This repository has been archived on 2026-03-09. You can view files and clone it. You cannot open issues or pull requests or push a commit.
Files
jeb-server-utils/src/jeb_server_utils/_core.py
T
2026-02-28 02:45:16 +01:00

510 lines
20 KiB
Python

# Copyright 2026 jCloud Services GbR
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# https://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from jeb_utils import exceptions, jeb_utils, utils
import os
import dbm
import config_parser
import shutil
import time
import binascii
import warnings
__all__ = [
'JEBServerUtils',
'Segment',
'Topic',
]
def _create_file_if_not_exists(path: str, content: bytes = b''):
'''
Creates a file if it does not exist.
:param path: The file path
:type path: str
:param content: Optional, the content of the file if it is newly created.
:type content: bytes
'''
if not os.path.exists(path):
with open(path, 'wb') as f:
f.write(content)
f.close()
class JEBServerUtils:
def __init__(self, data_dir: str, *, segment_max_size: int = 2 ** 32, fetch_records_chunk_size: int = 1024 ** 2):
'''
The utilities for the jeb server.
'''
self.data_dir = data_dir
self.segment_max_size = segment_max_size
self.fetch_records_chunk_size = fetch_records_chunk_size
self._log_end_offsets = {}
self._init()
def _init(self):
self._create_base_segment_files()
self._init_log_end_offsets()
def _create_base_segment_files(self):
'''
Creates the base segment files.
'''
for topic, v in self.get_topics().items():
partitions = jeb_utils.format_topic_partitions(topic, v['partitions'])
for p in partitions:
_create_file_if_not_exists(f'{self.data_dir}/topics/{p}/' + '0' * 16 + '.log')
_create_file_if_not_exists(f'{self.data_dir}/topics/{p}/' + '0' * 16 + '.index')
_create_file_if_not_exists(f'{self.data_dir}/topics/{p}/' + '0' * 16 + '.timeindex')
def _init_log_end_offsets(self):
'''
Initializes the log end offsets.
'''
global _log_end_offsets
topics = [f'{t}{jeb_utils.TOPIC_PARTITION_SEPARATOR}{p:02d}' for t, v in self.get_topics().items() for p in v['partitions']]
topic_segments = {t: Topic(t, self).segments for t in topics}
for topic, segments in topic_segments.items():
log_end_offset = 0
for segment in segments:
with open(f'{self.data_dir}/topics/{topic}/{segment.base_offset:016x}.log', 'rb') as logfile:
logfile.seek(0, os.SEEK_END)
file_size = logfile.tell()
logfile.seek(0)
while True:
record_size_bytes = logfile.read(8)
if not record_size_bytes or len(record_size_bytes) < 8:
break
record_size = int.from_bytes(record_size_bytes)
if logfile.tell() + record_size > file_size:
break
logfile.seek(record_size, os.SEEK_CUR)
log_end_offset += 1
logfile.close()
self._log_end_offsets[topic] = log_end_offset
def get_segment_base_timestamp(self, topic: str, segment: int) -> int:
'''
Returns the base timestamp of a segment.
:param topic: The topic
:type topic: str
:param segment: The segment base offset
:type segment: int
:return: The base timestamp (milliseconds since 1970)
:rtype: int
'''
with open(f'{self.data_dir}/topics/{topic}/{segment:016x}.log', 'rb') as logfile:
# go to the topic length
logfile.seek(20)
# skip topic
topic_length = int.from_bytes(logfile.read(2))
logfile.seek(topic_length + 1, os.SEEK_CUR)
timestamp_bytes = logfile.read(8)
if len(timestamp_bytes) < 8:
timestamp = None
else:
timestamp = int.from_bytes(timestamp_bytes)
return timestamp
def get_topics(self):
'''
Returns all topics.
:return: The topics
:rtype: dict
'''
with dbm.open(f'{self.data_dir}/conf/topics', 'c') as db:
topics = {k.decode(): config_parser.parse.json.parse_json(v.decode()) for k, v in db.items()}
db.close()
return topics
def mktopic(self, topic_name: str):
'''
Creates a topic.
:param topic_name: The name of the topic
:type topic_name: str
'''
with dbm.open(f'{self.data_dir}/conf/topics', 'c') as db:
if topic_name.encode() in db.keys():
raise exceptions.TopicExistsError(f'topic \'{topic_name}\' already exists')
db[topic_name] = config_parser.serialize.json.serialize({'partitions': [1]})
db.close()
os.mkdir(f'{self.data_dir}/topics/{topic_name}{jeb_utils.TOPIC_PARTITION_SEPARATOR}01/')
def rmtopic(self, topic_name: str):
'''
Removes a topic.
