If Funcy and Pipe had a baby. Decorates all Funcy methods with Pipe superpowers.
MIT License
If Funcy and Pipe had a baby. Deal with data transformation in python in a sane way.
I love Ruby. It's a great language and one of the things they got right was pipelined data transformation. Elixir got this
even more right with the explicit pipeline operator |>
.
However, Python is the way of the future. As I worked more with Python, it was driving me nuts that the data transformation options were not chainable.
This project fixes this pet peeve.
pip install funcy-pipe
Or, if you are using poetry:
poetry add funcy-pipe
Extract a couple key values from a sql alchemy model:
import funcy_pipe as fp
entities_from_sql_alchemy
| fp.lmap(lambda r: r.to_dict())
| fp.lmap(lambda r: r | fp.omit(["id", "created_at", "updated_at"]))
| fp.to_list
Or, you can be more fancy and use whatever and pmap
:
import funcy_pipe as f
import whatever as _
entities_from_sql_alchemy
| fp.lmap(_.to_dict)
| fp.pmap(fp.omit(["id", "created_at", "updated_at"]))
| fp.to_list
Create a map from an array of objects, ensuring the key is always an int
:
section_map = api.get_sections() | fp.group_by(f.compose(int, that.id))
Grab the ID of a specific user:
filter_user_id = (
collaborator_map().values()
| fp.where(email=target_user)
| fp.pluck("id")
| fp.first()
)
Get distinct values from a list (in this case, github events):
events = [
{
"type": "PushEvent"
},
{
"type": "CommentEvent"
}
]
result = events | fp.pluck("type") | fp.distinct() | fp.to_list()
assert ["PushEvent", "CommentEvent"] == result
What if the objects are not dicts?
filter_user_id = (
collaborator_map().values()
| fp.where_attr(email=target_user)
| fp.pluck_attr("id")
| fp.first()
)
How about creating a dict where each value is sorted:
data
# each element is a dict of city information, let's group by state
| fp.group_by(itemgetter("state_name"))
# now let's sort each value by population, which is stored as a string
| fp.walk_values(
f.partial(sorted, reverse=True, key=lambda c: int(c["population"])),
)
A more complicated example (lifted from this project):
comments = (
# tasks are pulled from the todoist api
tasks
# get all comments for each relevant task
| fp.lmap(lambda task: api.get_comments(task_id=task.id))
# each task's comments are returned as an array, let's flatten this
| fp.flatten()
# dates are returned as strings, let's convert them to datetime objects
| fp.lmap(enrich_date)
# no date filter is applied by default, we don't want all comments
| fp.lfilter(lambda comment: comment["posted_at_date"] > last_synced_date)
# comments do not come with who created the comment by default, we need to hit a separate API to add this to the comment
| fp.lmap(enrich_comment)
# only select the comments posted by our target user
| fp.lfilter(lambda comment: comment["posted_by_user_id"] == filter_user_id)
# there is no `sort` in the funcy library, so we reexport the sort built-in so it's pipe-able
| fp.sort(key="posted_at_date")
# create a dictionary of task_id => [comments]
| fp.group_by(lambda comment: comment["task_id"])
)
Want to grab the values of a list of dict keys?
def add_field_name(input: dict, keys: list[str]) -> dict:
return input | {
"field_name": (
keys
# this is a sneaky trick: if we reference the objects method, when it's called it will contain a reference
# to the object
| fp.map(input.get)
| fp.compact
| fp.join_str("_")
)
}
result = [{ "category": "python", "header": "functional"}] | fp.map(fp.rpartial(add_field_name, ["category", "header"])) | fp.to_list
assert result == [{'category': 'python', 'header': 'functional', 'field_name': 'python_functional'}]
You can also easily test multiple conditions across API data (extracted from this project)
all_checks_successful = (
last_commit.get_check_runs()
| fp.pluck_attr("conclusion")
# if you pass a set into `all` each element of the set is used to build a predicate
# this condition tests if the "conclusion" attribute is either "success" or "skipped"
| fp.all({"success", "skipped"})
)
Want to grab the values of a list of dict keys?
def add_field_name(input: dict, keys: list[str]) -> dict:
return input | {
"field_name": (
keys
# this is a sneaky trick: if we reference the objects method, when it's called it will contain a reference
# to the object
| fp.map(input.get)
| fp.compact
| fp.join_str("_")
)
}
result = [{ "category": "python", "header": "functional"}] | fp.map(fp.rpartial(add_field_name, ["category", "header"])) | fp.to_list
assert result == [{'category': 'python', 'header': 'functional', 'field_name': 'python_functional'}]
You can also easily group dictionaries by a key (or arbitrary function):
import operator
result = [{"age": 10, "name": "Alice"}, {"age": 12, "name": "Bob"}] | fp.group_by(operator.itemgetter("age"))
assert result == {10: [{'age': 10, 'name': 'Alice'}], 12: [{'age': 12, 'name': 'Bob'}]}
breakpoint()
on the input valueThere are some functions which are not yet merged upstream into funcy, and may never be. You can patch funcy
to add them using:
import funcy_pipe
funcy_pipe.patch()
where(some="Condition") | first
or where_attr(some="Condition") | first
Create a module alias for funcy-pipe
to make things clean (import *
always irks me):
# fp.py
from funcy_pipe import *
# code py
import fp
array |> map(fn) |> filter(fn)
array.map(&:fn).filter(&:fn)