/tibia

Monads in python for pipeline-based development with Rust-like interface

Primary LanguagePythonMIT LicenseMIT

tibia

If you ask why tibia is called tibia - i don't remember why 😄

Containers

tibia provides several containers for values:

  • Value for already computed (sync) values with pipe operator / fluent API
  • Future for not yet computed (async) values with pipe operator / fluent API
  • Maybe is well-known monad for present and non-present already competed values with pipe operator / fluent API; provides 2 containers:
    • Some - value is present and actually contained inside
    • Empty - nothing is contained (something like python None)
  • FutureMaybe is like Future for Value, but for Maybe
  • Result is well-known monad for successful or failed contained value states with pipe operator / fluent API; provides 2 containers:
    • Ok - value is successfully computed
    • Err - failed to compute value
  • FutureResult is like Future for Value, but for Maybe

Also tibia provides some utilitarian functions for Iterable and Mapping in pipeline-based context:

  • Iterable
    • eager
      • map - applies passed function to each element of iterable eagerly
      • filter - filter values in iterable with passed predicate eagerly
      • reduce & reduce_to - aggregates iterable values into single value
      • sort_asc & sort_desc - sort iterable values
      • take - take first or last N values from iterable (can fail on infinite iterable)
      • first - take first value or default
      • skip - skip first or last N values from iterable (can fail on infinite iterable)
      • join - flatten iterable of iterables into single iterable
    • async (aio)
      • map - applies passed async function to each element of iterable
      • filter - filters elements in iterable with async predicate function
    • lazy
      • map - applies passed async function to each element of iterable lazily
      • filter - filters elements in iterable with async predicate function lazily
      • take_while - takes elements from iterable until predicate is true
      • skip_while - skips elements from iterable until predicate is true
    • threaded
      • map - applies passed sync function to each element of iterable in threads
      • filter - filters in threads elements of iterable
  • Mapping
    • value
      • map - applies function to each value of mapping
      • filter - filters mapping pairs based on value
      • iterate - returns iterable of values (lazy as generator)
      • set - set value to key
      • get - get value by key (can raise KeyError)
      • get_or - get value or passed default by key
      • maybe_get - get value by key in Maybe container (Some if key is present, Empty if not)
    • key
      • map - applies function to each key of mapping
      • filter - filters mapping pairs based on key
      • iterate - returns iterable of keys (lazy as generator)
    • item
      • map - maps key-value pair to new key-value pair
      • map_to_value - maps value based on key-value pair
      • map_to_key - maps key based on key-value pair
      • filter - filters mapping elements based on key-value pairs
      • iterate - returns iterable of key-value pairs (lazy as generator of tuples)

Value

Value is a simple container that brings to python so called pipe operator.

So how to use it? Put some concrete value to the container, for example a number. Now instead of passing this value as an argument to a function pass a function as some action that you want to be performed on this value:

_ = (
    Value(1)
    .map(add, 1)
    .map(multiply_by, 3)
    .map(subtract, 10)
    .map(multiply_by, -1)
    .inspect(print)
    .unwrap()
)

Thus we build a declarative chain of actions we perform on some piece of data.

Value.map

Value.map is method that invokes passed function on contained value. It supports not only single-argument functions, but actually any function with only one requirement: contained value must be the first argument for the function. Other arguments can be passed as *args and/or **kwargs to the Value.map method.

So image we have a Value[int]. The following example function will be completely valid for Value.map:

# simple single-argument function, contained value will be x
def add_one(x: int) -> int:
    return x + 1

# 2 argument function, contained value will be x again, and argument for y must
# be passed following the function argument to map method
def add(x: int, y: int) -> int:
    return x + y

# again 2 argument function, but in this case y argument can be ommited in map
def multiply_by(x: int, y: int = 1) -> int:
    return x * y

Value.map returns new Value container for data returned from the passed function, thus with Value.map method chaining one can build pipelines on data.

Value.map works only with synchronous functions. For asynchronous functions use Value.map_async.

Value.map_async

Value.map_async is direct analogue of Value.map but for async functions. Main difference is that instead of Value it returns Future - tibias simple container over coroutine (Future is discussed further, but most things valid for Value are valid for Future and their interface are very similar).

Value.inspect

Value.inspect has nearly the same signature as Value.map with only difference: it does not wrap returned from the passed function value into new Value container, but returns the current self Value container.

Simply this is function for making side-effects - something we want to happen, but don't care about result. In the example above Value.inspect was used to print contained in container value. print returns None, so if we've used Value.map we would lost the data, but with Value.inspect we just performed an action and continued working on already contained data.

Like Value.map can be used only with synchronous functions.

Value.inspect_async

Like Value.map_async is analogue of Value.map for async function that returns Future containers Value.inspect_async is the same for Value.inspect.

Value.unwrap

Method for extracting value from container if one is not further needed.

Value.wraps

Decorator that changes signature of wrapped function by containing returned value in Value container. For example initial signature was:

fn: (int, int, str) -> str

With Value.wraps decorator applied it becomes:

fn: (int, int, str) -> Value[str]

Future

Future is direct analogue of Value, but for values that are not calculated (awaited) yet. For JS developers it might be widely know as Promise. It has the same as Value interface:

  • Future.map & Future.inspect for sync functions
  • Future.map_async & Future.inspect_async for async functions
  • Future.unwrap for extracting contained value (must be awaited)

Additionally one can await directly Future without unwrap:

_ = await future.unwrap()

# same as

_ = await future

This is just a shortcut, but I would generally recommend explicit unwrapping and waiting of contained value.

Maybe & FutureMaybe

Maybe is an alternative for python Optional. At first glance it might seem that it is not needed, but it provides to apply functions without checks on emptiness and also can consider None as actual value.

It consists of 2 containers:

  • Some - indicates that value is present (even if one is None)
  • Empty - indicates that there is no value

For example let's imagine one is creating data structure for updating some data (for example table in DB). Naive approach would be to create some class with all-optional fields:

class UpdateUser:
  first_name: str | None = None
  second_name: str | None = None
  birthdate: datetime | None = None

Imagine that birthdate is nullable in table we want to update. If we get this structure with None in birthdate field we cannot determine whether we want to set it to null or we do not want to perform any update on this column.

With Maybe this problem goes away. Empty state tells us that we do not want to perform action and Some unambiguously tells us that we want to set column to contained value:

class UpdateUser:
  first_name: Maybe[str] = Empty()
  second_name: Maybe[str] = Empty()
  birthdate: Maybe[datetime | None] = Empty()

With this approach it is much more clear what each value means and what is valid state.

Piping API

For making pipelines Maybe & FutureMaybe like Value & Future provide the following methods:

  • map / map_async - apply contained value to function if one is present and wrap result into Some
  • map_or / map_or_async - apply contained value to function if one is present and return result of function directly or replace it with passed default value
  • inspect / inspect_async - apply function as side-effect if value is present ignoring result of passed function and return current container

Imagine we apply function that returns R:

Container Method Returns
Ok[T] map Ok[R]
Ok[T] map_async FutureMaybe(Ok[R])
Empty map Empty
Empty map_async FutureMaybe(Empty)
Ok[T] map_or R
Ok[T] map_or_async Future[R]
Empty map_or R from default
Empty map_or_async Future[R] from default
Ok[T] inspect Ok[T]
Ok[T] inspect_async FutureMaybe(Ok[T])
Empty inspect Empty
Empty inspect_async FutureMaybe(Empty)

Mainly one would use map and map_async to perform desired actions without thinking if value is present, some side-effect functions (for debugging for example) with inspect & inspect_async and than unwrap value with map_or or map_or_async methods (or via Unwrapping API).

Unwrapping API

In order to extract value from container one can use one of the following methods:

  • expect
    • if Some: return contained value
    • if Empty: raise ValueError with passed error message
  • unwrap - same as expect, but uses built-in error message "must be some"
  • unwrap_or
    • if Some: return contained value
    • if Empty: return passed default value
  • unwrap_or_none
    • if Some: return contained value
    • if Empty: return None

Logical API

In order to check / validate contained value on can use following methods:

  • is_some - True if container is Some, otherwise False
  • is_empty - True if container is Empty, otherwise False
  • is_some_and - True if container is Some and passed predicate is True (function that checks contained value), otherwise False
  • is_empty_or - True if container is Empty or passed predicate is True (function that check contained in Some value), otherwise False

For example imagine we have is_even predicate:

Container Method Returns
Ok(2) is_some_and True
Ok(1) is_some_and False
Empty() is_some_and False
Ok(1) is_empty_or False
Ok(2) is_empty_or True
Empty() is_empty_or True

is_some_and & is_empty_or methods also have is_some_and_async & is_empty_or_async alternatives for async predicates, as base versions work only with sync functions. Async alternatives return Future[bool] for further piping if needed.

