oslo.utils/oslo_utils/imageutils.py
Bin Zhou bf80b23e77 Support json format output from qemu command
QemuImgInfo is realized by parsing the output string of 'qemu-img info
**.img' line by line, it's complicated and improvment can be done. The
command 'qemu-img info **.img --output json' will produce json style
output and arguments can be parsed easily by calling json.loads. This
patch add support json format output parsing, to simplify the code and
make it more clear and readable.

Change-Id: Iefa139dc4bcea864cb86371d8a24df0e4695169c
Closes-Bug:#1590252
2016-06-20 01:21:14 +00:00

178 lines
7.3 KiB
Python

# Copyright 2010 United States Government as represented by the
# Administrator of the National Aeronautics and Space Administration.
# All Rights Reserved.
# Copyright (c) 2010 Citrix Systems, Inc.
#
# 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
#
# http://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.
"""
Helper methods to deal with images.
.. versionadded:: 3.1
.. versionchanged:: 3.14.0
add paramter format.
"""
import json
import re
from oslo_utils._i18n import _
from oslo_utils import strutils
class QemuImgInfo(object):
"""Parse Qemu image information from command `qemu-img info`'s output.
The instance of :class:`QemuImgInfo` has properties: `image`,
`backing_file`, `file_format`, `virtual_size`, `cluster_size`,
`disk_size`, `snapshots` and `encrypted`.
The parameter format can be set to 'json' or 'human'. With 'json' format
output, qemu image information will be parsed more easily and readable.
"""
BACKING_FILE_RE = re.compile((r"^(.*?)\s*\(actual\s+path\s*:"
r"\s+(.*?)\)\s*$"), re.I)
TOP_LEVEL_RE = re.compile(r"^([\w\d\s\_\-]+):(.*)$")
SIZE_RE = re.compile(r"(\d*\.?\d+)(\w+)?(\s*\(\s*(\d+)\s+bytes\s*\))?",
re.I)
def __init__(self, cmd_output=None, format='human'):
if format == 'json':
details = json.loads(cmd_output or '{}')
self.image = details.get('filename')
self.backing_file = details.get('backing-filename')
self.file_format = details.get('format')
self.virtual_size = details.get('virtual-size')
self.cluster_size = details.get('cluster-size')
self.disk_size = details.get('actual-size')
self.snapshots = details.get('snapshots', [])
self.encrypted = details.get('encrypted')
else:
details = self._parse(cmd_output or '')
self.image = details.get('image')
self.backing_file = details.get('backing_file')
self.file_format = details.get('file_format')
self.virtual_size = details.get('virtual_size')
self.cluster_size = details.get('cluster_size')
self.disk_size = details.get('disk_size')
self.snapshots = details.get('snapshot_list', [])
self.encrypted = details.get('encrypted')
def __str__(self):
lines = [
'image: %s' % self.image,
'file_format: %s' % self.file_format,
'virtual_size: %s' % self.virtual_size,
'disk_size: %s' % self.disk_size,
'cluster_size: %s' % self.cluster_size,
'backing_file: %s' % self.backing_file,
]
if self.snapshots:
lines.append("snapshots: %s" % self.snapshots)
if self.encrypted:
lines.append("encrypted: %s" % self.encrypted)
return "\n".join(lines)
def _canonicalize(self, field):
# Standardize on underscores/lc/no dash and no spaces
# since qemu seems to have mixed outputs here... and
# this format allows for better integration with python
# - i.e. for usage in kwargs and such...
field = field.lower().strip()
for c in (" ", "-"):
field = field.replace(c, '_')
return field
def _extract_bytes(self, details):
# Replace it with the byte amount
real_size = self.SIZE_RE.search(details)
if not real_size:
raise ValueError(_('Invalid input value "%s".') % details)
magnitude = real_size.group(1)
unit_of_measure = real_size.group(2)
bytes_info = real_size.group(3)
if bytes_info:
return int(real_size.group(4))
elif not unit_of_measure:
return int(magnitude)
return strutils.string_to_bytes('%s%sB' % (magnitude, unit_of_measure),
return_int=True)
def _extract_details(self, root_cmd, root_details, lines_after):
real_details = root_details
if root_cmd == 'backing_file':
# Replace it with the real backing file
backing_match = self.BACKING_FILE_RE.match(root_details)
if backing_match:
real_details = backing_match.group(2).strip()
elif root_cmd in ['virtual_size', 'cluster_size', 'disk_size']:
# Replace it with the byte amount (if we can convert it)
if root_details in ('None', 'unavailable'):
real_details = 0
else:
real_details = self._extract_bytes(root_details)
elif root_cmd == 'file_format':
real_details = real_details.strip().lower()
elif root_cmd == 'snapshot_list':
# Next line should be a header, starting with 'ID'
if not lines_after or not lines_after.pop(0).startswith("ID"):
msg = _("Snapshot list encountered but no header found!")
raise ValueError(msg)
real_details = []
# This is the sprintf pattern we will try to match
# "%-10s%-20s%7s%20s%15s"
# ID TAG VM SIZE DATE VM CLOCK (current header)
while lines_after:
line = lines_after[0]
line_pieces = line.split()
if len(line_pieces) != 6:
break
# Check against this pattern in the final position
# "%02d:%02d:%02d.%03d"
date_pieces = line_pieces[5].split(":")
if len(date_pieces) != 3:
break
lines_after.pop(0)
real_details.append({
'id': line_pieces[0],
'tag': line_pieces[1],
'vm_size': line_pieces[2],
'date': line_pieces[3],
'vm_clock': line_pieces[4] + " " + line_pieces[5],
})
return real_details
def _parse(self, cmd_output):
# Analysis done of qemu-img.c to figure out what is going on here
# Find all points start with some chars and then a ':' then a newline
# and then handle the results of those 'top level' items in a separate
# function.
#
# TODO(harlowja): newer versions might have a json output format
# we should switch to that whenever possible.
# see: http://bit.ly/XLJXDX
contents = {}
lines = [x for x in cmd_output.splitlines() if x.strip()]
while lines:
line = lines.pop(0)
top_level = self.TOP_LEVEL_RE.match(line)
if top_level:
root = self._canonicalize(top_level.group(1))
if not root:
continue
root_details = top_level.group(2).strip()
details = self._extract_details(root, root_details, lines)
contents[root] = details
return contents