| Re: Returning a struct containing variable-length arrays [message #62521] |
Wed, 17 September 2008 14:01 |
franzpx125
Messages: 6
Registered: September 2008
|
Junior Member |
|
|
On 17 Set, 10:46, Nigel Wade <n...@ion.le.ac.uk> wrote:
> franzpx125 wrote:
>> Hi!
>
>> Within my DLM, I need to return a struct containing two arrays but the
>> length of the arrays is known only at run-time. Which is the correct
>> way to write the DLM code? I think I have to use IDL_MAX_ARRAY_DIM and
>> I defined:
>
>> static IDL_MEMINT my_tags_dims[] = { 1, IDL_MAX_ARRAY_DIM };
>
>> static IDL_STRUCT_TAG_DEF my_tags[] = {
>> { "X", my_tags_dims, (void *) IDL_TYP_DOUBLE },
>> { "Y", my_tags_dims, (void *) IDL_TYP_DOUBLE },
>> { 0 }
>> };
>
>> but I don't know how to fill the fields of this struct. Any help?
>
>> Thanks,
>> Brun Francesco
>
> You basically have two choices depending on whether you want to use your own
> memory storage allocated by your DLM code to hold the structure data, or have
> IDL allocate its own memory and copy the values into that. Each method has its
> own set of advantages and caveats/gotchas.
>
> If you use your own memory allocation then that memory has to persist after the
> DLM routine returns. This either means it has to be declared static (but since
> you want to use variable length arrays this isn't possible) or allocated by
> malloc. Using this method means that you are responsible for freeing the memory
> when IDL has finished with it. If it is simple throw away code which will only
> be used occasionally, and the function isn't called many times, and doesn't
> allocate much memory, you can ignore this but you *will* have a memory leak. To
> cure this you need to implement the callback routine, a function pointer to
> this is passed into the IDL_ImportArray function so that it can free whatever
> memory you allocate.
>
> If you create an IDL struct using your own memory then you don't have to worry
> about copying the data into the structure. If you use IDL_MakeStruct so that
> IDL allocates the storage (and handles de-allocation) then you do have to copy
> the data into that memory. This is a little tricky.
>
> In the first case, allocating your own storage, the sequence is to first
> determine how much storage is required. Then you set your dimensions arrays
> appropriately. Next you create the structure with IDL_MakeStruct and finally
> import your data into the structure using IDL_ImportArray. You are responsible
> for writing a callback function to free this memory when IDL has finished with
> the variable.
>
> To get IDL to handle the allocation the first two steps are the same. Then,
> rather than importing your own data you use IDL_MakeTempStruct to create a IDL
> temporary variable (which you can pass back to IDL). The tricky part is
> determining where to put your data. To do this you can use the function
> IDL_StructTagInfoByName, which returns the byte offset into the struct where
> the data for that tag is located. Then it's a case of some hairy pointer
> manipulation to write the data into the structure.
>
> --
> Nigel Wade
Thank you so much for your exhaustive explanation. As you can notice
from a previous post I adopted the second solution and although
results in slower code with greater memory occupation, I think it's
cleaner without a callback function and more coherent with other
routines included in my DLM. Thanks again.
Brun Francesco
|
|
|
|
| Re: Returning a struct containing variable-length arrays [message #62527 is a reply to message #62521] |
Wed, 17 September 2008 01:46  |
Nigel Wade
Messages: 286
Registered: March 1998
|
Senior Member |
|
|
franzpx125 wrote:
> Hi!
>
> Within my DLM, I need to return a struct containing two arrays but the
> length of the arrays is known only at run-time. Which is the correct
> way to write the DLM code? I think I have to use IDL_MAX_ARRAY_DIM and
> I defined:
>
> static IDL_MEMINT my_tags_dims[] = { 1, IDL_MAX_ARRAY_DIM };
>
> static IDL_STRUCT_TAG_DEF my_tags[] = {
> { "X", my_tags_dims, (void *) IDL_TYP_DOUBLE },
> { "Y", my_tags_dims, (void *) IDL_TYP_DOUBLE },
> { 0 }
> };
>
> but I don't know how to fill the fields of this struct. Any help?
>
> Thanks,
> Brun Francesco
You basically have two choices depending on whether you want to use your own
memory storage allocated by your DLM code to hold the structure data, or have
IDL allocate its own memory and copy the values into that. Each method has its
own set of advantages and caveats/gotchas.
