This repository has been archived on 2024-05-31. You can view files and clone it, but cannot push or open issues or pull requests.
devicehub-teal/ereuse_devicehub/resources/action/rate/v1_0.py

266 lines
10 KiB
Python
Raw Permalink Normal View History

from enum import Enum, unique
from itertools import groupby
2019-05-08 17:12:05 +00:00
from typing import Dict, Iterable, Tuple
from ereuse_devicehub.resources.action.models import BenchmarkDataStorage, BenchmarkProcessor, \
2020-08-17 14:45:18 +00:00
BenchmarkProcessorSysbench, RateComputer
from ereuse_devicehub.resources.action.rate.rate import BaseRate
2019-05-08 17:12:05 +00:00
from ereuse_devicehub.resources.device.models import Computer, DataStorage, Processor, \
RamModule
2019-04-11 16:29:51 +00:00
class RateAlgorithm(BaseRate):
2019-05-08 17:12:05 +00:00
"""The algorithm that generates the Rate v1.0.
Rate v1.0 rates only computers, counting their processor, ram,
data storage, appearance and functionality. This rate is only
triggered by a Snapshot from Workbench that has a VisualTest.
The algorithm is as follows:
1. Specialized subclasses of :class:`BaseRate` compute a rating
for each component. To perform this, each class normalizes first
the characteristics and benchmarks of the components between
0 and 1, and then they merge the values with specific formulas for
each components to get a resulting score.
The classes are:
* :class:`ProcessorRate`, using cores, speed, and ``BenchmarkProcessor``.
* :class:`RamRate`, using the total of RAM size and speed.
* :class:`DataStorageRate`, using the total of disk capacity,
and ``BenchmarkDataStorage``.
2. Merge the components individual rates into a single rate for
all components, using a weighted harmonic mean of
50% for the processor rating, 20% for the data storage rating,
and 30% for the RAM rating.
3. Merge the rate for the components with the appearance and
functionality from :class:`VisualTest`. ``Final Rate =
Components Rate + Functionality Rate + Appearance Rate``. The
value is between 0 and 4.7, included.
"""
@unique
class Appearance(Enum):
Z = 0.5
A = 0.3
B = 0
C = -0.2
D = -0.5
E = -1.0
@unique
class Functionality(Enum):
A = 0.4
B = -0.5
C = -0.75
D = -1
def __init__(self) -> None:
super().__init__()
self.RATES = {
# composition: type: (field, compute class)
Processor.t: ('processor', ProcessorRate()),
RamModule.t: ('ram', RamRate()),
DataStorage.t: ('data_storage', DataStorageRate())
2019-05-08 17:12:05 +00:00
} # type: Dict[str, Tuple[str, BaseRate]]
2019-05-08 17:12:05 +00:00
def compute(self, device: Computer) -> RateComputer:
"""Generates a new
:class:`ereuse_devicehub.resources.action.models.RateComputer`
2019-05-08 17:12:05 +00:00
for the passed-in device.
Do not call directly this class, but use
:meth:`ereuse_devicehub.resources.action.models.RateComputer.compute`,
which then calls this.
"""
2019-05-08 17:12:05 +00:00
assert isinstance(device, Computer), 'Can only rate computers'
2019-04-11 16:29:51 +00:00
2019-05-08 17:12:05 +00:00
rate = RateComputer()
rate.processor = rate.data_storage = rate.ram = 1 # Init
2019-05-08 17:12:05 +00:00
# Group cpus, rams, storage and compute their rate
# Treat the same way with HardDrive and SolidStateDrive like (DataStorage)
clause = lambda x: DataStorage.t if isinstance(x, DataStorage) else x.t
c = (c for c in device.components if clause(c) in set(self.RATES.keys()))
for type, components in groupby(sorted(c, key=clause), key=clause):
if type == Processor.t: # ProcessorRate.compute expects only 1 processor
components = next(components)
2019-05-08 17:12:05 +00:00
field, rate_cls = self.RATES[type]
result = rate_cls.compute(components)
if result:
setattr(rate, field, result)
2019-04-16 15:47:28 +00:00
rate_components = self.harmonic_mean_rates(rate.processor, rate.data_storage, rate.ram)
rate.rating = rate_components
device.actions_one.add(rate)
assert 0 <= rate.rating
2019-04-11 16:29:51 +00:00
return rate
class ProcessorRate(BaseRate):
"""Calculate a ProcessorRate of all Processor devices."""
# processor.xMin, processor.xMax
PROCESSOR_NORM = 3196.17, 17503.81
DEFAULT_CORES = 1
DEFAULT_SPEED = 1.6
DEFAULT_SCORE = 4000
2019-05-08 17:12:05 +00:00
def compute(self, processor: Processor):
"""Compute processor rate
We assume always exists a Benchmark Processor.
Obs: cores and speed are possible NULL value
:return: result is a rate (score) of Processor characteristics
"""
cores = processor.cores or self.DEFAULT_CORES
speed = processor.speed or self.DEFAULT_SPEED
2019-05-08 17:12:05 +00:00
benchmark_cpu = next(
e for e in reversed(processor.actions)
2019-05-08 17:12:05 +00:00
if isinstance(e, BenchmarkProcessor) and not isinstance(e, BenchmarkProcessorSysbench)
)
benchmark_cpu = benchmark_cpu.rate or self.DEFAULT_SCORE
# STEP: Fusion components
2019-05-14 18:31:43 +00:00
processor_rate = (benchmark_cpu + speed * 2000 * cores) / 2
# STEP: Normalize values
processor_norm = max(self.norm(processor_rate, *self.PROCESSOR_NORM), 0)
# STEP: Compute rate/score from every component
# Calculate processor_rate
if processor_norm >= self.CEXP:
processor_rate = self.rate_exp(processor_norm)
if self.CLIN <= processor_norm < self.CLOG:
processor_rate = self.rate_lin(processor_norm)
if processor_norm >= self.CLOG:
processor_rate = self.rate_log(processor_norm)
return processor_rate
class RamRate(BaseRate):
"""Calculate a RamRate of all RamModule devices."""
