of the associate-construct (F2018 / 11.1.3.1). ---
**Author:** ![certik](https://yyz2.discourse-cdn.com/free1/user_avatar/fortran-lang.discourse.group/certik/32/4_2.png) [@certik](https://fortran-lang.discourse.group/u/certik)\ **Post date:** [November 23, 2021, 4:19pm UTC](https://fortran-lang.discourse.group/t/should-associate-trigger-automatic-finalization/2324/9 "2021-11-23T16:19:21Z")
@Reinhold_Bader thanks for joining the forum and your feedback! ---
**Author:** ![ivanpribec](https://yyz2.discourse-cdn.com/free1/user_avatar/fortran-lang.discourse.group/ivanpribec/32/3290_2.png) [@ivanpribec](https://fortran-lang.discourse.group/u/ivanpribec)\ **Post date:** [November 23, 2021, 4:28pm UTC](https://fortran-lang.discourse.group/t/should-associate-trigger-automatic-finalization/2324/10 "2021-11-23T16:28:13Z")
Thanks Reinhold for the answer. The example I constructed here is artificial in the sense that the derived type does not need a custom finalizer. However for any object which requires custom finalization (e.g. because it contains pointers) it’s probably a good idea to overload the structure constructor. This seems very close to the programming idiom/pattern known as the “Rule of zero/three/five” in C++: - [The rule of three/five/zero (cppreference.com)](https://en.cppreference.com/w/cpp/language/rule_of_three) - [Rule of three (Wikipedia)](https://en.wikipedia.org/wiki/Rule_of_three_(C%2B%2B_programming)) ---
**Author:** ![Reinhold\_Bader](https://avatars.discourse-cdn.com/v4/letter/r/e95f7d/32.png) [@Reinhold\_Bader](https://fortran-lang.discourse.group/u/Reinhold_Bader)\ **Post date:** [November 23, 2021, 4:36pm UTC](https://fortran-lang.discourse.group/t/should-associate-trigger-automatic-finalization/2324/11 "2021-11-23T16:36:27Z")
I agree. The structure constructor is so restricted in semantics that executing a finalizer is simply not necessary. But for the same reason, it usually cannot do what needs to be done for POINTER components. Cheers.