:param topic_name: The topic name
:type topic_name: str
:raises TopicNotFoundError: If the topic does not exist.
'''
with dbm.open(f'{self.data_dir}/conf/topics', 'c') as db:
if topic_name.encode() not in db.keys():
raise exceptions.TopicNotFoundError(f'topic \'{topic_name}\' does not exist')
partitions = config_parser.parse.json.parse_json(db[topic_name.encode()].decode())['partitions']
del db[topic_name]
db.close()
partition_dirs = jeb_utils.format_topic_partitions(topic_name, partitions)
for pd in partition_dirs:
try:
shutil.rmtree(f'{self.data_dir}/topics/{pd}/')
except:
pass
def mktopicpart(self, topic_name: str, partition_number: int):
'''
Creates a partition of a topic.
:param topic_name: The topic name
:type topic_name: str
:param partition_number: The partition number
:type partition_number: int
'''
if partition_number < 1 or partition_number > 99:
raise ValueError('partition number must be between 1 and 99')
formatted_partition_number = f'{(int(partition_number)):02d}'
with dbm.open(f'{self.data_dir}/conf/topics', 'c') as db:
if topic_name.encode() not in db.keys():
raise exceptions.TopicNotFoundError(f'topic \'{topic_name}\' does not exist')
partitions = config_parser.parse.json.parse_json(db[topic_name.encode()].decode())['partitions']
if partition_number in partitions:
raise exceptions.TopicPartitionExistsError(f'partition \'{formatted_partition_number}\' already exists for topic \'{topic_name}\'')
db[topic_name] = config_parser.serialize.json.serialize({'partitions': sorted(list(set(partitions) | {int(formatted_partition_number)}))})
db.close()
os.mkdir(f'{self.data_dir}/topics/{topic_name}{jeb_utils.TOPIC_PARTITION_SEPARATOR}{formatted_partition_number}/')
def rmtopicpart(self, topic_name: str, partition_number: int):
'''
Removes a partition of a topic.
:param topic_name: The topic name
:type topic_name: str
:param partition_number: The partition number
:type partition_number: int
'''
formatted_partition_number = f'{(int(partition_number)):02d}'
with dbm.open(f'{self.data_dir}/conf/topics', 'c') as db:
if topic_name.encode() not in db.keys():
raise exceptions.TopicNotFoundError(f'topic \'{topic_name}\' does not exist')
partitions = config_parser.parse.json.parse_json(db[topic_name.encode()].decode())['partitions']
if partition_number not in partitions:
raise exceptions.TopicPartitionNotFoundError(f'partition \'{formatted_partition_number}\' does not exist for topic \'{topic_name}\'')
db[topic_name] = config_parser.serialize.json.serialize({'partitions': sorted(list(set(partitions) - {int(formatted_partition_number)}))})
db.close()
shutil.rmtree(f'{self.data_dir}/topics/{topic_name}{jeb_utils.TOPIC_PARTITION_SEPARATOR}{formatted_partition_number}/')
def check_topic_exists(self, topic: str):
'''
Checks whether a topic exists.
:param topic: The topic name
:type topic: str
:raises TopicNotFoundError: If the topic does not exist
:raises TopicPartitionNotFoundError: If the partition does not exist
'''
topics = self.get_topics()
topic_name, formatted_partition_number = topic.split(jeb_utils.TOPIC_PARTITION_SEPARATOR)
raw_partition_number = int(formatted_partition_number)
if topic_name not in topics:
raise exceptions.TopicNotFoundError(f'topic \'{topic_name}\' does not exist')
if raw_partition_number not in topics[topic_name]['partitions']:
raise exceptions.TopicPartitionNotFoundError(f'partition \'{formatted_partition_number}\' does not exist for topic \'{topic_name}\'')
def create_record(self, topic: str, timestamp: int, record_data: bytes, key: bytes = b'', compression_type: int = 0, headers: dict = {}):
'''
Creates a record.
:param topic: The topic name
:type topic: str
:param timestamp: The timestamp.
