Source code for eta_utility.connectors.sub_handlers

from __future__ import annotations

import csv
import pathlib
import queue
import re
import threading
from collections import deque
from contextlib import AbstractContextManager
from datetime import datetime
from logging import getLogger
from threading import Lock
from typing import TYPE_CHECKING

import numpy as np
import pandas as pd
from dateutil import tz

from eta_utility.connectors.node import Node

if TYPE_CHECKING:
    from collections.abc import Sequence
    from types import TracebackType
    from typing import Any, TextIO

    from typing_extensions import Self

    from eta_utility.type_hints import Number, Path, TimeStep

from .base_classes import SubscriptionHandler

log = getLogger(__name__)


[docs] class MultiSubHandler(SubscriptionHandler): """The MultiSubHandler can be used to distribute subscribed values to multiple different subscription handlers. The handlers can be registered using the register method. """ def __init__(self) -> None: super().__init__() self._handlers: list = []
[docs] def register(self, sub_handler: SubscriptionHandler) -> None: """Register a subscription handler. :param SubscriptionHandler sub_handler: SubscriptionHandler object to use for handling subscriptions. """ if not isinstance(sub_handler, SubscriptionHandler): raise TypeError("Subscription Handler should be an instance of the SubscriptionHandler class.") self._handlers.append(sub_handler)
[docs] def push(self, node: Node, value: Any, timestamp: datetime | None = None) -> None: """Receive data from a subscription. This should contain the node that was requested, a value and a timestamp when data was received. Push data to all registered sub-handlers. :param node: Node object the data belongs to. :param value: Value of the data. :param timestamp: Timestamp of receiving the data. """ for handler in self._handlers: handler.push(node, value, timestamp)
[docs] def close(self) -> None: """Finalize and close all subscription handlers.""" for handler in self._handlers: try: handler.close() except Exception: pass
[docs] class CsvSubHandler(SubscriptionHandler): """Handle data for a subscription and save it as a CSV file. :param output_file: CSV file to write data to. :param write_interval: Interval between rows in the CSV file (value that time is rounded to) :param size_limit: Size limit for the csv file. A new file with a unique name will be created when the size is exceeded. :param dialect: Dialect of the csv file. This takes objects, which correspond to the csv.Dialect interface from the python csv module. """ def __init__( self, output_file: Path, write_interval: TimeStep = 1, size_limit: int = 1024, dialect: type[csv.Dialect] = csv.excel, ) -> None: super().__init__(write_interval=write_interval) # Create the csv file handler object which writes data to disc self._csv_file = _CSVFileDB(output_file, size_limit, dialect) # Enable propagation of exceptions self.exc: BaseException | None = None # Create the queue and thread self._queue: queue.Queue = queue.Queue() self._thread: threading.Thread = threading.Thread(target=self._run) self._thread.start()
[docs] def push(self, node: Node, value: Any, timestamp: datetime | None = None) -> None: """Receive data from a subscription. THis should contain the node that was requested, a value and a timestamp when data was received. If the timestamp is not provided, current time will be used. :param node: Node object the data belongs to. :param value: Value of the data. :param timestamp: Timestamp of receiving the data. """ timestamp = timestamp if timestamp is not None else datetime.now() self._queue.put_nowait((node, value, timestamp)) # Reraise exceptions if any if self.exc: raise self.exc
def _run(self) -> None: """Take data from the queue, preprocess it and write to output file.""" cancelled = False with self._csv_file as f: try: while True: try: data = self._queue.get_nowait() except queue.Empty: if not cancelled: continue # Check for sentinel and finalize thread after completing open tasks if data is None: self._queue.task_done() cancelled = True continue if cancelled is True and self._queue.empty(): break node, value, timestamp = data # Make sure not to send lists to the file handler if not isinstance(value, str) and hasattr(value, "__len__"): value = self._format_list(value) if hasattr(timestamp, "__len__"): timestamp = timestamp[0] timestamp = self._round_timestamp(timestamp).astimezone(self._local_tz) f.write(timestamp, node.name, value) self._queue.task_done() except BaseException as e: self.exc = e
[docs] def close(self) -> None: """Finalize and close the subscription handler.""" # Reraise exceptions if any if self.exc: raise self.exc self._queue.put_nowait(None) self._queue.join() self._thread.join()
