Tutorials
These tutorials are written to get new users up and running with pandss
. It might be neccesary for users to understand some DSS concepts to fully understand what they are doing with pandss
. Where appropriate, USACE HEC-DSS documentation is linked.
Installation
Conda is the preferred package manager for pandss
due to certain preferred dependencies. In the future pandss
may be distributed via PyPI, but not as of now.
Conda
conda install pandss -c dwr-cvm
Basic Usage
The library is mostly used to interact with existing DSS files, but it is possible to create DSS files from scratch.
import pandss as pdss
# Specify the file to be created
file = "new.dss"
# The data to be written to file
data = dict(
path="/DOCS/MONTH_DAYS/TUTORIALS//1MON/2024/",
values=(31, 28, 31),
dates=("1921-01-31", "1921-02-28", "1921-03-31"),
period_type="PER-CUM",
units="days",
interval="1MON",
)
# Create the RegularTimeseries object
rts = pdss.RegularTimeseries.from_json(data)
# Write the data to a file
with pdss.DSS(file) as dss:
dss.write_rts(rts.path, rts)
Usually, you will have a DSS file handy so you will not need to create the RegularTimeseries objects from scratch.
import pandss as pdss
rts = pdss.read_rts("existing.dss", "/DOCS/MONTH_DAYS/TUTORIALS//1MON/2024/")
Once you have the data read, you can convert it to a pandas.DataFrame
and use your typical data analysis tools.
df = rts.to_frame() # Convert to DataFrame