Unstructured grid files axis by uramirez8707 · Pull Request #1114 · NOAA-GFDL/FMS

uramirez8707 · 2023-01-23T15:39:07Z

Description
This is an update for the unstructured grid files.

Axis that are in the "unstructured" domain require a "compress" attribute for the combiner and PP. This attribute is passed in via a diag_axis_add_attribute call in the model code. The compress attribute indicates the names of the axis that were compressed for example grid_index:compress = "grid_yt grid_xt".

This PR:

adds the metadata and the data for the axis structured axis in a unstructured grid file
addresses a gnu compiler bug where the size of the axis_data was equal to 2 after it was passed in fms_diag_axis_init

Fixes # (issue)

How Has This Been Tested?
CI

Checklist:

My code follows the style guidelines of this project
I have performed a self-review of my own code
I have commented my code, particularly in hard-to-understand areas
I have made corresponding changes to the documentation
My changes generate no new warnings
Any dependent changes have been merged and published in downstream modules
New check tests, if applicable, are included
make distcheck passes

uramirez8707 · 2023-01-23T15:44:07Z

+      !TODO Passing in the axis_length because of a gnu issue where inside fms_diag_axis_init, the size of DATA
+      !was 2 which was causing the axis_data to not be written correctly...
      diag_axis_init = fms_diag_object%fms_diag_axis_init(name, DATA, units, cart_name, long_name=long_name,&
       & direction=direction, set_name=set_name, edges=edges, Domain=Domain, Domain2=Domain2, DomainU=DomainU, &
-       & aux=aux, req=req, tile_count=tile_count, domain_position=domain_position )
+       & aux=aux, req=req, tile_count=tile_count, domain_position=domain_position, axis_length=size(DATA(:)) )


@rem1776 @thomas-robinson FYI i saw this with gnu 9.3. Let me know if you have a better fix for this.

If I were to print the size(DATA(:)) here it gave me the correct answers, but inside fms_diag_axis_init it was equal to 2

We discussed this, and it sounded good when we discussed it.

thomas-robinson · 2023-01-24T18:11:11Z

+      !TODO Passing in the axis_length because of a gnu issue where inside fms_diag_axis_init, the size of DATA
+      !was 2 which was causing the axis_data to not be written correctly...
      diag_axis_init = fms_diag_object%fms_diag_axis_init(name, DATA, units, cart_name, long_name=long_name,&
       & direction=direction, set_name=set_name, edges=edges, Domain=Domain, Domain2=Domain2, DomainU=DomainU, &
-       & aux=aux, req=req, tile_count=tile_count, domain_position=domain_position )
+       & aux=aux, req=req, tile_count=tile_count, domain_position=domain_position, axis_length=size(DATA(:)) )


We discussed this, and it sounded good when we discussed it.

thomas-robinson · 2023-01-25T18:09:27Z

+      !TODO Passing in the axis_length because of a gnu issue where inside fms_diag_axis_init, the size of DATA
+      !was 2 which was causing the axis_data to not be written correctly...


If this is a bug with a SPECIFIC gnu version, do we need to fix it or can we just say it's not compatible with that version? Is this a C5, CI, or dev box compiler?

The version of gnu is what we used in the CI
I can also reproduce in the dev box.

thomas-robinson · 2023-01-25T18:16:38Z

     TYPE(fmsDiagAttribute_type),allocatable , private :: attributes(:) !< Array to hold user definable attributes
     INTEGER                        , private :: num_attributes  !< Number of defined attibutes
     INTEGER                        , private :: domain_position !< The position in the doman (NORTH, EAST or CENTER)
+     integer                        , private :: structured_ids(2) !< If the axis is in the unstructured grid,


these are the mappings from unstructured to structured, correct?

these are not mappings, they are the axis IDs of the original structured axis

thomas-robinson · 2023-01-25T18:19:08Z

-      allocate(real(kind=r4_kind) :: this%axis_data(size(axis_data)))
+      allocate(real(kind=r4_kind) :: this%axis_data(axis_length))
      this%axis_data = axis_data
+      this%length = axis_length


This is optional. What happens if it's not included?

