UpsilonConf is a simple configuration library written in Python.
It might not be really obvious, but this library is inspired by the great OmegaConf library.
OmegaConf is also the backbone for the more advanced Hydra framework.
Concretely, the idea of this library is to provide an alternative to OmegaConf without the overhead of the variable interpolation (especially the antlr dependency).
It is also very similar to the (discontinued) AttrDict library.
In the meantime, there is also the ml_collections library, which seems to build on similar ideas as this project.
Nevertheless, I decided to release upsilonconf because there might be a few features that people might find interesting/useful:
- dict-like configuration object with attribute access (cf.
attrdict) - hierarchical indexing by means of tuples or dot-strings (cf.
omegaconf) - overwriting protection to prevent accidents
- read from/write to various file formats
- use hierarchical configs with options (cf. config groups in
hydra) - retrieve and manipulate config using CLI (cf.
omegaconf) - minimal dependencies (cf.
attrdict)
The name is inspired by OmegaConf. I decided to go for the Greek letter Upsilon because it is the first letter of ὑπέρ (hupér). This again comes from the fact that this library should mainly help me with managing hyper-parameters in neural networks.
import upsilonconfconf1 = upsilonconf.load("config.yaml") # from config file
conf2 = upsilonconf.Configuration(key1="value1", key2=2) # direct
dictionary = {"sub": conf2} # sub-configs allowed!
conf3 = upsilonconf.Configuration(**dictionary) # from dict
conf = conf1 | conf2 | conf3 # from other configurations# getters
conf["key1"] == conf.key1
conf.key2 == conf["sub", "key2"]
conf["sub", "key1"] == conf["sub.key1"]
conf["sub.key2"] == conf.sub.key2
# setters
conf["new_key"] = "new_value"
conf.other_key = "other_value"
conf.sub2 = {"a": .1, "b": 2}
conf["sub2", "c"] = 3.
conf["sub2.d"] = -4
# and deleters...
del conf["sub2.c"]try:
conf["key1"] = "overwrite1"
except ValueError:
print("overwriting")
conf.overwrite("key1", "overwrite1")
try:
conf.key1 = "overwrite2"
except AttributeError:
print("overwriting")
conf.overwrite("key1", "overwrite2")
try:
conf.update(key1="overwrite3")
except ValueError:
print("overwriting")
conf.overwrite_all(key1="overwrite3")# different file formats (with optional requirements)
conf = upsilonconf.load("config.yaml") # with patched float parsing
upsilonconf.save(conf, "config.json") # with indentation by default
# organise hierarchical configs in directories
upsilonconf.save({"key": "option1"}, "config_dir/config.json")
upsilonconf.save({"foo": 1, "bar": 2}, "config_dir/key/option1.json")
upsilonconf.save({"foo": 2, "baz": 3}, "config_dir/key/option2.json")
# load arbitrary parts of hierarchy
conf = upsilonconf.load("config_dir/key")
conf == upsilonconf.Configuration(
option1={"foo": 1, "bar": 2},
option2={"foo": 2, "baz": 3}
)
# hierarchies enable option feature
conf = upsilonconf.load("config_dir")
conf == upsilonconf.Configuration(key={"foo": 1, "bar": 2})
# store hierarchy to default file in specified directory
upsilonconf.save(conf, "backup")# read command-line arguments (from sys.argv)
conf = upsilonconf.from_cli()
# parse arbitrary arguments to construct config
conf = upsilonconf.from_cli(["key=1", "sub.test=2"])
conf == Configuration(key=1, sub={"test": 2})
# use file as base config
conf = upsilonconf.from_cli(["--config", "config.yaml", "key=1", "sub.test=2"])
result = upsilonconf.load("config.yaml")
result.overwrite_all(key=1, sub={"test": 2})
conf == result
# enhance existing argparser
from argparse import ArgumentParser
parser = ArgumentParser()
# add other arguments...
conf, args = upsilonconf.from_cli(parser=parser)
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent