Clicky
Showing changes from revision #8 to #9:
(追記) Added (追記ここまで) | (削除) Removed (削除ここまで) | (削除) Chan (削除ここまで)(追記) ged (追記ここまで)
NOTE: Gfortran has a bug that prevents using aliases via "USE, ONLY" as done here. See below for a work-around.
! FILE: c_interface_module.f
! PURPOSE: Supplement ISO-C-Binding to provide type aliases and interfaces
! to common ISO-C string functions to aid working with strings.
! AUTHOR: Joseph M. Krahn
! STATUS: Still in development. Reasonably complete, but somewhat limited testing.
!
! The idea is to provide type aliases for all ISO-C types, so that the
! Fortran interface code more explicitly defines the actual C interface.
! This should be updated to support F2008 variable-length allocatable
! strings.
!
! Entity names all have the "C_" prefix, as with ISO-C-Binding, with a
! few exceptions.
!
module C_interface_module
use, intrinsic :: ISO_C_Binding, &
! C type aliases for pointer derived types:
C_ptr => C_ptr , &
C_char_ptr => C_ptr, &
C_const_char_ptr => C_ptr, &
C_void_ptr => C_ptr, &
C_const_void_ptr => C_ptr
implicit none
!----------------------------------------------------------------------------
! C type aliases for intrinsic type KIND parameters:
! NOTE: a C enum may not always be a standard C int
integer, parameter :: C_enum = C_int
! Defining off_t is difficult, because it may depend on "LARGEFILE" selection.
! integer, parameter :: C_off_t = ??
! C string terminator alais using the 3-letter ASCII name.
! The C_ prefix is not used because it is just an ASCII character.
character(len=1,kind=C_char), parameter :: NUL = C_NULL_char
! NOTE: In C, "char" is distinct from "signed char", unlike integers.
! The plain "char" type is specific for text/string values, whereas
! "signed char" should indicate 1-byte integer data.
!
! Most ISO-C systems have wide chars "wchar_t", but Fortran compilers
! have limited support for different character kinds. UTF encoding
! adds more complexity. This should be updated as Fortran compilers
! include support for more character types.
!
! Fortran does not (yet) support unsigned types.
integer, parameter :: &
C_unsigned = C_int, &
C_unsigned_short = C_short, &
C_unsigned_long = C_long, &
C_unsigned_long_long = C_long_long, &
C_unsigned_char = C_signed_char, &
C_ssize_t = C_size_t, &
C_uint8_t = C_int8_t, &
C_uint16_t = C_int16_t, &
C_uint32_t = C_int32_t, &
C_uint64_t = C_int64_t, &
C_uint_least8_t = C_int_least8_t, &
C_uint_least16_t = C_int_least16_t, &
C_uint_least32_t = C_int_least32_t, &
C_uint_least64_t = C_int_least64_t, &
C_uint_fast8_t = C_int_fast8_t, &
C_uint_fast16_t = C_int_fast16_t, &
C_uint_fast32_t = C_int_fast32_t, &
C_uint_fast64_t = C_int_fast64_t, &
C_uintmax_t = C_intmax_t
! Note: ptrdiff_t cannot be reliably defined from other types.
! When practical, it is larger than a pointer because it benefits
! from the full unsigned range in both positive and negative directions.
! Integer versions including 'int', where the 'int' is optional:
integer, parameter :: &
C_short_int = C_short, &
C_long_int = C_long, &
C_long_long_int = C_long_long, &
C_unsigned_int = C_unsigned, &
C_unsigned_short_int = C_short, &
C_unsigned_long_int = C_long, &
C_unsigned_long_long_int = C_long_long
interface C_F_string
module procedure C_F_string_ptr
module procedure C_F_string_chars
end interface C_F_string
interface F_C_string
module procedure F_C_string_ptr
module procedure F_C_string_chars
end interface F_C_string
!=======================================================================
! Some useful ISO C library string functions from <string.h>
! These are based on GCC header sections marked as NAMESPACE_STD
interface
! Copy N bytes of SRC to DEST, no aliasing or overlapping allowed.
!extern void *memcpy (void *dest, const void *src, size_t n);
function C_memcpy(dest, src, n) result(result) bind(C,name="memcpy")
import C_void_ptr, C_size_t
type(C_void_ptr) :: result
type(C_void_ptr), value, intent(in) :: dest ! target=intent(out)
type(C_void_ptr), value, intent(in) :: src ! target=intent(in)
integer(C_size_t), value, intent(in) :: n
end function C_memcpy
! Copy N bytes of SRC to DEST, guaranteeing correct behavior for overlapping strings.