:type timestamp: int
:param record_data: The payload
:type record_data: bytes
:param key: The key
:type key: bytes
:param compression_type: The compression type
:type compression_type: int
:param headers: The header dictionary
:type headers: dict
'''
topic_name, formatted_partition_number = topic.split(jeb_utils.TOPIC_PARTITION_SEPARATOR)
if not jeb_utils.validate_topic_name(topic_name):
raise exceptions.FormatError(f'invalid topic name: \'{topic}\'')
self.check_topic_exists(topic)
topic_dir = f'{self.data_dir}/topics/{topic_name}{jeb_utils.TOPIC_PARTITION_SEPARATOR}{formatted_partition_number}/'
utils.create_file_if_not_exists(topic_dir + '0' * 16 + '.log')
segments = sorted([int(f.split('.')[0], 16) for f in os.listdir(topic_dir) if f.endswith('.log') and utils.is_number(f.split('.')[0], 16) and len(f.split('.')[0]) == 16])
last_segment = max(segments)
if timestamp == 0:
timestamp = int(time.time() * 1000)
timestamp_type = 1
else:
timestamp_type = 0
# Assemble record
record = b''
record += self._log_end_offsets[topic].to_bytes(8) # Offset
record += binascii.crc32(record_data).to_bytes(4) # CRC
record += (len(topic)).to_bytes(2) # Topic name length
record += topic.encode() # Topic
record += jeb_utils.AttributesByte(compression_type, timestamp_type).to_byte() # Attributes
record += timestamp.to_bytes(8) # Timestamp
record += (len(key)).to_bytes(4) # Key Length
record += key # Key
record += (len(record_data)).to_bytes(4) # Content Length
record += record_data # Content
record += (len(headers)).to_bytes(1)
for key, value in headers.items():
record += len(key).to_bytes(4)
record += key
record += len(value).to_bytes(4)
record += value
record = len(record).to_bytes(8) + record # Prepend Record Size
if os.path.getsize(topic_dir + f'{last_segment:016x}.log') + len(record) > self.segment_max_size:
last_segment = self._log_end_offsets[topic]
with open(topic_dir + f'{last_segment:016x}.log', 'ab') as logfile:
logfile.write(record)
logfile.close()
utils.create_file_if_not_exists(topic_dir + f'{last_segment:016x}.index')
utils.create_file_if_not_exists(topic_dir + f'{last_segment:016x}.timeindex')
if self._log_end_offsets[topic] % 1024 == 0:
index = self._log_end_offsets[topic].to_bytes(8) + os.path.getsize(topic_dir + f'{last_segment:016x}.log').to_bytes(4)
timeindex = timestamp.to_bytes(8) + self._log_end_offsets[topic].to_bytes(8)
with open(topic_dir + f'{last_segment:016x}.index', 'ab') as idxfile:
idxfile.write(index)
idxfile.close()
with open(topic_dir + f'{last_segment:016x}.timeindex', 'ab') as timeidxfile:
timeidxfile.write(timeindex)
timeidxfile.close()
self._log_end_offsets[topic] += 1
def fetch_records(self, topic: str, start_type: int, start: int, max_bytes: int):
'''
Fetches records from the topic and yields them in chunks.
:param topic: The topic name
:type topic: str
:param start_type: The start type
:type start_type: int
:param start: The start
:type start: int
:param max_bytes: The maximum of bytes
:type max_bytes: int
'''
topic = Topic(topic, self)
topic_name, formatted_partition_number = topic.topic_name.split(jeb_utils.TOPIC_PARTITION_SEPARATOR)
if not jeb_utils.validate_topic_name(topic_name):
raise exceptions.FormatError(f'invalid topic name: \'{topic.topic_name}\'')
if start_type not in (jeb_utils.FETCH_START_TYPE_OFFSET, jeb_utils.FETCH_START_TYPE_TIMESTAMP):
raise ValueError('invalid start type')
self.check_topic_exists(topic.topic_name)
segments = [int(f.split('.')[0], 16) for f in os.listdir(f'{self.data_dir}/topics/{topic.topic_name}/') if utils.is_number(f.split('.')[0], 16) and len(f.split('.')[0]) == 16 and f.endswith('.log')]
if start_type == jeb_utils.FETCH_START_TYPE_TIMESTAMP:
segment_base_timestamps = {s: s.base_timestamp for s in topic.segments}
s = {v: k for k, v in segment_base_timestamps.items()}[utils.find_nearest_lower_number(list(segment_base_timestamps.values()), start) or min(segment_base_timestamps.values())]
with open(f'{self.data_dir}/topics/{topic.topic_name}/{s.base_offset:016x}.timeindex', 'rb') as f:
timestamps = {}
f.seek(0)
while True:
timestamp_bytes = f.read(8)
if len(timestamp_bytes) < 8:
break
timestamp = int.from_bytes(timestamp_bytes)
offset_bytes = f.read(8)
if len(offset_bytes) < 8:
break
offset = int.from_bytes(offset_bytes)
timestamps[timestamp] = offset
f.close()
