rosu-pp-py 4.0.2

Difficulty and performance calculation for osu!

rosu-pp-py

Library to calculate difficulty and performance attributes for all osu! modes.

This is a python binding to the Rust library rosu-pp which was bootstrapped through PyO3. As such, its performance is much faster than a native python library.

Usage

The library exposes multiple classes:

• Beatmap: Parsed .osu file
• GameMode
• Calculators
  • Difficulty: Class to calculate difficulty attributes, strains, or create gradual calculators
  • Performance: Performance attributes calculator
  • GradualDifficulty: Calculator to calculate difficulty attributes after each hitobject
  • GradualPerformance: Calculator to calculate performance attributes after each hitresult
  • BeatmapAttributesBuilder: Beatmap attributes calculator
• Results
  • DifficultyAttributes
  • Strains: Strain values of a difficulty calculation, suitable to plot difficulty over time
  • PerformanceAttributes
  • BeatmapAttributes
• HitResultPriority: Passed to Performance, decides whether specified accuracy should be realized through good or bad hitresults
• HitResultGenerator: Passed to Performance, decides how hitresults are being generated
• ScoreState: Hitresults and max combo of a score, found in PerformanceAttributes and passed to gradual calculators

Example

Calculating performance

import rosu_pp_py as rosu

# either `path`, `bytes`, or `content` must be specified when parsing a map
map = rosu.Beatmap(path = "/path/to/file.osu")

# Optionally convert to a specific mode for optionally given mods
map.convert(rosu.GameMode.Mania, "6K")

# Whereas osu! simply times out on malicious maps, rosu-pp does not. To
# prevent potential performance/memory issues, it is recommended to check
# beforehand whether a map is too suspicious for further calculation.
if map.is_suspicious():
    exit()

perf = rosu.Performance(
    # various kwargs available
    accuracy = 98.76,
    lazer = False, # defaults to True if not specified
    misses = 2,
    combo = 700,
    hitresult_priority=rosu.HitResultPriority.BestCase,
)

# Each kwarg can also be specified afterwards through setters
perf.set_accuracy(99.11) # override previously specified accuracy
perf.set_mods(8 + 64)    # HDDT
perf.set_clock_rate(1.4)

# Second arg of map attributes specifies whether mods still need to be accounted for.
# `True`: mods already considered; `False`: value should still be adjusted
perf.set_ar(10.5, True)
perf.set_od(5, False)

# The 'Closest' hitresult generation may be significantly slower but provides
# hitresults that match the given accuracy as close as possible.
perf.set_hitresult_generator(rosu.HitResultGenerator.Closest, rosu.GameMode.Osu)
# Especially for mania, 'Closest' might be too slow in general so let's use
# 'Fast' instead (which is the default anyway for all modes if unspecified).
perf.set_hitresult_generator(rosu.HitResultGenerator.Fast, rosu.GameMode.Mania)

# Calculate for the map
attrs = perf.calculate(map)

# Note that calculating via map will have to calculate difficulty attributes
# internally which is fairly expensive. To speed it up, you can also pass in
# previously calculated attributes, but be sure they were calculated for the
# same difficulty settings like mods, clock rate, custom map attributes, ...

perf.set_accuracy(100)
perf.set_misses(None)
perf.set_combo(None)

# Calculate a new set of attributes by re-using previous attributes instead of the map
max_attrs = perf.calculate(attrs)

print(f'PP: {attrs.pp}/{max_attrs.pp} | Stars: {max_attrs.difficulty.stars}')

Gradual calculation

import rosu_pp_py as rosu

# Parsing the map, this time through the `content` kwarg
with open("/path/to/file.osu") as file:
    map = rosu.Beatmap(content = file.read())

# Specifying some difficulty parameters
diff = rosu.Difficulty(
    mods = 16 + 1024, # HRFL
    clock_rate = 1.1,
    ar = 10.2,
    fixed_ar = True,
)

# Gradually calculating *difficulty* attributes
gradual_diff = diff.gradual_difficulty(map)

for i, attrs in enumerate(gradual_diff, 1):
    print(f'Stars after {i} hitobjects: {attrs.stars}')

# Gradually calculating *performance* attributes
gradual_perf = diff.gradual_performance(map)
i = 1

while True:
    state = rosu.ScoreState(
        max_combo = i,
        n300 = i,
        n100 = 0,
        # ...
    )

    attrs = gradual_perf.next(state)

    if attrs is None:
        # All hitobjects have been processed
        break

    print(f'PP: {attrs.pp}')
    i += 1

Mods

Wherever mods are specified, their type should coincide with the following alias definition:

GameMods = int | str | GameMod | list[GameMod | str | int]
GameMod = dict[str, str | GameModSettings]
GameModSettings = dict[str, bool | float | str]

That means, mods can be given either through their (legacy) bitflags, a string for acronyms, a "GameMod" dict, or a sequence whose items are either a "GameMod" dict, a single acronym string, or bitflags for a single mod.

A "GameMod" dict must have the item 'acronym': str and an optional item 'settings': GameModSettings.

Some examples for valid mods look as follows:

mods = 8 + 64              # Hidden, DoubleTime
mods = "hRNcWIez"          # HardRock, Nightcore, Wiggle, Easy
mods = { 'acronym': "FI" } # FadeIn
mods = [
    1024,
    'nf',
    {
        'acronym': "AC",
        'settings': {
            'minimum_accuracy': 95,
            'restart': True
        }
    }
] # Flashlight, NoFail, AccuracyChallenge

import json
mods_json = '[{"acronym": "TC"}, {"acronym": "HT", "settings": {"speed_change": 0.6}}]'
mods = json.loads(mods_json) # Traceable, HalfTime

Installing rosu-pp-py

Installing rosu-pp-py requires a supported version of Python and Rust.

If a pre-built wheel is available for your architecture, you can even skip the Rust part.

Once Python and (optionally) Rust are ready to go, you can install the project with pip:

$ pip install rosu-pp-py

or

$ pip install git+https://github.com/MaxOhn/rosu-pp-py

Learn More

• rosu-pp
• Rust
• PyO3