!extern void *memmove (void *dest, const void *src, size_t n)
function C_memmove(dest, src, n) result(result) bind(C,name="memmove")
import C_void_ptr, C_size_t
type(C_void_ptr) :: result
type(C_void_ptr), value, intent(in) :: dest ! target=intent(out)
type(C_void_ptr), value, intent(in) :: src
integer(C_size_t), value, intent(in) :: n
end function C_memmove
! Set N bytes of S to C.
!extern void *memset (void *s, int c, size_t n)
function C_memset(s, c, n) result(result) bind(C,name="memset")
import C_void_ptr, C_int, C_size_t
type(C_void_ptr) :: result
type(C_void_ptr), value, intent(in) :: s ! target=intent(out)
integer(C_int), value, intent(in) :: c
integer(C_size_t), value, intent(in) :: n
end function C_memset
! Compare N bytes of S1 and S2.
!extern int memcmp (const void *s1, const void *s2, size_t n)
pure &
function C_memcmp(s1, s2, n) result(result) bind(C,name="memcmp")
import C_int, C_void_ptr, C_size_t
integer(C_int) :: result
type(C_void_ptr), value, intent(in) :: s1
type(C_void_ptr), value, intent(in) :: s2
integer(C_size_t), value, intent(in) :: n
end function C_memcmp
! Search N bytes of S for C.
!extern void *memchr (const void *s, int c, size_t n)
pure &
function C_memchr(s, c, n) result(result) bind(C,name="memchr")
import C_void_ptr, C_int, C_size_t
type(C_void_ptr) :: result
type(C_void_ptr), value, intent(in) :: s
integer(C_int), value, intent(in) :: c
integer(C_size_t), value, intent(in) :: n
end function C_memchr
! Copy SRC to DEST.
!extern char *strcpy (char *dest, const char *src)
function C_strcpy(dest, src) result(result) bind(C,name="strcpy")
import C_char_ptr, C_size_t
type(C_char_ptr) :: result
type(C_char_ptr), value, intent(in) :: dest ! target=intent(out)
type(C_char_ptr), value, intent(in) :: src
end function C_strcpy
! Copy no more than N characters of SRC to DEST.
!extern char *strncpy (char *dest, const char *src, size_t n)
function C_strncpy(dest, src, n) result(result) bind(C,name="strncpy")
import C_char_ptr, C_size_t
type(C_char_ptr) :: result
type(C_char_ptr), value, intent(in) :: dest ! target=intent(out)
type(C_char_ptr), value, intent(in) :: src
integer(C_size_t), value, intent(in) :: n
end function C_strncpy
! Append SRC onto DEST.
!extern char *strcat (char *dest, const char *src)
function C_strcat(dest, src) result(result) bind(C,name="strcat")
import C_char_ptr, C_size_t
type(C_char_ptr) :: result
type(C_char_ptr), value, intent(in) :: dest ! target=intent(out)
type(C_char_ptr), value, intent(in) :: src
end function C_strcat
! Append no more than N characters from SRC onto DEST.
!extern char *strncat (char *dest, const char *src, size_t n)
function C_strncat(dest, src, n) result(result) bind(C,name="strncat")
import C_char_ptr, C_size_t
type(C_char_ptr) :: result
type(C_char_ptr), value, intent(in) :: dest ! target=intent(out)
type(C_char_ptr), value, intent(in) :: src
integer(C_size_t), value, intent(in) :: n
end function C_strncat
! Compare S1 and S2.
!extern int strcmp (const char *s1, const char *s2)
pure &
function C_strcmp(s1, s2) result(result) bind(C,name="strcmp")
import C_int, C_char_ptr, C_size_t
integer(C_int) :: result
type(C_char_ptr), value, intent(in) :: s1
type(C_char_ptr), value, intent(in) :: s2
end function C_strcmp
! Compare N characters of S1 and S2.
!extern int strncmp (const char *s1, const char *s2, size_t n)
pure &
function C_strncmp(s1, s2, n) result(result) bind(C,name="strncmp")
import C_int, C_char_ptr, C_size_t
integer(C_int) :: result
type(C_char_ptr), value, intent(in) :: s1
type(C_char_ptr), value, intent(in) :: s2
integer(C_size_t), value, intent(in) :: n
end function C_strncmp
! Return the length of S.
!extern size_t strlen (const char *s)
pure &
function C_strlen(s) result(result) bind(C,name="strlen")
import C_char_ptr, C_size_t
integer(C_size_t) :: result
type(C_char_ptr), value, intent(in) :: s !character(len=*), intent(in)
end function C_strlen
end interface
! End of <string.h>
!=========================================================================
! Standard ISO-C malloc routines:
interface
! void *calloc(size_t nmemb, size_t size);
type(C_void_ptr) &
function C_calloc(nmemb, size) bind(C,name="calloc")
import C_void_ptr, C_size_t
integer(C_size_t), value, intent(in) :: nmemb, size
end function C_calloc
! void *malloc(size_t size);
type(C_void_ptr) &
function C_malloc(size) bind(C,name="malloc")
import C_void_ptr, C_size_t
integer(C_size_t), value, intent(in) :: size
end function C_malloc
! void free(void *ptr);
subroutine C_free(ptr) bind(C,name="free")
import C_void_ptr
type(C_void_ptr), value, intent(in) :: ptr
end subroutine C_free
! void *realloc(void *ptr, size_t size);
type(C_void_ptr) &
function C_realloc(ptr,size) bind(C,name="realloc")
import C_void_ptr, C_size_t
type(C_void_ptr), value, intent(in) :: ptr
integer(C_size_t), value, intent(in) :: size
end function C_realloc
end interface
interface assignment(=)
module procedure F_string_assign_C_string
end interface assignment(=)
!==========================================================================
contains
! HACK: For some reason, C_associated was not defined as pure.
pure logical &
function C_associated_pure(ptr) result(associated)
type(C_ptr), intent(in) :: ptr
integer(C_intptr_t) :: iptr
iptr = transfer(ptr,iptr)
associated = (iptr /= 0)
end function C_associated_pure
! Set a fixed-length Fortran string to the value of a C string.
subroutine F_string_assign_C_string(F_string, C_string)
character(len=*), intent(out) :: F_string
type(C_ptr), intent(in) :: C_string
character(len=1,kind=C_char), pointer :: p_chars(:)
integer :: i
if (.not. C_associated(C_string) ) then
F_string = ' '
else
call C_F_pointer(C_string,p_chars,[huge(0)])
i=1
do while(p_chars(i)/=NUL .and. i<=len(F_string))
F_string(i:i) = p_chars(i)
i=i+1
end do
if (i<len(F_string)) F_string(i:) = ' '
end if
end subroutine F_string_assign_C_string
! Copy a C string, passed by pointer, to a Fortran string.
! If the C pointer is NULL, the Fortran string is blanked.
! C_string must be NUL terminated, or at least as long as F_string.
! If C_string is longer, it is truncated. Otherwise, F_string is
! blank-padded at the end.
subroutine C_F_string_ptr(C_string, F_string)
type(C_ptr), intent(in) :: C_string
character(len=*), intent(out) :: F_string
character(len=1,kind=C_char), dimension(:), pointer :: p_chars
integer :: i
if (.not. C_associated(C_string)) then
F_string = ' '
else
call C_F_pointer(C_string,p_chars,[huge(0)])
i=1
do while(p_chars(i)/=NUL .and. i<=len(F_string))
F_string(i:i) = p_chars(i)
i=i+1
end do
if (i<len(F_string)) F_string(i:) = ' '
end if
end subroutine C_F_string_ptr
! Copy a C string, passed as a char-array reference, to a Fortran string.
subroutine C_F_string_chars(C_string, F_string)
character(len=1,kind=C_char), intent(in) :: C_string(*)
character(len=*), intent(out) :: F_string
integer :: i
i=1
do while(C_string(i)/=NUL .and. i<=len(F_string))
F_string(i:i) = C_string(i)
i=i+1
end do
if (i<len(F_string)) F_string(i:) = ' '
end subroutine C_F_string_chars
! Copy a Fortran string to an allocated C string pointer.
! If the C pointer is NULL, no action is taken. (Maybe auto allocate via libc call?)
