Global mean time series plotting script #334
Conversation
| # reference time series (DataArray) | ||
| ref_ts_da = adfobj.data.load_reference_timeseries_da(field) | ||
| # check if this is a "2-d" varaible: | ||
| has_lat_ref, has_lev_ref = pf.zm_validate_dims(ref_ts_da) |
There was a problem hiding this comment.
If the field doesn't exist (for some reason), the adfobj.data.load_reference_timeseries_da(field) will return a Nonetype, and
has_lat_ref, has_lev_ref = pf.zm_validate_dims(ref_ts_da)
fails, and ADF crashes.
I added
ref_ts_da = adfobj.data.load_reference_timeseries_da(field)
if ref_ts_da is None:
print(
f"\t Variable named {field} provides Nonetype. Skipping this variable"
)
continue
# check if this is a "2-d" varaible:
has_lat_ref, has_lev_ref = pf.zm_validate_dims(ref_ts_da)
which seems to work fine
There was a problem hiding this comment.
@adagj You're right, thanks for finding this! I can confirm this fix works
Additionally, if the ADF is comparing against obs, it will also trigger the Nonetype.
@brianpm We can just add this check and not have the time series plot made if comparing against obs, or I have coded up a quick fix to allow the time series plots to be created in the model vs obs case, but will just plot the test time series. I'm happy to share if this desired.
There was a problem hiding this comment.
@justin-richling -- I think your solution to keep making the plot (just skip the reference) is the desired behavior. Can you incorporate your fix and merge it in?
There was a problem hiding this comment.
@brianpm Will do!
| @@ -684,7 +684,7 @@ def jinja_list(seas_list): | |||
|
|
|||
| #List of ADF default plot types | |||
| avail_plot_types = res["default_ptypes"] | |||
|
|
|||
| #Check if current plot type is in ADF default. | |||
| #If not, add it so the index.html file can include it | |||
| for ptype in plot_types.keys(): | |||
There was a problem hiding this comment.
There was a problem causing ADF to crash in the for loop starting on line 687 in adf_web.py
I needed to change
for ptype in plot_types.keys():
if ptype not in avail_plot_types:
avail_plot_types.append(plot_types)
to
for ptype in plot_types.keys():
if ptype not in avail_plot_types:
avail_plot_types.append(ptype)
There was a problem hiding this comment.
@adagj Thanks for catching this! Although this should've caught the TimeSeries plot type since it is included in lib/adf_variable_defaults.yaml. I'm curious why this crashed for you, would you be able to provide any further details when this crashed? Thanks!
Remove testing print statements
This will be important when variables that aren't in the ref case
Fix bug when variable doesn't exist in reference files and allow for model vs obs to continue, just plotting the test case.
Add check for variable unit from variable defaults yaml in case none are listed in time series file. Also update check for `NoneType` for data arrays
Add a script to plot global mean time series (annual mean only right now).
Uses time series files directly. To do this, I added some functionality to
adf_dataset.pyto get time series data.Adds a TimeSeries type of plot category to
adf_variable_defaults.yaml.Modifies
plotting_functions.pyto try to keep units in tact while doing spatial and annual averages.Changes in
adf_info.pyis incidental (two print statements), and can be ignored.The change in
adf_web.pyis literally whitespace and can be ignored.