start = timestamps[utils.find_nearest_lower_number(timestamps.keys(), start) or min(timestamps.keys())]
if start < 0 or start >= self._log_end_offsets[topic.topic_name]:
# await send_block_function(b'\xb1\x31')
yield b'\xb1\x31'
return
idx_offset = jeb_utils.get_next_lower_index_entry(start)
start_seg = utils.find_nearest_lower_number(segments, start)
with open(f'{self.data_dir}/topics/{topic.topic_name}/' + f'{start_seg:016x}' + '.index', 'rb') as idxfile:
idxfile.seek(0, os.SEEK_END)
file_size = idxfile.tell()
idxfile.seek(idx_offset // 1024 * 12)
index_entry = idxfile.read(12)
idx_entry_pos = int.from_bytes(index_entry[8:])
if int.from_bytes(index_entry[:8]) != idx_offset:
warnings.warn(f'Index entry offset mismatch: expected {idx_offset}, got {int.from_bytes(index_entry[:8])}', jeb_utils.FileCorruptWarning)
idxfile.close()
with open(f'{self.data_dir}/topics/{topic.topic_name}/{start_seg:016x}.log', 'rb') as logfile:
logfile.seek(idx_entry_pos)
offset = idx_offset
record_position = 0
while offset < start:
record_length_bytes = logfile.read(8)
if len(record_length_bytes) < 8:
break
record_length = int.from_bytes(record_length_bytes)
offset_bytes = logfile.read(8)
if len(offset_bytes) < 8:
break
offset = int.from_bytes(offset_bytes)
record_position = logfile.tell() - 16
logfile.seek(record_length - 8, os.SEEK_CUR)
logfile.close()
# await send_block_function(b'\xa0\x30')
yield b'\xa0\x30'
segments = sorted(segments)
bytes_fetched = 0
for i, segment in enumerate(segments[segments.index(start_seg):]):
records_sent = 0
with open(f'{self.data_dir}/topics/{topic.topic_name}/' + f'{segment:016x}' + '.log', 'rb') as logfile:
if i == 0:
logfile.seek(record_position)
else:
logfile.seek(0)
while bytes_fetched <= max_bytes:
record_size_bytes = logfile.read(8)
if len(record_size_bytes) < 8:
break
record_size = int.from_bytes(record_size_bytes)
if bytes_fetched + 8 + record_size > max_bytes:
break
to_send = record_size_bytes
for _ in range(record_size // self.fetch_records_chunk_size):
to_send += logfile.read(self.fetch_records_chunk_size)
# await send_block_function(b'\x01' + to_send)
yield b'\x01' + to_send
to_send = b''
to_send += logfile.read(record_size % self.fetch_records_chunk_size)
# await send_block_function(b'\x01' + to_send)
yield b'\x01' + to_send
bytes_fetched += 8 + record_size
records_sent += 1
logfile.close()
if bytes_fetched >= max_bytes:
break
# await send_block_function(b'\x00') # EOF
yield b'\x00' # EOF
class Segment:
def __init__(self, topic: str, base_offset: int, jeb_server_utils: JEBServerUtils):
'''
Represents a segment of a topic.
:param topic: The topic name
:type topic: str
:param base_offset: The base offset of the segment
:type base_offset: int
:param jeb_server_utils: The ``JEBServerUtils`` instance
:type jeb_server_utils: JEBServerUtils
'''
self.topic = topic
self.base_offset = base_offset
self.jeb_server_utils = jeb_server_utils
self.base_timestamp = self.get_base_timestamp()
def __repr__(self):
return f'Segment({self.topic}, {self.base_offset})'
def get_base_timestamp(self):
'''
Returns the base timestamp of the segment.
:return: The base timestamp (milliseconds since 1970)
:rtype: int
'''
# with open(f'{self.jeb_server_utils.data_dir}/topics/{self.topic}/{self.base_offset:016x}.timeindex', 'rb') as logfile:
# timestamp_bytes = logfile.read(8)
# if len(timestamp_bytes) < 8:
# timestamp = None
# else:
# timestamp = int.from_bytes(timestamp_bytes)
# return timestamp
return self.jeb_server_utils.get_segment_base_timestamp(self.topic, self.base_offset)
class Topic:
def __init__(self, topic_name: str, jeb_server_utils: JEBServerUtils):
'''
Represents a topic.
:param topic_name: The name of the topic
:type topic_name: str
:param jeb_server_utils: The ``JEBServerUtils`` instance
:type jeb_server_utils: JEBServerUtils
'''
self.topic_name = topic_name
self.jeb_server_utils = jeb_server_utils
def __repr__(self):
return f'Topic({self.topic_name})'
@property
def segments(self):
'''
Returns the segments of the topic.
:return: The segments
:rtype: list
'''
return [Segment(self.topic_name, bo, self.jeb_server_utils) for bo in sorted([int(s.split('.')[0], 16) for s in os.listdir(f'{self.jeb_server_utils.data_dir}/topics/{self.topic_name}') if s.endswith('.log') and utils.is_number(s.split('.')[0], 16) and len(s.split('.')[0]) == 16])]