1
0
mirror of https://github.com/moparisthebest/SickRage synced 2024-11-06 01:15:05 -05:00
SickRage/cherrypy/lib/covercp.py
echel0n 0d9fbc1ad7 Welcome to our SickBeard-TVRage Edition ...
This version of SickBeard uses both TVDB and TVRage to search and gather it's series data from allowing you to now have access to and download shows that you couldn't before because of being locked into only what TheTVDB had to offer.

Also this edition is based off the code we used in our XEM editon so it does come with scene numbering support as well as all the other features our XEM edition has to offer.

Please before using this with your existing database (sickbeard.db) please make a backup copy of it and delete any other database files such as cache.db and failed.db if present, we HIGHLY recommend starting out with no database files at all to make this a fresh start but the choice is at your own risk!

Enjoy!
2014-03-09 22:39:12 -07:00

365 lines
11 KiB
Python

"""Code-coverage tools for CherryPy.
To use this module, or the coverage tools in the test suite,
you need to download 'coverage.py', either Gareth Rees' original
implementation:
http://www.garethrees.org/2001/12/04/python-coverage/
or Ned Batchelder's enhanced version:
http://www.nedbatchelder.com/code/modules/coverage.html
To turn on coverage tracing, use the following code:
cherrypy.engine.subscribe('start', covercp.start)
DO NOT subscribe anything on the 'start_thread' channel, as previously
recommended. Calling start once in the main thread should be sufficient
to start coverage on all threads. Calling start again in each thread
effectively clears any coverage data gathered up to that point.
Run your code, then use the covercp.serve() function to browse the
results in a web browser. If you run this module from the command line,
it will call serve() for you.
"""
import re
import sys
import cgi
from urllib import quote_plus
import os, os.path
localFile = os.path.join(os.path.dirname(__file__), "coverage.cache")
try:
from coverage import the_coverage as coverage
def start():
coverage.start()
except ImportError:
# Setting coverage to None will raise errors
# that need to be trapped downstream.
coverage = None
import warnings
warnings.warn("No code coverage will be performed; coverage.py could not be imported.")
def start():
pass
start.priority = 20
TEMPLATE_MENU = """<html>
<head>
<title>CherryPy Coverage Menu</title>
<style>
body {font: 9pt Arial, serif;}
#tree {
font-size: 8pt;
font-family: Andale Mono, monospace;
white-space: pre;
}
#tree a:active, a:focus {
background-color: black;
padding: 1px;
color: white;
border: 0px solid #9999FF;
-moz-outline-style: none;
}
.fail { color: red;}
.pass { color: #888;}
#pct { text-align: right;}
h3 {
font-size: small;
font-weight: bold;
font-style: italic;
margin-top: 5px;
}
input { border: 1px solid #ccc; padding: 2px; }
.directory {
color: #933;
font-style: italic;
font-weight: bold;
font-size: 10pt;
}
.file {
color: #400;
}
a { text-decoration: none; }
#crumbs {
color: white;
font-size: 8pt;
font-family: Andale Mono, monospace;
width: 100%;
background-color: black;
}
#crumbs a {
color: #f88;
}
#options {
line-height: 2.3em;
border: 1px solid black;
background-color: #eee;
padding: 4px;
}
#exclude {
width: 100%;
margin-bottom: 3px;
border: 1px solid #999;
}
#submit {
background-color: black;
color: white;
border: 0;
margin-bottom: -9px;
}
</style>
</head>
<body>
<h2>CherryPy Coverage</h2>"""
TEMPLATE_FORM = """
<div id="options">
<form action='menu' method=GET>
<input type='hidden' name='base' value='%(base)s' />
Show percentages <input type='checkbox' %(showpct)s name='showpct' value='checked' /><br />
Hide files over <input type='text' id='pct' name='pct' value='%(pct)s' size='3' />%%<br />
Exclude files matching<br />
<input type='text' id='exclude' name='exclude' value='%(exclude)s' size='20' />
<br />
<input type='submit' value='Change view' id="submit"/>
</form>
</div>"""
TEMPLATE_FRAMESET = """<html>
<head><title>CherryPy coverage data</title></head>
<frameset cols='250, 1*'>
<frame src='menu?base=%s' />
<frame name='main' src='' />
</frameset>
</html>
"""
TEMPLATE_COVERAGE = """<html>
<head>
<title>Coverage for %(name)s</title>
<style>
h2 { margin-bottom: .25em; }
p { margin: .25em; }
.covered { color: #000; background-color: #fff; }
.notcovered { color: #fee; background-color: #500; }
.excluded { color: #00f; background-color: #fff; }
table .covered, table .notcovered, table .excluded
{ font-family: Andale Mono, monospace;
font-size: 10pt; white-space: pre; }
.lineno { background-color: #eee;}
