420 lines
15 KiB
Python
420 lines
15 KiB
Python
from __future__ import annotations
|
||
|
||
import re
|
||
from typing import Iterable, List, TypeVar
|
||
from collections.abc import Sequence
|
||
|
||
import pandas as pd
|
||
|
||
T = TypeVar("T")
|
||
|
||
CLASSES = ("w", "d", "c", "t", "s", "h", "b", "e", "l", "r")
|
||
DISTRICTS_PREFIXES = ("мо ", "р-н","городское","лесхоз")
|
||
COUNTRYSIDE_PREFIXES = (
|
||
"г", "п", "д", "гп", "рп", "кп", "пгт", "c", "хутор", " урочище")
|
||
TERRITORY_PREFIXES = (
|
||
"тер.", " тер", "снт ", "ст ", "дск ", "днп ", "дпк ", "нп ", "пдк ", "т/б ", "садоводство", "массив", "хозя", "сад-во")
|
||
STREET_PREFIXES = (
|
||
" ул", " бул", " пр", " ш", " пер", " дор", " маг", " наб", " пл", " просп", " туп", "шоссе", "лини", "аллея",
|
||
"мост", " парк", "кольцо", "проезд", "съезд","переулок",
|
||
"ул.", "бул.", "пр.", "ш.", "пер.", "дор.", "маг.", "наб.", "пл.", "просп.", "туп.")
|
||
HOUSES_PREFIXES = ("д.", "уч.", "участок", "мкд", "тп","дом","дома")
|
||
BUILDING_PREFIXES = ("к.", "к ","корп", "корпус")
|
||
EDIFICE_PREFIXES=("стр.", "строение")
|
||
LETTER = ("лит.", "литера", " л.")
|
||
PREFIXES = (DISTRICTS_PREFIXES, COUNTRYSIDE_PREFIXES, TERRITORY_PREFIXES, STREET_PREFIXES, HOUSES_PREFIXES, BUILDING_PREFIXES, EDIFICE_PREFIXES,LETTER)
|
||
|
||
|
||
def unfold_house_ranges(token: str) -> List[str]:
|
||
addresses = []
|
||
pairs_strings = re.findall(r"([\d]+-[\d]+)", token)
|
||
for pair_string in pairs_strings:
|
||
a, b = pair_string.split("-")
|
||
a, b = int(a), int(b)
|
||
|
||
if b > a:
|
||
addresses += [re.sub(r"([\d]+-[\d]+)", number, token) for number in map(str, range(a, b + 1))]
|
||
else:
|
||
token = token.replace("-", "/")
|
||
if not addresses:
|
||
addresses.append(token)
|
||
return addresses
|
||
|
||
def any_of_in(substrings: Iterable[str], string: str) -> bool:
|
||
return any(map(lambda substring: substring in string, substrings))
|
||
|
||
|
||
def flatten(arr: Iterable[List[T]]) -> List[T]:
|
||
return sum(arr, [])
|
||
|
||
|
||
def find_room(token: pd.Series, pre_token: pd.Series) -> str:
|
||
if re.search(r"\bпом\.?", token['obj']):
|
||
return "r"
|
||
return ""
|
||
|
||
|
||
def find_litera(token: pd.Series, pre_token: pd.Series) -> str:
|
||
if find_room(token, pre_token):
|
||
return ""
|
||
if any_of_in(LETTER, token['obj'].lower()) \
|
||
or re.search(r"\d{1,3}([А-Я]|[а-я])( |$)", token['obj']):
|
||
return "l"
|
||
if (re.search(r"\b([А-Я]|[а-я]){1}$", token['obj']) \
|
||
and ("l" in pre_token['class'] or "h" in pre_token['class'])) \
|
||
and (" ш" not in token["obj"]) \
|
||
and not find_countryside(token, pre_token):
|
||
return "l"
|
||
return ""
|
||
def find_edifice(token: pd.Series, pre_token: pd.Series) -> str:
|
||
if any_of_in(EDIFICE_PREFIXES, token['obj'].lower()):
|
||
return "e"
|
||
return ""
|
||
|
||
def find_building(token: pd.Series, pre_token: pd.Series) -> str:
|
||
if re.search(r"\d", token['obj']) and not find_room(token,pre_token):
|
||
if any_of_in(BUILDING_PREFIXES, token['obj'].lower()) \
|
||
or "b" in pre_token['class'] and ("h" not in token['class']) and not find_edifice(token,pre_token)\
|
||
or re.search(r"к\.* ?\d", token['obj']):
|
||
return "b"
|
||
return ""
|
||
|
||
|
||
def find_house(token: pd.Series, pre_token: pd.Series) -> str:
|
||
if re.search(r"\d{1,4}", token['obj']) and not find_room(token,pre_token):
|
||
if any_of_in(HOUSES_PREFIXES, token['obj'].lower()):
|
||
return "h"
|
||
