grid not identical to model.grid

[francispoulin]

@ali-ramadhan and I discovered today in developing fjp/flat-for-shallow-water-model that grid is set up with default halos but model.grid actually has the correct halos needed for the advection scheme. Note that the advection scheme is not specified until after the grid is created so given the current framework I think this is necessary. I realize this might be less than ideal but as long as you are focusing on the physcial domain, the two agree exactly.

Is this a concern?

[glwagner]

I think we should change the defaults as suggested by #1245 so this is uncommon.

[francispoulin]

Good idea.

Following the discussion, and the fact that people seem to be using high order advection schemes like WENO5, I changed the defalt in this branch from 1 to 3. This can be revisited when it becomes a PR so I will close this issue.

[ali-ramadhan]

Changing the default would make inflating halos less common. I still think it would be useful to print a warning/info when halos are inflated so the user knows grid and model.grid are different now.

[francispoulin]

I'm happy to put a warning if people agree. What would we say exactly and where would we say it?

[ali-ramadhan]

How about this?

@warn "Inflating model grid halo size to ($Hx, $Hy, $Hz) and recreating grid. " *
      "The model grid will be different from the input grid. To avoid this warning, " *
      "pass halo=($Hx, $Hy, $Hz) when constructing the grid."

[francispoulin]

How about this?

@warn "Inflating model grid halo size to ($Hx, $Hy, $Hz) and recreating grid. " *
      "The model grid will be different from the input grid. To avoid this warning, " *
      "pass halo=($Hx, $Hy, $Hz) when constructing the grid."

Sounds great! I have added it in my branch. I will close this issue and we cna modify this when the branch makes its way as a PR. Soon I hope!