def _format_list(self, value: Any) -> str: delim = ";" if self._csv_file.dialect.delimiter == "," else "," return repr(value).replace(self._csv_file.dialect.delimiter, delim)
class _CSVFileDB(AbstractContextManager): """Handle CSV file content. :param file: Path to the csv file. :param file_size_limit: Size limit for the file in MB. A new file will be created, once the limit is exceeded. :param dialect: Dialect of the csv file. This takes objects, which correspond to the csv.Dialect interface from the python csv module. """ def __init__( self, file: Path, file_size_limit: int = 1024, dialect: type[csv.Dialect] = csv.excel, ): #: Path to the file that is being written to. self.filepath: pathlib.Path = file if isinstance(file, pathlib.Path) else pathlib.Path(file) #: File descriptor. self._file: TextIO | None = None #: Size limit for written files in bytes. self.file_size_limit: int = file_size_limit * 1024 * 1024 #: CSV dialect to be used for reading and writing data. self.dialect: type[csv.Dialect] = dialect #: List of header fields. self._header: list[str] = [] #: Ending position of the header in the file stream (used for extending the header). self._endof_header: int = 0 #: Write buffer. self._buffer: deque[dict[str, str]] = deque() self._timebuffer: deque[datetime] = deque() #: Latest timestamp in the write-buffer. self._latest_ts: datetime = datetime.fromtimestamp(10000, tz=tz.tzlocal()) #: Latest known value for each of the names in the header. self._latest_values: dict[str, str] = {} #: Length of the line terminator in bytes (for finding file positions). self._len_lineterminator: int = len(bytes(self.dialect.lineterminator, "UTF-8")) self.check_file(exclusive_creation=False) self.check_valid_csv() def __enter__(self) -> Self: """Enter the context managed file database.""" self._open_file() return self def _open_file(self, exclusive_creation: bool = False) -> None: """Open a new file and check whether it is writable. If the file exists, try to figure out the dialect and header of the existing file. :param exclusive_creation: Set to True, to request exclusive creation of a new file. If set to False, an existing file may be updated. """ self.check_file(exclusive_creation) self.check_valid_csv() assert self._file is not None # Go to the end to get ready for updating/extending the file. self._file.seek(0, 2) def check_valid_csv(self) -> None: """Check whether the file is a valid csv file.""" assert self._file is not None, "Open a file before calling check_valid_csv." # If the file is not empty, go to the beginning and try to figure out, whether existing data could be extended. if self._file.readline() in "": valid = True self._header = ["Timestamp"] self._write_file(self._header) self._endof_header = self._file.tell() - self._len_lineterminator log.debug(f"The '.csv' file was empty, dialect set to {self.dialect}, started writing header.") else: self._file.seek(0) # Read a maximum of 30 lines from the file to use as a sample for figuring out, whether the file # is valid csv sample_lines = [] for _ in range(30): sample_lines.append(self._file.readline()) if sample_lines[-1] == "": break sample = "".join(sample_lines) try: valid = csv.Sniffer().has_header(sample) self.dialect = csv.Sniffer().sniff(sample) self.dialect.delimiter = "," if self.dialect.delimiter not in {",", ";"} else self.dialect.delimiter self._len_lineterminator = len(bytes(self.dialect.lineterminator, "UTF-8")) except csv.Error: valid = False self.dialect = csv.excel # Determine the header of the existing csv file self._file.seek(0) self._header = list(re.sub(r"\s+", "", self._file.readline()).split(self.dialect.delimiter)) self._endof_header = self._file.tell() - self._len_lineterminator if not valid: raise ValueError(f"Output file for writing to '.csv' is not a valid '.csv' file: {self.filepath}.") def check_file(self, exclusive_creation: bool = False) -> None: # Try opening or creating the specified file. try: if exclusive_creation: raise FileNotFoundError self._file = self.filepath.open("r+t", newline="", encoding="UTF-8") log.debug(f"Opened existing '.csv' file for updating: {self.filepath}.") except FileNotFoundError: try: self._file = self.filepath.open("x+t", newline="", encoding="UTF-8") log.debug(f"Created a new '.csv' file: {self.filepath}.") except OSError as e: raise OSError(f"Unable to read or write the requested '.csv' file: {self.filepath}.") from e # Check whether the file is accessible in the required ways. if self._file is None or not self._file.readable() or not self._file.seekable() or not self._file.writable(): raise ValueError("Output file for writing to '.csv' is not readable or writable.") log.debug("Successfully verified full '.csv' file access") def _write_file(self, field_list: list[str], insert_pos: int | None = None) -> int: """Write data to the file. :param field_list: List of strings to be inserted into the csv file. :param insert_pos: Position to insert the fields (stream position). If None, insertion will be at end of file. :return: Ending position of the last insertion (stream position). """ # Check whether the file is accessible in the required ways. if self._file is None or not self._file.readable() or not self._file.seekable() or not self._file.writable(): raise ValueError("Output file for writing to '.csv' is not readable or writable.") if insert_pos is None: string = self.dialect.delimiter.join(field_list) + self.dialect.lineterminator try: self._file.write(string) except ValueError: self._open_file() self._file.write(string) pos = self._file.tell() else: # When inserting, everything else in the file must be moved along as well # (otherwise it would be overwritten). Therefore, a chunk of the file is read before inserting string = self.dialect.delimiter + self.dialect.delimiter.join(field_list) chunksize = max(100000, len(string)) self._file.seek(insert_pos) chunk = self._file.read(chunksize) nextpos_read = self._file.tell() self._file.seek(insert_pos) self._file.write(string) pos = nextpos_insert = self._file.tell() while nextpos_insert < nextpos_read: self._file.seek(nextpos_read) newchunk = self._file.read(chunksize) nextpos_read = self._file.tell() # Insert last chunk into file and store next chunk self._file.seek(nextpos_insert) self._file.write(chunk) # Store current position for next insertion, then read next chunk nextpos_insert = self._file.tell() chunk = newchunk self._file.seek(nextpos_insert) self._file.write(chunk) return pos def write( self, timestamp: datetime | None = None, name: str | None = None, value: Number | None = None, flush: bool = False, _len_buffer: int = 20, ) -> None: """Write value to the file and manage the data buffer. :param timestamp: Timestamp of the value to be written (can be empty if only flushing the buffer is intended). :param name: Name/Header for the value to be written (can be empty if only flushing the buffer is intended). :param value: Value to be written to the file (can be empty if only flushing the buffer is intended). :param flush: Flush the entire buffer to file if set to True. :param _len_buffer: Length of the buffer in lines. Does not usually need to be changed. """ if self._file is None: raise RuntimeError("Enter context manager before trying to write to CSVFileDB.") # Check whether the file size limit is exceeded to initiate switching to a new file. size_limit_exceeded = self.filepath.stat().st_size > self.file_size_limit # Determine how large the buffer should be, depending on whether data is being flushed to file or preparing to # switch to a new file. if flush: buffer_target = 0 elif size_limit_exceeded and not len(self._buffer) >= 2 * _len_buffer: buffer_target = 2 * _len_buffer log.debug("Preparing to switch files due to exceeded CSV file size limit.") else: buffer_target = _len_buffer # New values are inserted into the buffer, depending on their timestamp if timestamp is not None and name is not None and value is not None: # Extend header, if the value does not exist there yet if name not in self._header: self._header.append(name) self._endof_header = self._write_file([name], self._endof_header) # Find out, where to insert the current timestamp if len(self._timebuffer) == 0 or self._timebuffer[-1] < timestamp: # If the new timestamp occurred later than the latest timestamp self._buffer.append({name: str(value)}) self._timebuffer.append(timestamp) elif self._timebuffer[-1] == timestamp: # If the timestamp is equal to the latest timestamp self._buffer[-1].update({name: str(value)}) elif self._timebuffer[0] > timestamp: # If the new timestamp occurred before the earliest buffered timestamp self._buffer.appendleft({name: str(value)}) self._timebuffer.appendleft(timestamp) log.debug(f"Buffer time for CSV file exceeded, older value received with {timestamp}") else: # If none of the special cases above apply, search through the time buffer to figure out, where to # insert the value last_ts = self._timebuffer[0] for idx, ts in enumerate(self._timebuffer): if timestamp == ts: self._buffer[idx].update({name: str(value)}) break if last_ts < timestamp < ts: self._buffer.insert(idx, {name: str(value)}) self._timebuffer.insert(idx, timestamp) break last_ts = ts # Write any rows in the buffer which exceed the size of buffer_target to the file. while len(self._buffer) >= buffer_target and len(self._buffer) > 0: row = self._buffer.popleft() row[self._header[0]] = self._timebuffer.popleft().strftime("%Y-%m-%d %H:%M:%S.%f") processed_row: list[str] = [""] * len(self._header) for idx, col in enumerate(self._header): if col in row: v = self._latest_values[col] = row[col] processed_row[idx] = str(v) else: processed_row[idx] = str(self._latest_values.get(col, "")) log.debug(f"Writing line with index {processed_row[0]} to CSV file .") self._write_file(processed_row) # Close current file and create a new file with a different name if the size limit was exceeded. if size_limit_exceeded and buffer_target <= _len_buffer: log.info(f"CSV File size limit exceeded. Closing current file {self.filepath}.") self._file.close() self._file = None self.filepath = self.filepath.with_name(f"{self.filepath.stem}_{datetime.now().strftime('%y%m%d%H%M')}.csv") self._open_file(exclusive_creation=True) if flush: self.write(flush=True) def __exit__( self, __exc_type: type[BaseException] | None, __exc_value: BaseException | None, __traceback: TracebackType | None, ) -> None: """Exit the context manager :param exc_details: Execution details """ if self._file is not None: self.write(flush=True) self._file.close()
[docs] class DFSubHandler(SubscriptionHandler): """Subscription handler for returning pandas.DataFrames when requested. :param write_interval: Interval between index values in the data frame (value to which time is rounded). :param size_limit: Number of rows to keep in memory. :param auto_fillna: If True, missing values in self._data are filled with the pandas-method df.ffill() each time self.data is called. """ def __init__(self, write_interval: TimeStep = 1, size_limit: int = 100, auto_fillna: bool = True) -> None: super().__init__(write_interval=write_interval) self._data: pd.DataFrame = pd.DataFrame() self._data_lock: threading.Lock = Lock() self.keep_data_rows: int = size_limit self.auto_fillna: bool = auto_fillna
[docs] def push( self, node: Node, value: Any | pd.Series | Sequence[Any], timestamp: datetime | pd.DatetimeIndex | TimeStep | None = None, ) -> None: """Append values to the dataframe. :param node: Node object the data belongs to. :param value: Value of the data or Series of values. There must be corresponding timestamps for each value. :param timestamp: Timestamp of receiving the data or DatetimeIndex if pushing multiple values. Alternatively an integer/timedelta can be provided to determine the interval between data points. Use negative numbers to describe past data. Integers are interpreted as seconds. If value is a pd.Series and has a pd.DatetimeIndex, timestamp is ignored. """ # Check if node.name is in _data.columns self._data_lock.acquire() if node.name not in self._data.columns: self._data[node.name] = np.nan self._data_lock.release() # Multiple values if not isinstance(value, str) and hasattr(value, "__len__"): value = self._convert_series(value, timestamp) # Push Series # Values are rounded to self.write_interval in _convert_series for _timestamp, _value in value.items(): _timestamp = self._assert_tz_awareness(_timestamp) self._data_lock.acquire() # Replace NaN with -inf to distinguish between the 'real' NaN and the 'fill' NaN if pd.isna(_value): _value = -np.inf self._data.loc[_timestamp, node.name] = _value self._data_lock.release() # Single value else: if not isinstance(timestamp, datetime) and timestamp is not None: raise ValueError("Timestamp must be a datetime object or None.") timestamp = self._round_timestamp(timestamp if timestamp is not None else datetime.now()) self._data_lock.acquire() # Replace NaN with -inf to distinguish between the 'real' NaN and the 'fill' NaN if pd.isna(value): value = -np.inf self._data.loc[timestamp, node.name] = value self._data_lock.release() # Housekeeping (Keep internal data short) self._housekeeping()
[docs] def get_latest(self) -> pd.DataFrame | None: """Return a copy of the dataframe, this ensures they can be worked on freely. Returns None if data is empty.""" self._data_lock.acquire() if len(self._data.index) == 0: self._data_lock.release() return None # If no data in self._data, return None self._data_lock.release() return self.data.iloc[[-1]]
@property def data(self) -> pd.DataFrame: """This contains the interval dataframe and will return a copy of that.""" self._data_lock.acquire() if self.auto_fillna: self._data = self._data.ffill() data = self._data.replace(-np.inf, np.nan, inplace=False) self._data_lock.release() return data
[docs] def reset(self) -> None: """Reset the internal data and restart collection.""" self._data_lock.acquire() self._data = pd.DataFrame() self._data_lock.release() log.info(f"Subscribed DataFrame {hash(self._data)} was reset successfully.")
def _housekeeping(self) -> None: """Keep internal data short by only keeping last rows as specified in self.keep_data_rows.""" self._data_lock.acquire() self._data = self._data.drop(index=self._data.index[: -self.keep_data_rows]) self._data_lock.release()
[docs] def close(self) -> None: """This is just here to satisfy the interface, not needed in this case.""" pass