latin1prober.py

Go to the documentation of this file.

 
 
from typing import List, Union
 
from .charsetprober import CharSetProber
from .enums import ProbingState
 
FREQ_CAT_NUM = 4
 
UDF = 0  # undefined
OTH = 1  # other
ASC = 2  # ascii capital letter
ASS = 3  # ascii small letter
ACV = 4  # accent capital vowel
ACO = 5  # accent capital other
ASV = 6  # accent small vowel
ASO = 7  # accent small other
CLASS_NUM = 8  # total classes
 
# fmt: off
Latin1_CharToClass = (
    OTH, OTH, OTH, OTH, OTH, OTH, OTH, OTH,   # 00 - 07
    OTH, OTH, OTH, OTH, OTH, OTH, OTH, OTH,   # 08 - 0F
    OTH, OTH, OTH, OTH, OTH, OTH, OTH, OTH,   # 10 - 17
    OTH, OTH, OTH, OTH, OTH, OTH, OTH, OTH,   # 18 - 1F
    OTH, OTH, OTH, OTH, OTH, OTH, OTH, OTH,   # 20 - 27
    OTH, OTH, OTH, OTH, OTH, OTH, OTH, OTH,   # 28 - 2F
    OTH, OTH, OTH, OTH, OTH, OTH, OTH, OTH,   # 30 - 37
    OTH, OTH, OTH, OTH, OTH, OTH, OTH, OTH,   # 38 - 3F
    OTH, ASC, ASC, ASC, ASC, ASC, ASC, ASC,   # 40 - 47
    ASC, ASC, ASC, ASC, ASC, ASC, ASC, ASC,   # 48 - 4F
    ASC, ASC, ASC, ASC, ASC, ASC, ASC, ASC,   # 50 - 57
    ASC, ASC, ASC, OTH, OTH, OTH, OTH, OTH,   # 58 - 5F
    OTH, ASS, ASS, ASS, ASS, ASS, ASS, ASS,   # 60 - 67
    ASS, ASS, ASS, ASS, ASS, ASS, ASS, ASS,   # 68 - 6F
    ASS, ASS, ASS, ASS, ASS, ASS, ASS, ASS,   # 70 - 77
    ASS, ASS, ASS, OTH, OTH, OTH, OTH, OTH,   # 78 - 7F
    OTH, UDF, OTH, ASO, OTH, OTH, OTH, OTH,   # 80 - 87
    OTH, OTH, ACO, OTH, ACO, UDF, ACO, UDF,   # 88 - 8F
    UDF, OTH, OTH, OTH, OTH, OTH, OTH, OTH,   # 90 - 97
    OTH, OTH, ASO, OTH, ASO, UDF, ASO, ACO,   # 98 - 9F
    OTH, OTH, OTH, OTH, OTH, OTH, OTH, OTH,   # A0 - A7
    OTH, OTH, OTH, OTH, OTH, OTH, OTH, OTH,   # A8 - AF
    OTH, OTH, OTH, OTH, OTH, OTH, OTH, OTH,   # B0 - B7
    OTH, OTH, OTH, OTH, OTH, OTH, OTH, OTH,   # B8 - BF
    ACV, ACV, ACV, ACV, ACV, ACV, ACO, ACO,   # C0 - C7
    ACV, ACV, ACV, ACV, ACV, ACV, ACV, ACV,   # C8 - CF
    ACO, ACO, ACV, ACV, ACV, ACV, ACV, OTH,   # D0 - D7
    ACV, ACV, ACV, ACV, ACV, ACO, ACO, ACO,   # D8 - DF
    ASV, ASV, ASV, ASV, ASV, ASV, ASO, ASO,   # E0 - E7
    ASV, ASV, ASV, ASV, ASV, ASV, ASV, ASV,   # E8 - EF
    ASO, ASO, ASV, ASV, ASV, ASV, ASV, OTH,   # F0 - F7
    ASV, ASV, ASV, ASV, ASV, ASO, ASO, ASO,   # F8 - FF
)
 
# 0 : illegal
# 1 : very unlikely
# 2 : normal
# 3 : very likely
Latin1ClassModel = (
# UDF OTH ASC ASS ACV ACO ASV ASO
    0,  0,  0,  0,  0,  0,  0,  0,  # UDF
    0,  3,  3,  3,  3,  3,  3,  3,  # OTH
    0,  3,  3,  3,  3,  3,  3,  3,  # ASC
    0,  3,  3,  3,  1,  1,  3,  3,  # ASS
    0,  3,  3,  3,  1,  2,  1,  2,  # ACV
    0,  3,  3,  3,  3,  3,  3,  3,  # ACO
    0,  3,  1,  3,  1,  1,  1,  3,  # ASV
    0,  3,  1,  3,  1,  1,  3,  3,  # ASO
)
# fmt: on
 
 
class Latin1Prober(CharSetProber):
    def __init__(self) -> None:
        super().__init__()
        self._last_char_class = OTH
        self._freq_counter: List[int] = []
        self.resetreset()
 
    def reset(self) -> None:
        self._last_char_class = OTH
        self._freq_counter = [0] * FREQ_CAT_NUM
        super().reset()
 
    @property
    def charset_name(self) -> str:
        return "ISO-8859-1"
 
    @property
    def language(self) -> str:
        return ""
 
    def feed(self, byte_str: Union[bytes, bytearray]) -> ProbingState:
        byte_str = self.remove_xml_tags(byte_str)
        for c in byte_str:
            char_class = Latin1_CharToClass[c]
            freq = Latin1ClassModel[(self._last_char_class * CLASS_NUM) + char_class]
            if freq == 0:
                self._state_state = ProbingState.NOT_ME
                break
            self._freq_counter[freq] += 1
            self._last_char_class = char_class
 
        return self.statestate
 
    def get_confidence(self) -> float:
        if self.statestate == ProbingState.NOT_ME:
            return 0.01
 
        total = sum(self._freq_counter)
        confidence = (
            0.0
            if total < 0.01
            else (self._freq_counter[3] - self._freq_counter[1] * 20.0) / total
        )
        confidence = max(confidence, 0.0)
        # lower the confidence of latin1 so that other more accurate
        # detector can take priority.
        confidence *= 0.73
        return confidence