# Length(s) of allocatable character array? **URL:** \ **Category:** Help\ **Created:** [April 6, 2025, 1:34am UTC](https://fortran-lang.discourse.group/t/length-s-of-allocatable-character-array/9547 "2025-04-06T01:34:24Z")\ **Posts on this page:** 13\ **Page:** 1
**Author:** ![Harper](https://avatars.discourse-cdn.com/v4/letter/h/b5ac83/32.png) [@Harper](https://fortran-lang.discourse.group/u/Harper)\ **Post date:** [April 6, 2025, 1:34am UTC](https://fortran-lang.discourse.group/t/length-s-of-allocatable-character-array/9547/1 "2025-04-06T01:34:24Z")
One compiler printed ```auto words: "firstsecond" lengths: 5 6 ``` with this program: ```auto implicit none character(:),allocatable::a,b,words(:) a = 'first' b = 'second' allocate(character(max(len(a),len(b))) :: words(2)) words(1) = a words(2) = b print "(4A,2I2)",'words: "',words,'" lengths:',len(words(1)),len(words(2)) end program ``` I suspect that different elements of an allocatable character array can’t have different lengths, but does the Fortran standard say so? If so, where? ---
**Author:** ![DavidB](https://avatars.discourse-cdn.com/v4/letter/d/76d3ee/32.png) [@DavidB](https://fortran-lang.discourse.group/u/DavidB)\ **Post date:** [April 6, 2025, 6:16am UTC](https://fortran-lang.discourse.group/t/length-s-of-allocatable-character-array/9547/2 "2025-04-06T06:16:54Z")
See the first response in [Character array constructor](https://fortran-lang.discourse.group/t/character-array-constructor/5591) ---
**Author:** ![msz59](https://avatars.discourse-cdn.com/v4/letter/m/3d9bf3/32.png) [@msz59](https://fortran-lang.discourse.group/u/msz59)\ **Post date:** [April 6, 2025, 7:39am UTC](https://fortran-lang.discourse.group/t/length-s-of-allocatable-character-array/9547/3 "2025-04-06T07:39:28Z")
But here, there is no array constructor. Instead, we have two assignments to `words` elements 1 and 2. The array surely has all elements of the same length, 6. So the only guess that comes to my mind is that your compiler treats ‘A’ descriptor as the F2023 new ‘AT’, i.e. with auto-trimming. _Edit:_ my guess is wrong, as it also gives `len(words(1))` as `5`. NB. `gfortran` 14.2, `ifx` 2025.0.4 and `ifort` 2021.13.1 all give, IMHO poperly: `words: "first second" lengths: 6 6` ---
**Author:** ![sblionel](https://yyz2.discourse-cdn.com/free1/user_avatar/fortran-lang.discourse.group/sblionel/32/853_2.png) [@sblionel](https://fortran-lang.discourse.group/u/sblionel)\ **Post date:** [April 6, 2025, 3:34pm UTC](https://fortran-lang.discourse.group/t/length-s-of-allocatable-character-array/9547/4 "2025-04-06T15:34:40Z")
> [@Harper](#): > > I suspect that different elements of an allocatable character array can’t have different lengths, but does the Fortran standard say so? If so, where? This is a very interesting question. Length is a type parameter, so changing the length of an individual array element would imply changing that element’s type parameter, but elements don’t have type parameters, only variables do; I haven’t found words saying so explicitly, though (they must exist somewhere!) I then was looking at the description of intrinsic assignment, which is where you’re changing the length. 10.2.1.3 is where this is found, and it says, in part, “If the variable is an allocated allocatable variable, it is deallocated if expr is an array of different shape, any corresponding length type parameter values of the variable and expr differ, or the variable is polymorphic and the dynamic type or any corresponding kind type parameter values of the variable and expr differ.” You can’t, of course, deallocate an array element, but it seems to me that this wording could use some improvement. I will ask. ---
**Author:** ![RonShepard](https://avatars.discourse-cdn.com/v4/letter/r/a3d4f5/32.png) [@RonShepard](https://fortran-lang.discourse.group/u/RonShepard)\ **Post date:** [April 6, 2025, 4:10pm UTC](https://fortran-lang.discourse.group/t/length-s-of-allocatable-character-array/9547/5 "2025-04-06T16:10:58Z")
> [@Harper](#): > > `words: "firstsecond" lengths: 5 6` If the programmer wants this behavior, that is different array elements can have different lengths, or even different allocation status, then in fortran it is done with a derived type array where each element has an allocatable component. An example would be ```auto type string character(:), allocatable :: s end type string type(string) :: words(1:2) ... words(1)%s = 'first' words(2)%s = 'second' ``` There are a couple of downsides to this approach. The programmer cannot initialize such an array ```auto type(string) :: words(1:2)=['first', 'second'] ``` or define such an array as a parameter constant. It would be nice if the language were extended to allow those things, but for now it is not allowed. A second downside is that this is a user defined type, not an intrinsic type, so that introduces compatibility issues when codes and libraries from different origins are mixed together. In these case, the programmer must manually copy or `move_alloc()` the components among the various derived types. ---
**Author:** ![rwmsu](https://avatars.discourse-cdn.com/v4/letter/r/48db29/32.png) [@rwmsu](https://fortran-lang.discourse.group/u/rwmsu)\ **Post date:** [April 6, 2025, 5:21pm UTC](https://fortran-lang.discourse.group/t/length-s-of-allocatable-character-array/9547/6 "2025-04-06T17:21:25Z")
I have always found the requirement to add the allocatable keyword to deferred length strings to be confusing. For folks relatively new to Fortran and who don’t understand that the elements of an array have to have a fixed length, the current syntax might imply that an array of variable length strings is possible (and I would love a way to construct an array of variable length strings without using a derived type). My personal preference would be that the allocatable keyword is only necessary when you want to create an array of strings of fixed length. For a scalar, I would think doing character(`:`) should be enough. You would only need to add the ALLOCATABLE if you want to create an array of fixed length strings at run time ala ```auto character(:) , allocatable :: strings(:) allocate(character(20):: strings(50)) and for just a scalar you want to set at runtime character(:) :: a_string a_string = "This is a string" ``` ---
**Author:** ![wspector](https://avatars.discourse-cdn.com/v4/letter/w/47e85d/32.png) [@wspector](https://fortran-lang.discourse.group/u/wspector)\ **Post date:** [April 6, 2025, 6:34pm UTC](https://fortran-lang.discourse.group/t/length-s-of-allocatable-character-array/9547/7 "2025-04-06T18:34:08Z")
> [@RonShepard](#): > > There are a couple of downsides to this approach. The programmer cannot initialize such an array > > ```auto > type(string) :: words(1:2)=['first', 'second'] > > ``` > > or define such an array as a parameter constant. It would be nice if the language were extended to allow those things, but for now it is not allowed. One can write an assignment like: ` words = [string ('first'), string ('second')]` ---
**Author:** ![RonShepard](https://avatars.discourse-cdn.com/v4/letter/r/a3d4f5/32.png) [@RonShepard](https://fortran-lang.discourse.group/u/RonShepard)\ **Post date:** [April 6, 2025, 8:38pm UTC](https://fortran-lang.discourse.group/t/length-s-of-allocatable-character-array/9547/8 "2025-04-06T20:38:24Z")
> [@wspector](#): > > One can write an assignment like: Yes, the assignment works, but not initialization or as a parameter constant. ```auto 6 | type(string) :: words(1:2) = [string('first'),string('second')] | 1 Error: Invalid initialization expression for ALLOCATABLE component 's' in structure constructor at (1) ``` This is a general limitation of allocatable entities of any type, character, integer, real, etc., or in this case, of an allocatable component. ---
**Author:** ![Harper](https://avatars.discourse-cdn.com/v4/letter/h/b5ac83/32.png) [@Harper](https://fortran-lang.discourse.group/u/Harper)\ **Post date:** [April 6, 2025, 9:23pm UTC](https://fortran-lang.discourse.group/t/length-s-of-allocatable-character-array/9547/9 "2025-04-06T21:23:26Z")
The F2023 standard does forbid character array elements having different lengths: 3.3 says an array is a set of scalar data, all of the same type and type parameters, and 7.2 para 2 says a type parameter is either a kind type parameter or a length type parameter. I have reported the bug in the offending compiler. We shall all be interested to see what happens to 10.2.1.3. ---
**Author:** ![sblionel](https://yyz2.discourse-cdn.com/free1/user_avatar/fortran-lang.discourse.group/sblionel/32/853_2.png) [@sblionel](https://fortran-lang.discourse.group/u/sblionel)\ **Post date:** [April 6, 2025, 9:40pm UTC](https://fortran-lang.discourse.group/t/length-s-of-allocatable-character-array/9547/10 "2025-04-06T21:40:22Z")
Thanks for finding that - I had hoped such wording appeared, but I had not found it. As for 10.2.1.3, that’s not a problem I now realize because, as we’ve said before, an array element cannot itself have the allocatable (nor pointer) attribute. ---
**Author:** ![Harper](https://avatars.discourse-cdn.com/v4/letter/h/b5ac83/32.png) [@Harper](https://fortran-lang.discourse.group/u/Harper)\ **Post date:** [April 6, 2025, 11:48pm UTC](https://fortran-lang.discourse.group/t/length-s-of-allocatable-character-array/9547/11 "2025-04-06T23:48:58Z")
My problem with 10.2.1.3 was due to the declaration ```auto character(:),allocatable::a,b,words(:) ``` where the second `(:)` says `words` is an allocatable array but the first `(:)` says the elements have an allocatable length. ---
**Author:** ![sblionel](https://yyz2.discourse-cdn.com/free1/user_avatar/fortran-lang.discourse.group/sblionel/32/853_2.png) [@sblionel](https://fortran-lang.discourse.group/u/sblionel)\ **Post date:** [April 7, 2025, 12:16am UTC](https://fortran-lang.discourse.group/t/length-s-of-allocatable-character-array/9547/12 "2025-04-07T00:16:01Z")
I don’t see the problem with this declaration. You have two allocatable scalars of deferred length, and one allocatable array of deferred length. Can you elaborate? Maybe I’m overlooking something. ---
**Author:** ![Harper](https://avatars.discourse-cdn.com/v4/letter/h/b5ac83/32.png) [@Harper](https://fortran-lang.discourse.group/u/Harper)\ **Post date:** [April 7, 2025, 4:06am UTC](https://fortran-lang.discourse.group/t/length-s-of-allocatable-character-array/9547/13 "2025-04-07T04:06:36Z")
There was indeed no problem with that declaration. My mistake was thinking of the two `(:)` as indicating the same sort of allocatability. The difference between deferred length and deferred shape became clearer when gfortran and ifx reported that one of these declarations was not valid: ```auto character(:), allocatable:: x(2) character(2), allocatable:: y(:) ```