VisIt scripts for plotting data from FVCOM output files
Instructions are on the GitHub README, or in the README.md after cloning the repo. To clone the repo, from Linux/Mac terminal or Windows git bash (and set your HOME directory to the new cloned repo) do:
git clone https://github.com/l3-hpc/visit-for-fvcom
cd visit-for-fvcom
In that directory, there is a file .gitignore. Files listed there will be ignored by git. It includes files that are likely to change between users, like setpaths.py, and also files that may be created in the directory but should not be saved, such as core files and SLURM/LSF output.
The paths to the input files and output directories are defined in the file setpaths.py.
The file setpaths.py will be different for each user, and for each user will be different from PC to HPC. To avoid having git track changes to all of our filepaths, a template setpaths.py.template is included in this git repo.
Copy the template and modify setpaths.py. The file setpaths.py is listed in the .gitignore and it will not be tracked or overwritten with new pulls.
cp setpaths.py.template setpaths.py
Modify the following in setpaths.py to point to the appropriate directory paths:
EPA_directory
COMPARE_directory
IMGS_DIR
If your netCDF files do not begin with mi_, then change these as well:
file_prefix_epa
file_prefix_compare
Note, if IMGS_DIR does not already exist, it will be created. Also, any existing files in the IMGS_DIR will be overwritten.
Do not modify any lines below the file declarations.
Set the parameters for various plots in file setparams.py. The commenting is decent, but there is no error checking.
The images are named after the type of plot and the time/date stamp and RUN_NAME, defined in setpaths.py, is appended. If you rerun the script for a different transect or colormap, etc., the old plots will be overwritten if you do not choose a different RUN_NAME. You may alternatively or additionally change the name of IMGS_DIR before a new run.
To print the paths and parameters for basic error checking, run test_params.py, which prints the value and type of each function. When adding variables to the code, please add them here as well.
python test_params.py
Sample submission scripts were created for Henry2 and atmos. Before modifying, make a local copy. For example,
cp submit.csh.atmos submit.csh
or
cp submit.csh.henry2 submit.csh
After copying, modify submit.csh to change the wall clock time and, for atmos, the run directory, which is the directory from where you submit the job. To submit the job, on atmos, use
sbatch submit.csh
and for Henry2, use
bsub < submit.csh
You can check the status of the batch job by
squeue -u <username>
The code that calls VisIt is plot_any.py. To run interactively, start an interactive session, load the visit module, and type the commands which are listed in the submit scripts but remove the 'nowin' argument:
[On HPC, start interactive session]
[On HPC, module load the visit module via 'module load visit']
visit -cli -s plot_any.py
For HPC, X11 forwarding is enabled with MobaXterm by default. If using a Mac, you may need to use 'ssh -X' and XQuartz. On Henry2, use the HPC-VCL.
After the images were created by VisIt, run the Python scripts. First, set the environment for Python 3 by loading an appropriate module or activating a Conda environment. For example:
module load intelpython3
A YAML file is provided for folks who want to make a Conda environment themselves. So far, the Python scripts use libraries that are available with any Python 3 module on atmos.
After loading the module or activating the Conda environment, run the Python script linkandmovie.py, which creates a movie of a single plot for each type of plot listed in the setparams.py. This script assumes you will be making movies of whatever was produced in your last run...i.e., it reads setpaths.py and setparams.py. Even if your goal is to make a movie with four images in one plot, this script must be run first, as the script that contatenates the images uses the symbolic links created in this script. To run, do
python linkandmovie.py
After loading the module or activating the Conda environment, run the Python script for concatenating 4 plots onto a single plot. The script multiplot.py currently makes plots for a comparison, with the top two plots the Layer=1 view and the bottom two plots are transects. In multi-layer.py, the top are Layer=1 and the bottom are Layer=19. Run by typing python and then the name of the script, e.g.,
python multiplot.py
Ideally, if you are on an HPC, do the command in a batch script or interactive session.
The script simply takes multiple image files and contatenate them into a single image, i.e. 4 PNGs in a 2x2 layout.
The scripts have the plot names hardcoded. Right now the looping index is also hardcoded, which is bad. It is assumed that you already ran the script linkandmovie to create Symbolic links of the images named with a regular expression and a formated integer index.
On a Mac from command line, I can look at images with open. On Henry2, with display. On atmos use 'display' to open Image Magick and navigate to the file you want to open(you may need to click on the window that opens to have the menu window pop up). Or navigate to the directory with the images on atmos and open from there.
I didn't put the movie command in the script that concatenates the images, so the call to ffmpeg is just hardcoded in the file movie-fourplots.py. The directory name, file names, number of images...all hardcoded for now. Sorry 'bout that.
To use, make a copy to another name, modify for your case, and then run:
python movie-fourplots.py
dir()
help()
ClearCache
a = GetAnnotationAttributes()
print(a)
Sometimes during extended VisIt runs, you might want to periodically clear the compute engine’s network cache to reduce the amount of memory being used by the compute engine
SetPipelineCachingMode
SetPipelineCachingMode(0) # Disable caching
Disabling caching is still grinding script to a halt on the Mac while on Linux it is fine. I suspect a memory leak.
If you are just using and not modifying the code, to get new changes, do:
git pull
Note, since setpaths.py is in the gitignore, it will not notice if there were changes made. If setpaths.py.template has changed, you may need to modify your setpaths.py. Best practice is to do this after doing a new git pull.
diff setpaths.py setpaths.py.template
If you changed or created files in that directory since you last got the code, it will give an error message on trying to git pull.
- If you added files to the directory, and need those files but don't need them specifically in that directory, you should move them.
- If they are junk files, you can remove them or do
git stash, which will remove anything that changed since you last didgit cloneorgit pull. If you expect to aways make those same named files in that directory, add the name of the file (or directory) to the .gitignore. - If you made changes and what to keep them, then you need to do fetch and merge. Instructions for Fetch and Merge will be added later.
To check if the changes you have are actually important, you can check what exactly was changed by doing
git status
git diff
YAML file to create a Conda environment for multiplots.py:
name: fvcom
channels:
- conda-forge
dependencies:
- matplotlib
- numpy
To create the environment:
conda env create --prefix /[your_path]/env_fvcom -f fvcom.yml
To use
conda activate /[your_path]/env_fvcom
Then run multiplot.py:
python multiplot.py
Even though these made nice plots on my Mac, the ones on Henry2 were huge. I think that has to do with the save window settings. (I assume the same will happen on atmos. So I need to fix that.)