if re.search(r"(д|д\.) ?\d{1,4} ?\/*\d* ?", token['obj']):
|
||
return "h"
|
||
if ("s" in pre_token['class'] or "h" in pre_token['class'] or "s" in token['class']) \
|
||
and not any_of_in(("-я", "-й", "-Я"), token['obj']) \
|
||
and not find_building(token, pre_token)\
|
||
and not find_edifice(token,pre_token):
|
||
return "h"
|
||
if find_building(token, pre_token) \
|
||
and not any_of_in(("-я", "-й", "-Я"), token['obj']) \
|
||
and True:
|
||
if len(re.findall(r"\d{1,4}", token['obj'])) > 1:
|
||
return "h"
|
||
if int(re.search(r"\d{1,4}", token['obj']).group()) // 10 >0:
|
||
return "h"
|
||
return ""
|
||
|
||
|
||
def find_street(token: pd.Series, pre_token: pd.Series) -> str:
|
||
if any_of_in(STREET_PREFIXES, token['obj'].lower()):
|
||
return "s"
|
||
if re.search(r"\b[А-Яа-я]{4,}\b", token['obj']) \
|
||
and not any([el in token["obj"].lower() for pr in PREFIXES for el in pr if len(el)>2]) \
|
||
and not ("d" in token["class"] or "t" in token["class"] or "c" in token["class"]):
|
||
return "s"
|
||
return ""
|
||
|
||
|
||
def find_territory(token: pd.Series, pre_token: pd.Series) -> str:
|
||
if any_of_in(TERRITORY_PREFIXES, token['obj'].lower()):
|
||
return "t"
|
||
return ""
|
||
|
||
|
||
def find_countryside(token: pd.Series, pre_token: pd.Series) -> str:
|
||
if any_of_in(COUNTRYSIDE_PREFIXES, token['obj'].lower()) \
|
||
and re.search(r"\b[гпдрпктc]{1,3}(\b|\. )", token['obj']) \
|
||
and not find_house(token, pre_token) \
|
||
and not any_of_in(STREET_PREFIXES, token['obj'].lower()):
|
||
return "c"
|
||
return ""
|
||
|
||
|
||
def find_district(token: pd.Series, pre_token: pd.Series) -> str:
|
||
if any_of_in(DISTRICTS_PREFIXES, token['obj'].lower()):
|
||
return "d"
|
||
return ""
|
||
|
||
def address_classification(token: pd.Series, pre_token: pd.Series) -> pd.Series:
|
||
brackets = re.search(r"\(.+\)", token["obj"])
|
||
if brackets:
|
||
token["obj"] = re.sub(r"\(.+\)", "()", token["obj"])
|
||
token["class"] += find_district(token, pre_token)
|
||
token["class"] += find_countryside(token, pre_token)
|
||
token["class"] += find_territory(token, pre_token)
|
||
token["class"] += find_street(token, pre_token)
|
||
token["class"] += find_house(token, pre_token)
|
||
token["class"] += find_building(token, pre_token)
|
||
token["class"] += find_edifice(token, pre_token)
|
||
token["class"] += find_litera(token, pre_token)
|
||
token["class"] += find_room(token, pre_token)
|
||
if token['class'] == "":
|
||
token['class'] = "w"
|
||
if brackets:
|
||
token["obj"] = re.sub(r"\(\)", brackets.group(), token["obj"])
|
||
return token
|
||
|
||
def cut_address(ad: pd.Series, cl: str) -> pd.Series:
|
||
while ad["class"] and CLASSES.index(ad["class"][-1]) > CLASSES.index(cl[0]):
|
||
if ad["class"][-1] == "h":
|
||
ad["address"] = re.sub(r"[мкдтпучасток]*\.? ?\d{1,4} ?\/*\d* ?", "",
|
||
ad["address"].lower())
|
||
elif ad["class"][-1] == "b":
|
||
num = re.findall(r"к{0,1}\.? ?\d", ad["address"])[-1]
|
||
ad["address"] = re.sub(num, "", ad["address"])
|
||
elif ad["class"][-1] == "e":
|
||
ad["address"] = re.sub(r"cтр\.? ?\d", "", ad["address"])
|
||
elif ad["class"][-1] == "l":
|
||
ad["address"] = re.sub(r"[литера]*\.? ?[А-Яа-я]{1}$", "", ad["address"])
|
||
elif ad["class"][-1] == "r":
|
||
ad["address"] = re.sub(r"пом\.? ?\d+", "", ad["address"])
|
||
ad["class"] = ad["class"][:-1]
|
||
return ad
|
||
|
||
|
||
def is_valid_token(string: str) -> bool:
|
||
return string not in ("", "уг.", "д.")
