Refactor Chem_MieTableMod2G.F90

[tclune]

No longer needed as we do not aggregate across tracers.

[weiyuan-jiang]

Where to save channels(:) after deleting Chem_Mie type?

[tclune]

There is a laundry list of changes needed:

• The module name should change
  • "2G" should be before "Mod"
  • "Table" should go away
  • Filename should be updated accordingly (git mv)
• Chem_Mie type goes away
• Chem_MieTable is renamed to Chem_Mie
• Use modern Fortran
  • POINTER components should be ALLOCATABLE
  • Chem_MieQuery should be a type-bound procedure
  • Constructor should be the same name as the type (requires interface block)
• "nlambda" component should be "nch" and documented as number of channels
• Create and Read methods should be fused into a single constructor
• QueryByInt should be QueryByScalar
• WithPmom should be QueryByVector
• New interface should allow passing frequency instead of channel
  • need new helper procedure that computes channel (int) from frequency (float)
  • just use placeholder for now

[tclune]

I also want to temporarily write another interface that works with 3D arrays that is not elemental. If ELEMENTAL gives comparable performance, we delete the 3d version (and leave a comment that elemental was benchmarked) Otherwise we eliminate ELEMENTAL and keep both versions and also create a 1D version that is needed outside of GEOS.

[weiyuan-jiang]

In the elemental procedure, I suspect it creates 3d channels and idx which are not necessary when 3d is called

[weiyuan-jiang]

• WithPmom should be QueryByVector
  Comparing Withpmom with byInt, there is one more 2d optional argument pmom(:,:) in Withpmom. I saw the comment that this subroutine may be deleted.

[weiyuan-jiang]

• "nlambda" component should be "nch" and documented as number of channels

Should we change lambda(:) to channels(:) in the component too?

[tclune]

ELEMENTAL only "creates arrays" for the arguments that are arrays. The "THIS" object remains a scalar, as does idx. Maybe others - have not looked.

• WithPmom should be QueryByVector
  Comparing Withpmom with byInt, there is one more 2d optional argument pmom(:,:) in Withpmom. I saw the comment that this subroutine may be deleted.

No - pmom is a different shape and thus prevents ELEMENTAL. I pulled it into a separate interface precisely because of that.

[tclune]

• "nlambda" component should be "nch" and documented as number of channels

Should we change lambda(:) to channels(:) in the component too?

Actually, it should be changed to wavelengths(:). Channel is the integer, wavelength is the float.

[amdasilva]

It has been done with PR on 3/17/2022.