%PDF- %PDF-
Mini Shell

Mini Shell

Direktori : /proc/self/root/proc/self/root/opt/cloudlinux/venv/lib/python3.11/site-packages/clcommon/
Upload File :
Create Path :
Current File : //proc/self/root/proc/self/root/opt/cloudlinux/venv/lib/python3.11/site-packages/clcommon/clfunc.py

# -*- coding: utf-8 -*-

# Copyright © Cloud Linux GmbH & Cloud Linux Software, Inc 2010-2018 All Rights Reserved
#
# Licensed under CLOUD LINUX LICENSE AGREEMENT
# http://cloudlinux.com/docs/LICENSE.TXT
#

import os
import re
import signal
import subprocess
import sys

LVE_FILE = '/proc/lve/list'


# GET VERSION from /proc/lve/list
def get_lve_version():
    """
    Obtain lve process filesystem version
    """
    try:
        with open(LVE_FILE, encoding='utf-8') as f:
            line = f.read(3)  # we need only first three symbols for parse version
        lve_procfs_version = [int(line.rsplit(':', 1)[0]), 'OK']
    except IOError:
        lve_procfs_version = [None, f'clcommon: get_lve_version: Can`t open file {LVE_FILE}']
    except IndexError:
        lve_procfs_version = [None, f'clcommon: get_lve_version: Can`t get data from {LVE_FILE}']
    except ValueError:
        lve_procfs_version = [None, f'clcommon: get_lve_version: Can`t parse {LVE_FILE}']

    return lve_procfs_version


BYTES_CONVERSION_TABLE = {
    'K': 1,
    'M': 1024,
    'G': 1024 * 1024,
    'T': 1024 * 1024 * 1024
}


def validate_cpu(val: str | int) -> str | int | None:
    """
    Check that the value is a valid CPU limit (0-100 int or speed (% or MHz/GHz)).
    Return the value if valid, otherwise None.
    If the value contains decimal part, return only integer part.
    """
    data = str(val)

    # Validate integer CPU limits
    regexp_int = re.compile(r'^([1-9][0-9]?|100)(?:\.(\d+))?$')
    regex_match = regexp_int.match(data)
    if regex_match:
        if isinstance(val, (float, int)):
            return int(val)
        return regex_match.group(1)  # return only integer part

    # Validate percentage speeds
    regexp_speedp = re.compile(r'^([1-9]|[1-9][0-9]*)(?:\.(\d+))?%$')
    regex_match = regexp_speedp.match(data)
    if regex_match:
        return f'{regex_match.group(1)}%'

    # Validate frequency speeds
    regexp_speedf = re.compile(r'^([1-9]|[1-9][0-9]*)(?:\.(\d+))?(mhz|ghz)$', re.IGNORECASE)
    regex_match = regexp_speedf.match(data)
    if regex_match:
        integer_part, _, unit = regex_match.groups()
        return f'{integer_part}{unit}'

    return None


def validate_int(val, min_val=0, max_val=sys.maxsize):
    """
    Check that val - is a string number
    return val as a string
    """
    try:
        dig_val = int(val)
    except ValueError:
        return None
    if max_val >= dig_val >= min_val:
        return val


def memory_to_page(val, min_val=0, max_val=sys.maxsize):
    try:
        suffix = val[-1]
        if suffix.isdigit():
            suffix = 'K'
            val = val + suffix
        result = int(float(val[:-1]) * BYTES_CONVERSION_TABLE[suffix.upper()] / 4)
        if max_val >= result >= min_val:
            return result
    except (IndexError, ValueError, KeyError):
        pass
    return None


def page_to_memory(pages):
    if pages < 256:  # KB
        return str(pages * 4) + 'K'
    if pages < 262144:  # MB
        return str(round(float(pages) * 4 / 1024, 2)) + 'M'
    return str(round(float(pages) * 4 / (1024 * 1024), 2)) + 'G'


def reload_processes(item, username):
    with subprocess.Popen(
        ['/bin/ps', '-U', username, '-u', username],
        stdout=subprocess.PIPE,
    ) as proc:
        lines = proc.communicate()[0].split(b"\n")
    for row in lines:
        parts = row.rstrip().split()
        try:
            parts[-1].index(item.encode())
            os.kill(int(parts[0]), signal.SIGHUP)
        except (IndexError, ValueError, OSError):
            continue


def login_defs(key, default=None, _path='/etc/login.defs'):
    with open(_path, encoding='utf-8') as login_defs:
        for raw_line in login_defs:
            if raw_line.startswith('#'):
                continue  # ignore commented lines
            line = raw_line.split('#', 1)[0]  # clear line of comments
            line_splited = line.split()
            if len(line_splited) >= 2 and line_splited[0] == key:
                return line_splited[1]
    return default


def uid_max(default=60000):
    try:
        uid_max_ = int(login_defs('UID_MAX'))
    except (IOError, ValueError, TypeError):
        uid_max_ = default
    return uid_max_


def exit_with_error(message, status=1):
    sys.stderr.write(f"{message}\n")
    sys.exit(status)


def safe_escaped_unicode_to_utf(s):
    # transforms unicode-escaped string into utf-8 encoded str
    if '\\u' in s:  # str should have escaped unicode symbols
        try:
            s = s.decode('unicode_escape').encode('utf-8')
        except UnicodeDecodeError:
            pass
    return s


# Copy/paste helper to convert unicode results of json.load back to str
# https://stackoverflow.com/questions/956867/how-to-get-string-objects-instead-of-unicode-from-json/13105359#13105359
def byteify(data):
    if isinstance(data, dict):
        return {byteify(key): byteify(value)
                for key, value in data.items()}
    elif isinstance(data, list):
        return [byteify(element) for element in data]
    elif isinstance(data, tuple):
        return tuple(byteify(element) for element in data)
    elif isinstance(data, str):
        return data.encode()
    else:
        return data


def unicodeify(data):
    if isinstance(data, dict):
        return {unicodeify(key): unicodeify(value)
                for key, value in data.items()}
    elif isinstance(data, list):
        return [unicodeify(element) for element in data]
    elif isinstance(data, tuple):
        return tuple(unicodeify(element) for element in data)
    elif isinstance(data, bytes):
        return data.decode()
    else:
        return data


def is_ascii_string(s):
    """
    Check is string contains only ASCII characters
    :param s: string to check
    :return: True - string contains only ASCII characters
    """
    try:
        s.encode(encoding='utf-8').decode('ascii')
    except UnicodeDecodeError:
        return False
    else:
        return True


def escape_formatting_chars(text: str) -> str:
    """
    Escape '%' characters inside text, except '%' followed by '('
    """
    def replace(match_obj):
        """
        Generate string to replace from matched string
        '% ' -> '%% '
        '%%c' -> '%%%%c'
        """
        return match_obj.group()[:-1] * 2 + match_obj.group()[-1]

    # twice all '%' inside text except '%' followed by '(' and '%' at the end
    # '% hello' -> '%% hello'
    # '%% hello' -> '%%%% hello'
    text = re.sub(r"%+([^(])", replace, text)
    # replace '%' at the end
    text = re.sub(r"%$", r"%%", text)
    return text

Zerion Mini Shell 1.0