|
||
|
||
|
||
def create_token(obj: str = "", token_class: str = ""):
|
||
return pd.Series(
|
||
{
|
||
"obj": obj,
|
||
"class": token_class,
|
||
}
|
||
)
|
||
|
||
|
||
class AddressSplitter(Sequence):
|
||
def __init__(self, address: str):
|
||
self.input = address
|
||
|
||
self.addresses = self.split()
|
||
|
||
## Sequence abstract methods implementation
|
||
|
||
def __getitem__(self, key: int):
|
||
if key < len(self.addresses):
|
||
return self.addresses[key]
|
||
else:
|
||
raise IndexError()
|
||
|
||
def __len__(self):
|
||
return len(self.addresses)
|
||
|
||
## Address token class manipulations
|
||
|
||
def next_class(self) -> str:
|
||
return self.token["class"][0]
|
||
|
||
def correct_order(self) -> bool:
|
||
prev_class = self.accumulator["class"][-1]
|
||
|
||
return (
|
||
CLASSES.index(prev_class) < CLASSES.index(self.next_class())
|
||
and self.accumulator["class"] != "w"
|
||
)
|
||
|
||
def next_class_is(self, comparing_class: str) -> bool:
|
||
return len(self.token["class"]) > 0 and self.next_class() == comparing_class[0]
|
||
|
||
def has_no_class(self, comparing_class: str) -> bool:
|
||
return comparing_class[0] not in self.accumulator["class"]
|
||
|
||
def pop_token_class(self):
|
||
self.token["class"] = self.token["class"][1:]
|
||
|
||
## Accumulator constrains
|
||
|
||
def next_is_street_or_upper(self) -> bool:
|
||
is_unknown_class = self.accumulator["class"] in ("", "w")
|
||
|
||
return (
|
||
CLASSES.index(self.next_class()) <= CLASSES.index("s") or is_unknown_class
|
||
)
|
||
|
||
def has_numbered_street(self) -> bool:
|
||
return any_of_in(("-я", "-й", "-Я"), self.accumulator["address"])
|
||
|
||
## Accumulator manipulation
|
||
|
||
# House
|
||
|
||
def substitue_house(self) -> str:
|
||
house_regex = re.compile(r"\d{1,4} ?[\/\-]?\d* ?")
|
||
|
||
number = house_regex.findall(self.token['obj'])[0]
|
||
|
||
if self.has_numbered_street():
|
||
house_number_index = 1
|
||
else:
|
||
house_number_index = 0
|
||
|
||
number_in_accumulator = house_regex.findall(self.accumulator["address"])
|
||
|
||
if number_in_accumulator:
|
||
return re.sub(number_in_accumulator[house_number_index], number, self.accumulator["address"])
|
||
else:
|
||
return self.accumulator["address"]
|
||
|
||
# Building
|
||
|
||
def append_building(self, number: int) -> pd.Series:
|
||
self.accumulator["class"] += "b"
|
||
self.accumulator["address"] += "к." + number
|
||
|
||
return self.accumulator
|
||
|
||
def substitue_building(self, number: int) -> str:
|
||
return re.sub(r"\d$", number, self.accumulator["address"])
|
||
|
||
def insert_building(self):
|
||
number = re.findall(r"\d", self.token["obj"])[-1]
|
||
|
||
if number and self.has_no_class("building"):
|
||
self.accumulator = self.append_building(number)
|
||
else:
|
||
self.accumulator["address"] = self.substitue_building(number)
|
||
|
||
# Edifice
|
||
|
||
def substitue_edifice(self, number: int) -> str:
|
||
return re.sub(r"cтр\. ?\d", number, self.accumulator["address"].strip())
|
||
|
||
def insert_edifice(self):
|
||
number = re.findall("стр\.? ?\d", self.token["obj"])[-1]
|
||
|
||
self.accumulator["address"] = self.substitue_edifice(number)
|
||
|
||
if number and self.has_no_class("edifice"):
|
||
self.accumulator["class"] += "e"
|
||
|
||
# Letter
|
||
|
||
def without_letter(self) -> str:
|
||
return re.sub(r"[А-Яа-я]$", "", self.accumulator["address"].strip())
|
||
|
||
def substitue_letter(self, letter: str) -> str:
|
||
address_without_letter = self.without_letter()
|
||
|
||
return address_without_letter + letter
|
||
|
||
def insert_letter(self):
|
||
letter = re.findall(r"[А-Яа-я]", self.token["obj"])[-1]
|
||
self.accumulator["address"] = self.substitue_letter(letter)
|
||
|
||
if letter and self.has_no_class("litera"):
|
||
self.accumulator["class"] += "l"
|
||
|
||
def has_letter_in(self) -> bool:
|
||
return (
|
||
re.search(r"\d{1,3}([А-Я]|[а-я])( |$)", self.accumulator["address"])
|
||
)
|
||
|
||
# Room
|
||
|
||
def substitue_room(self, number: int) -> str:
|
||
return re.sub(r"пом\. ?\d\-?\d*\w?", number, self.accumulator["address"].strip())
|
||
|
||
def insert_room(self):
|
||
number = re.findall("пом\. ?\-?\d*\w?", self.token["obj"])[-1]
|
||
self.accumulator["address"] = self.substitue_room(number)
|
||
|
||
if number and self.has_no_class("room"):
|
||
self.accumulator["class"] += "r"
|
||
|
||
## Data preprocessing
|
||
|
||
def split_tokens(self) -> list[pd.Series]:
|
||
address = self.input.replace(";", ",")
|
||
|
||
parts = address.split(",")
|
||
parts = map(str.strip, parts)
|
||
parts = filter(is_valid_token, parts)
|
||
|
||
tokens = map(lambda part: create_token(part, ""), parts)
|
||
|
||
return list(tokens)
|
||
|
||
def split(self):
|
||
self.tokens = self.split_tokens()
|
||
|
||
result = []
|
||
|
||
self.accumulator = pd.Series({"address": "", "class": ""})
|
||
|
||
prev_token = create_token()
|
||
|
||
for cursor in self.tokens:
|
||
self.token = address_classification(cursor, prev_token)
|
||
prev_token = self.token.copy()
|
||
|
||
if self.accumulator["class"] == "":
|
||
self.accumulator = self.token.rename({"obj": "address"})
|
||
continue
|
||
|
||
if self.correct_order():
|
||
self.accumulator["address"] += " "
|
||
self.accumulator += self.token.rename({"obj": "address"})
|
||
else:
|
||
unfolded_address = unfold_house_ranges(self.accumulator["address"])
|
||
self.accumulator["address"] = unfolded_address[-1]
|
||
|
||
result.extend(unfolded_address)
|
||
|
||
self.accumulator = cut_address(self.accumulator, self.token["class"])
|
||
|
||
if self.next_is_street_or_upper():
|
||
self.accumulator = self.token.rename({"obj": "address"})
|
||
|
||
if self.next_class_is("house"):
|
||
self.accumulator["address"] = self.substitue_house()
|
||
self.pop_token_class()
|
||
|
||
if self.next_class_is("building"):
|
||
self.insert_building()
|
||
self.pop_token_class()
|
||
|
||
if self.next_class_is("edifice"):
|
||
self.insert_edifice()
|
||
self.pop_token_class()
|
||
|
||
if self.next_class_is("letter"):
|
||
self.insert_letter()
|
||
elif self.has_letter_in():
|
||
self.accumulator["address"] = self.without_letter()
|
||
|
||
if self.next_class_is("room"):
|
||
self.insert_room()
|
||
self.pop_token_class()
|
||
|
||
result.extend(unfold_house_ranges(self.accumulator["address"]))
|
||
|
||
return result
|
||
|
||
|
||
def split_pesoch_res(address: str) -> List[str]:
|
||
t = re.sub(r",", " ", address)
|
||
t = re.split(r"(Санкт-Петербург|Ленинградская обл|Л\.О)", t)
|
||
t = list(map(str.strip, filter(lambda token: token != "", t)))
|
||
tokens = [t[i] + " " + t[i+1] for i in range(0, len(t)-1, 2)]
|
||
|
||
if tokens:
|
||
return list(set(tokens))
|
||
return [address]
|
||
|
||
def process_row(row: pd.Series[str]) -> pd.Series[str]:
|
||
row = row.copy()
|
||
|
||
if pd.isnull(row["Улица"]):
|
||
row["Улица"] = [None]
|
||
else:
|
||
if row["РЭС"] == "Песочинский РЭС":
|
||
addresses = split_pesoch_res(row["Улица"])
|
||
else:
|
||
addresses = AddressSplitter(row["Улица"])
|
||
row["Улица"] = addresses
|
||
|
||
return row
|
||
|
||
|
||
def split_addresses(df: pd.DataFrame) -> pd.DataFrame:
|
||
merged_df = df.apply(process_row, axis=1).reset_index()
|
||
|
||
return merged_df.explode("Улица", ignore_index=True) |