Only apply add_time0 function if time0 if needed.

Round times to nearest millisecond. This fixes times like 2019-06-15 18:03:0.000003,
which is supposed to be 2019-06-15 18:03 but has imprecise floating-point value.

Allow metpy.quantify() to protect units
This eliminates the need to restore units after applying .mean to DataArray.

Round times to nearest millisecond. Model output times are sometimes
imprecise due to floating point error.

Let argparse assume arguments are strings unless specified otherwise.

Allow time window to be a float, not just an int.

Refer to sum of all tendencies as "all_tendencies" instead of "total".

"dstate_variable" ->  "actual_tendency"

Don't treat nophys tendency differently. UFS output changed.
Removed bug where final units are assumed to be temperature tendency "K/hour".
Keep the tendencies in their native units of "[something] per second" instead
of converting to "[something] per hour".

Define time slice starting with time-after-time0 and ending with validtime.
We use the time *after* time0 because the time range of the tendency field is the period
immediately prior to the tendency timestamp.
That way, slice(time_after_time0, validtime) has a time range of [time0,validtime].
Compute the time-weighted mean tendencies for time index slice.

Be clear about "all" tendencies. It is the sum of all tendencies,
physics and non-physics. It does not include the
physics tendency variable, which may be part of the history file and
included in the tendency_vars list. That would be double-counting.

Subtract physics tendency variable if it was in tendency_vars.

Command line changes:
 Read file-type command line args as strings, not argparse.FileType('r').
 For simpler code.

 --norobust flag (to use data extremes for colormap range, not 99th
 and 2nd percentiles.)

logging info about renaming variables.
 Keep characters after final underscore. The first part is redundant.
 For example dtend_u_pbl -> pbl

actual_change -> actual_tendency

Test for shp is not None, not shp.
(Accept shp=something evaluated as False, like 0)

Print entire arg Namespace as fineprint, not just select variables.
Use plt.figtext instead of plt.annotate. Makes for cleaner code.

Reduce xarray time selection tolerance from 1 sec to 1 millisecond.

pep8 changes

reorder load modules

module and function documentation

shorten lines

lazy logging messages (no f-string)

specify encoding on open

Fix names of tendency variables in config file. The
state variables are not the same as the tendency variable
prefixes. For example spfh vs. qv.

new variable names

add table for derived tendencies

note about time window starting at initialization time and dealing
with it by appending state variables at time0.

equation for residual in description

new date for new history file in Users Guide

new figures for Users Guide

fill -> tendencytype

simplify construction of da2plot in planview_fv3.py

don't allow user to plot state variable at validtime or time0.
it makes code so much uglier for such a simple request.

No explicit debug option for physics_tend.pts_in_shp

Restore statevariable in default output filename in planview_fv3.py.

100 dpi default instead of 150

…m/ahijevyc/METplotpy into feature_380_phys_tend_plot_actual

fix link to cross section png

…m/ahijevyc/METplotpy into feature_380_phys_tend_plot_actual

Remove time_varname from yaml config file.
Assert time0 exists, and if not, help the user by listing the
nearest time available.

Make longitude the xaxis of cross section instead of 'index'.

Dequantify *after* cross section to preserve units.
Specify lat lon order of endpoints in cross_section_vert.py.

Remove dindex argument. xarray does a good job of choosing the
major axis interval.

Note vertical interpolation method in planview_fv3.py and reveal
actual pressure level in title instead of requested level.

only pull out variables you need

Use fewer times and change cross section example to use new
time range.

extract ugrd in cross section code block, not tmp.

include --resid option in --subtract example

…feature_380_phys_tend_plot_actual

…m/dtcenter/METplotpy into feature_380_phys_tend_plot_actual

…ure default name like "stack-23412h4234kj"

avoids error about Fancy indexing not implemented yet.

remove total area information. you'd have to mask it too, which
makes the code ugly.

And correct order of nophys and phys.

Dequantify before cross_section to avoid NotImplementedError: Don't yet support
 nd fancy indexing from cross_section(). something to do with dask and units.

cartopy changed the default params of the LambertConformal class from
`standard_parallels = None` to `standard_parallels = (33, 45)` between v0.20.2 and v0.22.

I fixed the problem by removing `standard_parallels=None`.

to avoid TypeError: The dtype and signature arguments to ufuncs only select the general DType and not details such as the byte order or time unit. You can avoid this error by using the scalar types np.float64 or the dtype string notation.

xarray 2022.3.0 was okay with how I handled xarray DataArrays with units (pint quantities), but xarray 2023.5.0 is not okay.

…feature_380_phys_tend_plot_actual