Changed it to a required argument fms_diag_axis_init is only used locally here:

FMS/diag_manager/diag_axis.F90

Lines 143 to 145 in cf55696

diag_axis_init = fms_diag_object%fms_diag_axis_init(name, DATA, units, cart_name, size(DATA(:)), &

& long_name=long_name, direction=direction, set_name=set_name, edges=edges, Domain=Domain, Domain2=Domain2, &

& DomainU=DomainU, aux=aux, req=req, tile_count=tile_count, domain_position=domain_position)

thomas-robinson · 2023-01-25T18:27:25Z


    this%nsubaxis = 0
    this%num_attributes = 0
+    this%structured_ids = diag_null


Can this be allocatable instead of used for every axis? Or should we make a new type that extends the fmsDiagFullAxis_type

Changed this to allocatables

thomas-robinson · 2023-01-25T18:28:37Z

+
+    select type (this)
+    type is (fmsDiagFullAxis_type)
+      this%structured_ids = axis_ids


this%structured_ids should be allocated here and then set to axis_ids.

thomas-robinson · 2023-01-25T18:29:49Z

+    class(*), intent(in) :: compress_att(:) !< The compress attribute to parse
+    character(len=120)   :: axis_names(2)
+
+    integer            :: ios           !< Errorcode after parsting the compress attribute


should say "after parsing"

thomas-robinson · 2023-01-25T18:31:18Z

+  integer                              :: i,k            !< For do loops
  integer                              :: j              !< diag_file%axis_ids(i) (for less typing)


confusing...

switched this to a pointer

axis_ptr => diag_axis(diag_file%axis_ids(i))

… structured_axis to allocatables

thomas-robinson · 2023-01-31T12:21:50Z

 !> @return Axis id
-FUNCTION fms_diag_axis_init(this, axis_name, axis_data, units, cart_name, long_name, direction,&
-  & set_name, edges, Domain, Domain2, DomainU, aux, req, tile_count, domain_position, axis_length ) &
+FUNCTION fms_diag_axis_init(this, axis_name, axis_data, units, cart_name, axis_length, long_name, direction,&


Why change the interface?

This is now a required arugment, so the interface changed.

* adds the compress attribute and axis data to the netcdf file * Fix some typos, change axis_length to a required argument, change the structured_axis to allocatables

…GFDL#1114)

uramirez8707 added 4 commits January 20, 2023 13:36

adds the compress attribute and axis data to the netcdf file

deb196e

Renames the variables and subroutines

9596033

Attempt to fix gnu shenanigans

fab037c

Fix the gnu error ...

1216773

uramirez8707 requested a review from thomas-robinson as a code owner January 23, 2023 15:39

uramirez8707 commented Jan 23, 2023

View reviewed changes

thomas-robinson requested changes Jan 25, 2023

View reviewed changes

Fix some typos, change axis_length to a required argument, change the…

cf55696

… structured_axis to allocatables

thomas-robinson reviewed Jan 31, 2023

View reviewed changes

Merge branch 'dmUpdate' into compress_axis

c39c5cf

thomas-robinson approved these changes Jan 31, 2023

View reviewed changes

fix mistake when fixing the merge conflicts

28c666b

thomas-robinson merged commit 34ef885 into NOAA-GFDL:dmUpdate Feb 1, 2023

rem1776 pushed a commit to rem1776/FMS that referenced this pull request May 1, 2024

feat: modern diag add support for unstructured grid files axis (NOAA-…

4c87fc7

…GFDL#1114)

rem1776 pushed a commit to rem1776/FMS that referenced this pull request May 1, 2024

feat: modern diag add support for unstructured grid files axis (NOAA-…

130b232

…GFDL#1114)

rem1776 pushed a commit to rem1776/FMS that referenced this pull request May 1, 2024

feat: modern diag add support for unstructured grid files axis (NOAA-…

552ae01

…GFDL#1114)

		!TODO Passing in the axis_length because of a gnu issue where inside fms_diag_axis_init, the size of DATA
		!was 2 which was causing the axis_data to not be written correctly...

	diag_axis_init = fms_diag_object%fms_diag_axis_init(name, DATA, units, cart_name, size(DATA(:)), &
	& long_name=long_name, direction=direction, set_name=set_name, edges=edges, Domain=Domain, Domain2=Domain2, &
	& DomainU=DomainU, aux=aux, req=req, tile_count=tile_count, domain_position=domain_position)

		integer :: i,k !< For do loops
		integer :: j !< diag_file%axis_ids(i) (for less typing)

Conversation

uramirez8707 commented Jan 23, 2023

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants