The generate
command is used to interop with third-party
tools that read tree-structured configuratoin data from a supported
format.
Consider a fictional tool named acme
. This tool reads
configuration files in JSON format.
acme
import sys
import json
import pathlib
= pathlib.Path(sys.argv[1])
config_file
print(f"Running simulation described in {config_file}.")
= json.loads(config_file.read_text())
config
print(f"Writing results to {config['simulation']['output_file']}")
'simulation']['output_file']).write_text("done")
pathlib.Path(config[
print("Done")
A typical configuration file looks something like this:
import sys
import json
import pathlib
config_file = pathlib.Path(sys.argv[1])
print(f"Running simulation described in {config_file}.")
config = json.loads(config_file.read_text())
print(f"Writing results to {config['simulation']['output_file']}")
pathlib.Path(config['simulation']['output_file']).write_text("done")
print("Done")
and we run it like this
$ python acme ACME-solver.json
$ ls
We would like to powerup our acme
config file to add
unit support. We start by writing a powerconf configuration file. We are
free to structure the configuration however we want, config parameters
can be given as quantities with units, and we can use expressions that
calculate the value of some parameters based on the value of others.
POWERCONFIG.yml
simulation:
grid:
res: 1 um
x:
min: 0 um
max: 100 um
n: $(int(${max}/${../res}) + 1)
y:
min: 0 um
max: 1 cm
n: $(int(${max}/${../res}) + 1)
We can test that this configuration works with the
print-instances
command:
$ powerconf print-instances
simulation:
grid:
res: 1 micrometer
x:
max: 100 micrometer
min: 0 micrometer
n: 101
y:
max: 1 centimeter
min: 0 micrometer
n: 10001
Next, we add a section to our configuration file for the acme tool. We replicate the acme config file tree computing the value for each parameter using expressions that reference the main config. The purpose of this section is to unit convert all parameters to the unit expected by the acme tool and strip the units.
POWERCONFIG.yml
simulation:
grid:
res: 1 um
x:
min: 0 um
max: 100 um
n: $(int(${max}/${../res}) + 1)
y:
min: 0 um
max: 1 cm
n: $(int(${max}/${../res}) + 1)
# parameters that will be inserted diredctly into ACME config file.
# these parameters _MUST_ be given.
# all quantities need to be converted to ACME internal units (cgs)
# and turned into plain numbers.
acme:
simulation:
output_file: acme-ouput-$(${../grid/x/n}).txt
grid:
x:
n: $($/simulation/grid/x/n)
min: $($/simulation/grid/x/min.to('cm').magnitude)
max: $($/simulation/grid/x/max.to('cm').magnitude)
y:
n: $($/simulation/grid/y/n)
min: $($/simulation/grid/y/min.to('cm').magnitude)
max: $($/simulation/grid/y/max.to('cm').magnitude)
Now we can generate an acme config file
$ powerconf generate POWERCONFIG.yml ACME-solver.json.generated -n acme -f json
$ ls
acme
ACME-simulation.json.generated
POWERCONFIG.yml
The -n
option takes the name of a node. The
generate
command will write the tree below the node to the
output file. The -f
option specifies the output file
format. The generated JSON file:
{
"simulation": {
"output_file": "acme-ouput-101.txt"
},
"grid": {
"x": {
"n": 101,
"min": 0,
"max": 0.009999999999999998
},
"y": {
"n": 10001,
"min": 0,
"max": 1
}
}
}
Running acme
:
$ python acme ACME-solver.json.generated
$ ls
acme
ACME-simulation.json.generated
POWERCONFIG.yml
An there you are, full-blown unit support for a tool that does not support them.