! If the length is not passed, the C string must be at least: len(F_string)+1
! If the length is passed and F_string is too long, it is truncated.
subroutine F_C_string_ptr(F_string, C_string, C_string_len)
character(len=*), intent(in) :: F_string
type(C_ptr), intent(in) :: C_string ! target = intent(out)
integer, intent(in), optional :: C_string_len ! Max string length,
! INCLUDING THE TERMINAL NUL
character(len=1,kind=C_char), dimension(:), pointer :: p_chars
integer :: i, strlen
strlen = len(F_string)
if (present(C_string_len)) then
if (C_string_len <= 0) return
strlen = min(strlen,C_string_len-1)
end if
if (.not. C_associated(C_string)) then
return
end if
call C_F_pointer(C_string,p_chars,[strlen+1])
forall (i=1:strlen)
p_chars(i) = F_string(i:i)
end forall
p_chars(strlen+1) = NUL
end subroutine F_C_string_ptr
pure &
function C_strlen_safe(s) result(length)
integer(C_size_t) :: length
type(C_char_ptr), value, intent(in) :: s
if (.not. C_associated_pure(s)) then
length = 0
else
length = C_strlen(s)
end if
end function C_strlen_safe
function C_string_value(C_string) result(F_string)
type(C_ptr), intent(in) :: C_string
character(len=C_strlen_safe(C_string)) :: F_string
character(len=1,kind=C_char), dimension(:), pointer :: p_chars
integer :: i, length
length = len(F_string)
if (length/=0) then
call C_F_pointer(C_string,p_chars,[length])
forall (i=1:length)
F_string(i:i) = p_chars(i)
end forall
end if
end function C_string_value
! Copy a Fortran string to a C string passed by char-array reference.
! If the length is not passed, the C string must be at least: len(F_string)+1
! If the length is passed and F_string is too long, it is truncated.
subroutine F_C_string_chars(F_string, C_string, C_string_len)
character(len=*), intent(in) :: F_string
character(len=1,kind=C_char), dimension(*), intent(out) :: C_string
integer, intent(in), optional :: C_string_len ! Max string length,
! INCLUDING THE TERMINAL NUL
integer :: i, strlen
strlen = len(F_string)
if (present(C_string_len)) then
if (C_string_len <= 0) return
strlen = min(strlen,C_string_len-1)
end if
forall (i=1:strlen)
C_string(i) = F_string(i:i)
end forall
C_string(strlen+1) = NUL
end subroutine F_C_string_chars
! NOTE: Strings allocated here must be freed by the
! C library, such as via C_free() or C_string_free(),
type(C_ptr) &
function F_C_string_dup(F_string,length) result(C_string)
character(len=*), intent(in) :: F_string
integer, intent(in), optional :: length
character(len=1,kind=C_char), pointer :: C_string_ptr(:)
integer :: i
integer(C_size_t) :: strlen
if (present(length)) then
strlen = length
else
strlen = len(F_string)
end if
if (strlen <= 0) then
C_string = C_NULL_ptr
else
C_string = C_malloc(strlen+1)
if (C_associated(C_string)) then
call C_F_pointer(C_string,C_string_ptr,[strlen+1])
forall (i=1:strlen)
C_string_ptr(i) = F_string(i:i)
end forall
C_string_ptr(strlen+1) = NUL
end if
end if
end function F_C_string_dup
! NOTE: Strings allocated here must be freed by the
! C library, such as via C_free() or C_string_free(),
type(C_ptr) &
function C_string_alloc(length) result(C_string)
integer(C_size_t), intent(in) :: length
character(len=1,kind=C_char), pointer :: C_charptr
C_string = C_malloc((削除) lenth (削除ここまで)(追記) length (追記ここまで)+1)
if (C_associated(C_string)) then
call C_F_pointer(C_string,C_charptr)
C_charptr = NUL
end if
end function C_string_alloc
subroutine C_string_free(string)
type(C_ptr), intent(inout) :: string
if (C_associated(string)) then
call C_free(string)
string = C_NULL_ptr
end if
end subroutine C_string_free
end module C_interface_moduleGFortran does not support multiple local-names for USE, ONLY: statements with the ISO_C_Binding module. This is a bug that only occurs with the intrinsic module. You can create a simple wrapper module that USEs ISO_C_BINDING, and the local-name aliases work. However, the IMPORT statements are still broken, so a similar work-around is required for every interface. Until this bug is fixed, it is easier just to avoid using the C_ptr aliases. The kind parameters are not affected, because they are written as separate parameters.