Construction API

There are a few static function in Maybe class used for constructing Maybe containers:

  • Maybe.from_value - always wraps value into Some
  • Maybe.from_value_when - wraps value into Some if value satisfies passed predicate, otherwise Empty
  • Maybe.from_optional - wraps value into Some if one is not None, otherwise Empty
  • Maybe.from_optional_when - wraps value into Some if value is not None and satisfies passed predicate , otherwise Empty

Decorator API

Based on from_value and from_optional Maybe provides 2 decorators for functions:

  • Maybe.wraps - wraps returned from function value via Maybe.from_value
  • Maybe.wraps_optional - wraps returned from function value via Maybe.from_optional

Point-free API

All discussed above methods can also be used as functions contained in tibia.maybe module. They can be found useful when working with iterables of Maybe for example (for massive filtering and unwrapping without loosing type hints).

FutureMaybe provides exactly the same API, but for "futurized" Maybe value.

Result & FutureResult

Result & FutureResult monads provide the ability for indicating computation success and error states without implicit error raises. Yes, yes, raising errors is actually implicit way, that highly resembles to goto operator widely forbidden or not implemented for increasing code complexity exponentially. With Result one does not raise exception, but explicitly returns it (like Rust and Go do for example).

It consists of 2 containers:

  • Ok - indicates successful result
  • Err - indicates failed result

Both of the containers store some value, Ok - what was actually computed, Err - some error representation (not always Exception).

Piping API

It is a bit extended relative to Maybe or Value and consist of the following methods:

  • map / map_async - mapping for Ok container
  • map_err / map_err_async - mapping for Err container
  • map_or / map_or_async - mapping and unwrapping for Ok container
  • map_err_or / map_err_or_async - mapping and unwrapping for Err container
  • inspect / inspect_async - applying side-effect function for Ok container
  • inspect_err / inspect_err_async - applying side-effect function for Err container

Unwrapping API

Also provides more options for unwrapping both containers:

  • expect
    • Ok - returns container value
    • Err - raises ValueError with passed error message
  • unwrap
    • Ok - returns container value
    • Err - raises ValueError with default error message "must be ok"
  • unwrap_or
    • Ok - returns container value
    • Err - returns passed default value
  • expect_err
    • Ok - raises ValueError with passed error message
    • Err - returns container value
  • unwrap_err
    • Ok - raises ValueError with default error message "must be err"
    • Err - returns container value
  • unwrap_err_or
    • Ok - returns passed default value
    • Err - returns container value

Logical API

Many more options for validating containers and values:

  • is_ok - True if Ok, otherwise False
  • is_ok_and / is_ok_and_async - True if Ok and contained value satisfies passed predicate, otherwise False (returns Future[bool] for async version)
  • is_ok_or / is_ok_or - True if Ok or Err contained value satisfies passed predicate, otherwise False (returns Future[bool] for async version)
  • is_err_ - True if Err, otherwise False
  • is_err__and / is_err__and_async - True if Err and contained value satisfies passed predicate, otherwise False (returns Future[bool] for async version)
  • is_err__or / is_err__or - True if Err or Ok contained value satisfies passed predicate, otherwise False (returns Future[bool] for async version)

Decorator API

Provides 2 decorators:

  • Result.wraps - simply always returns Ok-wrapped function result
  • Result.safe - if wrapped function:
    • returned value successfully - wraps it into Ok
    • raised exception in exceptions - wraps catches error into Err
    • raised exception not in exceptions - exception is raised (undesired behavior)

Point-free API

Same as Maybe, Result provides simple functions that can replace any method.

ResultMaybe provides exactly the same API, but for "futurized" Result value.