Sequential plot display with unified postage stamp plots #379
Conversation
jfrost-mo
force-pushed
the
307_sequential_plot_display
branch
from
e771950
to
ad263ef
Compare
|
|
jfrost-mo
force-pushed
the
307_sequential_plot_display
branch
2 times, most recently
from
e08729e
to
8b9a1f1
Compare
jfrost-mo
force-pushed
the
307_sequential_plot_display
branch
from
b26c010
to
ac582e3
Compare
This comment was marked as outdated.
This comment was marked as outdated.
jfrost-mo
force-pushed
the
307_sequential_plot_display
branch
from
ac582e3
to
f3a0cd4
Compare
|
Do we want the postage stamp plotting code merged into the same operator, so there is just a single spacial plot operator? I think we want to do that, but in a second PR, as this one is already quite big. Issue #381 exists for this. EDIT: I did it anyway, and it made more sense in the one PR. |
jfrost-mo
force-pushed
the
307_sequential_plot_display
branch
from
f3a0cd4
to
ebf284c
Compare
jfrost-mo
force-pushed
the
307_sequential_plot_display
branch
2 times, most recently
from
93e4902
to
dcb840d
Compare
cset-workflow/app/install_website_skeleton/file/html/status.html
Outdated
Show resolved
Hide resolved
src/CSET/recipes/surface_air_temperature_spatial_plot_time_series.yaml
Outdated
Show resolved
Hide resolved
cset-workflow/app/install_website_skeleton/file/html/status.html
Outdated
Show resolved
Hide resolved
src/CSET/recipes/surface_air_temperature_spatial_plot_time_series.yaml
Outdated
Show resolved
Hide resolved
jfrost-mo
force-pushed
the
307_sequential_plot_display
branch
from
ef16d52
to
157a8ce
Compare
jfrost-mo
changed the title
jfrost-mo
changed the title
There was a problem hiding this comment.
Changes look good, though note that I have not been able to test them here. However, I am holding approval pending a discussion on creation additional external dependencies (in this case with simple_template). In my opinion, it is better for portability if these are as few as possible, even if they are also in-house tools; otherwise they add to the maintenance burden, and can cause problems a few years down the line.
A recent example has been the recent need to upgrade TRUI to python 3, to be able to upgrade the RES; I don't know if the RES was the only tool that depended on TRUI, but it was contemplated for some time to upgrade the RES with reduced capability if TRUI was not made available under python3.
Having said that. I am on the fence on this one.
SVG plots were too slow to render when large. Fixes #369
The previous code was catching other ValueErrors we wanted to see the stacktrace for.
importlib.resources.files() works on the current module by default.
The viewer page now adapts to handle both plot sequences, and single plots to keep the generation code simple. For now I have used a simple templating library, rather than pulling in Jinja, but it is syntax compatible if we find we need more templating in future. Also moved template into a seperate file. The plot is constructed from the figure to prevent duplicated colourbar. Fixes #307 Template plots into output page Needed to get around CORS, but also beneficial for removing a network round trip. Make time series recipe more explicit
Its behaviour changed in 3.12.
This allows them to be faster, and more effectively pinpoint problems.
This makes the plotting interface much easier to use, as you no longer need to worry about the shape of your data. Reasonable defaults have been chosen for the postage stamp and sequence coordinates (realization and time respectively) so it will work as expected most of the time. The existing postage_stamp_contour_plot operator has been marked as deprecated, but retained, as there are recipes that use it. Fixes #381 The tests have also been modified: - Continue to test deprecated operator - Copy test cube before modifying to prevent other tests being affected - Assert that plots contains some plots - Make stamp_coordinate optional - Test plotting without realization coord
Series makes one think of a line plot, rather than multiple plots.
jfrost-mo
force-pushed
the
307_sequential_plot_display
branch
from
4a497bf
to
7dbb3c0
Compare
|
For the web page templating I was initially thinking of using jinja, as that is the definitive python templating library, and would be familiar to people. But that is quite heavy a dependency, and also might give more power than we want by allowing imperitive logic to be put into recipes. simple_template is something I've actually extracted from some of my personal code, and is a single file pure python templater that just does the variable insertion aspect of jinja. As it is a single file with no dependencies, and is licenced under 0BSD, we could copy it into CSET entirely and remove the external dependency. |
|
Also are you still having trouble running CSET or is it just time constraints? |
As it is a small dependency only needing the standard library, and is liberally licenced (0BSD) we can just integrate it directly.
|
@JorgeBornemann I've now (in b85c034) copied the simple_template module's code inside CSET, and removed the external dependency. |
Example of output: https://tmpweb.net/k7pKV1n2XtNq/ (Valid until 2024-02-23)
Fixes #307.