Scalpl provides a lightweight wrapper that helps you to operate
on nested dictionaries seamlessly through the built-in dict
API, by using dot-separated string keys.
It's not a drop-in replacement for your dictionnaries, just syntactic
sugar to avoid this['annoying']['kind']['of']['things'] and
prefer['a.different.approach'].
No conversion cost, a thin computation overhead: that's Scalpl in a nutshell.
There are a lot of good libraries to operate on nested dictionaries, such as Addict or Box , but if you give Scalpl a try, you will find it:
- ⚡ Fast
- 🚀 Powerful as the standard dict API
- 👌 Well tested
Scalpl is a Python3-only module that you can install via pip
pip3 install scalplScalpl provides two classes that can wrap around your dictionaries:
- LightCut: a wrapper that handles operations on nested
dict. - Cut: a wrapper that handles operations on nested
dictand that can cut accrosslistitem.
Usually, you will only need to use the Cut wrapper, but if you do
not need to operate through lists, you should work with the LightCut
wrapper as its computation overhead is a bit smaller.
These two wrappers strictly follow the standard dict
API, that
means you can operate seamlessly on dict,
collections.defaultdict or collections.OrderedDict.
Let's see what it looks like with a toy dictionary ! 👇
from scalpl import Cut
data = {
'pokemons': [
{
'name': 'Bulbasaur',
'type': ['Grass', 'Poison'],
'category': 'Seed',
'ability': 'Overgrow'
},
{
'name': 'Charmander',
'type': 'Fire',
'category': 'Lizard',
'ability': 'Blaze',
},
{
'name': 'Squirtle',
'type': 'Water',
'category': 'Tiny Turtle',
'ability': 'Torrent',
}
],
'trainers': [
{
'name': 'Ash',
'hometown': 'Pallet Town'
}
]
}
# Just wrap your data, and you're ready to go deeper !
proxy = Cut(data)You can use the built-in dict API to access its values.
proxy['pokemons[0].name']
# 'Bulbasaur'
proxy.get('pokemons[1].sex', 'Unknown')
# 'Unknown'
'trainers[0].hometown' in proxy
# TrueBy default, Scalpl uses dot as a key separator, but you are free to use a different character that better suits your needs.
# You just have to provide one when you wrap your data.
proxy = Cut(data, sep='->')
# Yarrr!
proxy['pokemons[0]->name']You can also easily create or update any key/value pair.
proxy['pokemons[1].weaknesses'] = ['Ground', 'Rock', 'Water']
proxy['pokemons[1].weaknesses']
# ['Ground', 'Rock', 'Water']
proxy.update({
'trainers[0].region': 'Kanto',
})Following its purpose in the standard API, the setdefault method allows you to create any missing dictionary when you try to access a nested key.
proxy.setdefault('pokemons[2].moves.Scratch.power', 40)
# 40And it is still possible to iterate over your data.
proxy.items()
# [('pokemons', [...]), ('trainers', [...])]
proxy.keys()
# ['pokemons', 'trainers']
proxy.values()
# [[...], [...]]By the way, if you have to operate on a list of dictionaries, the
Cut.all method is what you are looking for.
# Let's teach these pokemons some sick moves !
for pokemon in proxy.all('pokemons'):
pokemon.setdefault('moves.Scratch.power', 40)Also, you can remove a specific or an arbitrary key/value pair.
proxy.pop('pokemons[0].category')
# 'Seed'
proxy.popitem()
# ('trainers', [...])
del proxy['pokemons[1].type']Because Scalpl is only a wrapper around your data, it means you can get it back at will without any conversion cost. If you use an external API that operates on dictionary, it will just work.
import json
json.dumps(proxy.data)
# "{'pokemons': [...]}"Finally, you can retrieve a shallow copy of the inner dictionary or remove all keys.
shallow_copy = proxy.copy()
proxy.clear()This humble benchmark is an attempt to give you an overview of the performance
of Scalpl compared to Addict,
Box and the built-in dict.
It will summarize the number of operations per second that each library is able to perform on the JSON dump of the Python subreddit main page.
You can run this benchmark on your machine with the following command:
python3 ./performance_tests.py
Here are the results obtained on an Intel Core i5-7500U CPU (2.50GHz) with Python 3.6.4.
Addict:
instanciate:-------- 18,485 ops per second. get:---------------- 18,806 ops per second. get through list:--- 18,599 ops per second. set:---------------- 18,797 ops per second. set through list:--- 18,129 ops per second.
Box:
instanciate:--------- 4,150,396 ops per second. get:----------------- 1,424,529 ops per second. get through list:---- 110,926 ops per second. set:----------------- 1,332,435 ops per second. set through list:---- 110,833 ops per second.
Scalpl:
instanciate:-------- 136,517,371 ops per second. get:---------------- 24,918,648 ops per second. get through list:--- 12,624,630 ops per second. set:---------------- 26,409,542 ops per second. set through list:--- 13,765,265 ops per second.
dict:
instanciate:--------- 92,119,547 ops per second. get:----------------- 186,290,996 ops per second. get through list:---- 178,747,154 ops per second. set:----------------- 159,224,669 ops per second. set through list :--- 79,294,520 ops per second.
As a conclusion and despite being ~10 times slower than the built-in
dict, Scalpl is ~20 times faster than Box on simple read/write
operations, and ~100 times faster when it traverse lists. Scalpl is
also ~1300 times faster than Addict.
However, do not trust benchmarks and test it on a real use-case.
- What if my keys contain dots ?
If your keys contain a lot of dots, you should use an other key separator when wrapping your data:
proxy = Cut(data, sep='->') proxy['computer->network->127.0.0.1']
Otherwise, split your key in two part:
proxy = Cut(data) proxy['computer.network']['127.0.0.1']
What if my keys contain spaces ?:
proxy = Cut(data) proxy['it works perfectly'] = 'fine'
Contributions are welcomed and anyone can feel free to submit a patch, report a bug or ask for a feature. Please open an issue first in order to encourage and keep tracks of potential discussions ✍️
Scalpl is released into the Public Domain. 🎉
Ps: If we meet some day, and you think this small stuff worths it, you can give me a beer, a coffee or a high-five in return: I would be really happy to share a moment with you ! 🍻