# ram.size.xMin; ram.size.xMax
SIZE_NORM = 256, 8192
RAM_SPEED_NORM = 133, 1333
# ram.speed.factor
RAM_SPEED_FACTOR = 3.7
# ram.size.weight; ram.speed.weight;
RAM_WEIGHTS = 0.7, 0.3
2019-05-08 17:12:05 +00:00
def compute(self, ram_devices: Iterable[RamModule]):
"""If ram speed or ram size, we assume default values before declared.
:return: result is a rate (score) of all RamModule components
"""
size = 0.0
speed = 0.0
# STEP: Filtering, data cleaning and merging of component parts
for ram in ram_devices:
_size = ram.size or 0
size += _size
if ram.speed:
speed += (ram.speed or 0) * _size
else:
speed += (_size / self.RAM_SPEED_FACTOR) * _size
# STEP: Fusion components
# To guarantee that there will be no 0/0
if size:
speed /= size
# STEP: Normalize values
size_norm = max(self.norm(size, *self.SIZE_NORM), 0)
ram_speed_norm = max(self.norm(speed, *self.RAM_SPEED_NORM), 0)
# STEP: Compute rate/score from every component
# Calculate size_rate
if self.CEXP <= size_norm < self.CLIN:
size_rate = self.rate_exp(size_norm)
if self.CLIN <= size_norm < self.CLOG:
size_rate = self.rate_lin(size_norm)
if size_norm >= self.CLOG:
size_rate = self.rate_log(size_norm)
# Calculate ram_speed_rate
if self.CEXP <= ram_speed_norm < self.CLIN:
ram_speed_rate = self.rate_exp(ram_speed_norm)
if self.CLIN <= ram_speed_norm < self.CLOG:
ram_speed_rate = self.rate_lin(ram_speed_norm)
if ram_speed_norm >= self.CLOG:
ram_speed_rate = self.rate_log(ram_speed_norm)
# STEP: Fusion Characteristics
return self.harmonic_mean(self.RAM_WEIGHTS, rates=(size_rate, ram_speed_rate))
class DataStorageRate(BaseRate):
"""Calculate the rate of all DataStorage devices."""
# drive.size.xMin; drive.size.xMax
SIZE_NORM = 4, 265000
READ_SPEED_NORM = 2.7, 109.5
WRITE_SPEED_NORM = 2, 27.35
# drive.size.weight; drive.readingSpeed.weight; drive.writingSpeed.weight;
DATA_STORAGE_WEIGHTS = 0.5, 0.25, 0.25
2019-05-08 17:12:05 +00:00
def compute(self, data_storage_devices: Iterable[DataStorage]):
"""Obs: size != NULL and 0 value & read_speed and write_speed != NULL
:return: result is a rate (score) of all DataStorage devices
"""
size = 0
read_speed = 0
write_speed = 0
# STEP: Filtering, data cleaning and merging of component parts
for storage in data_storage_devices:
# We assume all hdd snapshots have BenchmarkDataStorage
benchmark = storage.last_action_of(BenchmarkDataStorage)
# prevent NULL values
_size = storage.size or 0
size += _size
read_speed += benchmark.read_speed * _size
write_speed += benchmark.write_speed * _size
# STEP: Fusion components
# Check almost one storage have size, try catch exception 0/0
if size:
read_speed /= size
write_speed /= size
# STEP: Normalize values
size_norm = max(self.norm(size, *self.SIZE_NORM), 0)
read_speed_norm = max(self.norm(read_speed, *self.READ_SPEED_NORM), 0)
write_speed_norm = max(self.norm(write_speed, *self.WRITE_SPEED_NORM), 0)
# STEP: Compute rate/score from every component
# Calculate size_rate
if size_norm >= self.CLOG:
size_rate = self.rate_log(size_norm)
elif self.CLIN <= size_norm < self.CLOG:
size_rate = self.rate_lin(size_norm)
elif self.CEXP <= size_norm < self.CLIN:
size_rate = self.rate_exp(size_norm)
# Calculate read_speed_rate
if read_speed_norm >= self.CLOG:
read_speed_rate = self.rate_log(read_speed_norm)
elif self.CLIN <= read_speed_norm < self.CLOG:
read_speed_rate = self.rate_lin(read_speed_norm)
elif self.CEXP <= read_speed_norm < self.CLIN:
read_speed_rate = self.rate_exp(read_speed_norm)
# write_speed_rate
if write_speed_norm >= self.CLOG:
write_speed_rate = self.rate_log(write_speed_norm)
elif self.CLIN <= write_speed_norm < self.CLOG:
write_speed_rate = self.rate_lin(write_speed_norm)
elif self.CEXP <= write_speed_norm < self.CLIN:
write_speed_rate = self.rate_exp(write_speed_norm)
# STEP: Fusion Characteristics
return self.harmonic_mean(self.DATA_STORAGE_WEIGHTS,
rates=(size_rate, read_speed_rate, write_speed_rate))
2019-04-11 16:29:51 +00:00
rate_algorithm = RateAlgorithm()
class CannotRate(Exception):
pass