If you use your own memory allocation then that memory has to persist after the
DLM routine returns. This either means it has to be declared static (but since
you want to use variable length arrays this isn't possible) or allocated by
malloc. Using this method means that you are responsible for freeing the memory
when IDL has finished with it. If it is simple throw away code which will only
be used occasionally, and the function isn't called many times, and doesn't
allocate much memory, you can ignore this but you *will* have a memory leak. To
cure this you need to implement the callback routine, a function pointer to
this is passed into the IDL_ImportArray function so that it can free whatever
memory you allocate.
If you create an IDL struct using your own memory then you don't have to worry
about copying the data into the structure. If you use IDL_MakeStruct so that
IDL allocates the storage (and handles de-allocation) then you do have to copy
the data into that memory. This is a little tricky.
In the first case, allocating your own storage, the sequence is to first
determine how much storage is required. Then you set your dimensions arrays
appropriately. Next you create the structure with IDL_MakeStruct and finally
import your data into the structure using IDL_ImportArray. You are responsible
for writing a callback function to free this memory when IDL has finished with
the variable.
To get IDL to handle the allocation the first two steps are the same. Then,
rather than importing your own data you use IDL_MakeTempStruct to create a IDL
temporary variable (which you can pass back to IDL). The tricky part is
determining where to put your data. To do this you can use the function
IDL_StructTagInfoByName, which returns the byte offset into the struct where
the data for that tag is located. Then it's a case of some hairy pointer
manipulation to write the data into the structure.
--
Nigel Wade
|
|
|
|
| Re: Returning a struct containing variable-length arrays [message #62528 is a reply to message #62527] |
Tue, 16 September 2008 14:11 |
franzpx125
Messages: 6
Registered: September 2008
|
Junior Member |
|
|
On 16 Set, 15:37, Mike <Michael.Mill...@gmail.com> wrote:
> On Sep 15, 6:28 pm, franzpx125 <franzpx...@gmail.com> wrote:
>
>> Hi!
>
>> Within my DLM, I need to return a struct containing two arrays but the
>> length of the arrays is known only at run-time. Which is the correct
>> way to write the DLM code?
>
> You can put avariable lengtharray into a structure by putting a
> pointer to the array in the structure:
>
> result = { x: x, y: y, ptr: ptr_new()}
> ...
> result.ptr = ptr_new(data)
> return, result
>
> I don't know if there are any DLM issues that will affect this, but it
> works well forvariable lengtharrays in objects and other structures.
>
Thanks for your reply. I don't know if it's the correct way to perform
this task, anyway I've tried the following workaround and it works.
I'll post the code if you're interested in:
IDL_VPTR idl_out_struct;
unsigned int* rev_x;
double* rev_y;
unsigned int num_el;
void* s;
IDL_LONG tmp_dims[IDL_MAX_ARRAY_DIM];
IDL_MEMINT offset;
double* y_ptr;
unsigned int* x_ptr;
char* s_data;
int i;
/* .... omitted code .... */
// Format struct for IDL output:
REV_tags_dims[0] = 1;
REV_tags_dims[1] = num_el;
s = IDL_MakeStruct("REV", REV_tags);
tmp_dims[0] = 1;
s_data = (char *)IDL_MakeTempStruct(s, 1, tmp_dims,
&idl_out_struct, 0);
// Get the field of the structure:
offset = IDL_StructTagInfoByName(s, "Y", IDL_MSG_LONGJMP, NULL);
// Get a pointer to that location:
y_ptr = (double *)(s_data + offset);
// Store values into array:
for ( i = 0; i < num_el; i++)
*(y_ptr++) = rev_y[i];
// Get the field of the structure:
offset = IDL_StructTagInfoByName(s, "X", IDL_MSG_LONGJMP, NULL);
// Get a pointer to that location:
x_ptr = (unsigned int*)(s_data + offset);
// Store values into array:
for ( i = 0; i < num_el; i++)
*(x_ptr++) = rev_x[i];
// Return output in IDL Format:
return idl_out_struct;
Any criticism is well accepted.
Regards,
Brun Francesco
|
|
|
|
| Re: Returning a struct containing variable-length arrays [message #62536 is a reply to message #62528] |
Tue, 16 September 2008 06:37 |
Mike[2]
Messages: 99
Registered: December 2005
|
Member |
|
|
On Sep 15, 6:28 pm, franzpx125 <franzpx...@gmail.com> wrote:
> Hi!
>
> Within my DLM, I need to return a struct containing two arrays but the
> length of the arrays is known only at run-time. Which is the correct
> way to write the DLM code?
You can put a variable length array into a structure by putting a
pointer to the array in the structure:
result = { x: x, y: y, ptr: ptr_new()}
...
result.ptr = ptr_new(data)
return, result
I don't know if there are any DLM issues that will affect this, but it
works well for variable length arrays in objects and other structures.
Mike
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