.notcovered .lineno { background-color: #000;}
table { border-collapse: collapse;
</style>
</head>
<body>
<h2>%(name)s</h2>
<p>%(fullpath)s</p>
<p>Coverage: %(pc)s%%</p>"""
TEMPLATE_LOC_COVERED = """<tr class="covered">
<td class="lineno">%s&nbsp;</td>
<td>%s</td>
</tr>\n"""
TEMPLATE_LOC_NOT_COVERED = """<tr class="notcovered">
<td class="lineno">%s&nbsp;</td>
<td>%s</td>
</tr>\n"""
TEMPLATE_LOC_EXCLUDED = """<tr class="excluded">
<td class="lineno">%s&nbsp;</td>
<td>%s</td>
</tr>\n"""
TEMPLATE_ITEM = "%s%s<a class='file' href='report?name=%s' target='main'>%s</a>\n"
def _percent(statements, missing):
s = len(statements)
e = s - len(missing)
if s > 0:
return int(round(100.0 * e / s))
return 0
def _show_branch(root, base, path, pct=0, showpct=False, exclude=""):
# Show the directory name and any of our children
dirs = [k for k, v in root.items() if v]
dirs.sort()
for name in dirs:
newpath = os.path.join(path, name)
if newpath.lower().startswith(base):
relpath = newpath[len(base):]
yield "| " * relpath.count(os.sep)
yield "<a class='directory' href='menu?base=%s&exclude=%s'>%s</a>\n" % \
(newpath, quote_plus(exclude), name)
for chunk in _show_branch(root[name], base, newpath, pct, showpct, exclude):
yield chunk
# Now list the files
if path.lower().startswith(base):
relpath = path[len(base):]
files = [k for k, v in root.items() if not v]
files.sort()
for name in files:
newpath = os.path.join(path, name)
pc_str = ""
if showpct:
try:
_, statements, _, missing, _ = coverage.analysis2(newpath)
except:
# Yes, we really want to pass on all errors.
pass
else:
pc = _percent(statements, missing)
pc_str = ("%3d%% " % pc).replace(' ', '&nbsp;')
if pc < float(pct) or pc == -1:
pc_str = "<span class='fail'>%s</span>" % pc_str
else:
pc_str = "<span class='pass'>%s</span>" % pc_str
yield TEMPLATE_ITEM % ("| " * (relpath.count(os.sep) + 1),
pc_str, newpath, name)
def _skip_file(path, exclude):
if exclude:
return bool(re.search(exclude, path))
def _graft(path, tree):
d = tree
p = path
atoms = []
while True:
p, tail = os.path.split(p)
if not tail:
break
atoms.append(tail)
atoms.append(p)
if p != "/":
atoms.append("/")
atoms.reverse()
for node in atoms:
if node:
d = d.setdefault(node, {})
def get_tree(base, exclude):
"""Return covered module names as a nested dict."""
tree = {}
coverage.get_ready()
runs = list(coverage.cexecuted.keys())
if runs:
for path in runs:
if not _skip_file(path, exclude) and not os.path.isdir(path):
_graft(path, tree)
return tree
class CoverStats(object):
def __init__(self, root=None):
if root is None:
# Guess initial depth. Files outside this path will not be
# reachable from the web interface.
import cherrypy
root = os.path.dirname(cherrypy.__file__)
self.root = root
def index(self):
return TEMPLATE_FRAMESET % self.root.lower()
index.exposed = True
def menu(self, base="/", pct="50", showpct="",
exclude=r'python\d\.\d|test|tut\d|tutorial'):
# The coverage module uses all-lower-case names.
base = base.lower().rstrip(os.sep)
yield TEMPLATE_MENU
yield TEMPLATE_FORM % locals()
# Start by showing links for parent paths
yield "<div id='crumbs'>"
path = ""
atoms = base.split(os.sep)
atoms.pop()
for atom in atoms:
path += atom + os.sep
yield ("<a href='menu?base=%s&exclude=%s'>%s</a> %s"
% (path, quote_plus(exclude), atom, os.sep))
yield "</div>"
yield "<div id='tree'>"
# Then display the tree
tree = get_tree(base, exclude)
if not tree:
yield "<p>No modules covered.</p>"
else:
for chunk in _show_branch(tree, base, "/", pct,
showpct == 'checked', exclude):
yield chunk
yield "</div>"
yield "</body></html>"
menu.exposed = True
def annotated_file(self, filename, statements, excluded, missing):
source = open(filename, 'r')
buffer = []
for lineno, line in enumerate(source.readlines()):
lineno += 1
line = line.strip("\n\r")
empty_the_buffer = True
if lineno in excluded:
template = TEMPLATE_LOC_EXCLUDED
elif lineno in missing:
template = TEMPLATE_LOC_NOT_COVERED
elif lineno in statements:
template = TEMPLATE_LOC_COVERED
else:
empty_the_buffer = False
buffer.append((lineno, line))
if empty_the_buffer:
for lno, pastline in buffer:
yield template % (lno, cgi.escape(pastline))
buffer = []
yield template % (lineno, cgi.escape(line))
def report(self, name):
coverage.get_ready()
filename, statements, excluded, missing, _ = coverage.analysis2(name)
pc = _percent(statements, missing)
yield TEMPLATE_COVERAGE % dict(name=os.path.basename(name),
fullpath=name,
pc=pc)
yield '<table>\n'
for line in self.annotated_file(filename, statements, excluded,
missing):
yield line
yield '</table>'
yield '</body>'
yield '</html>'
report.exposed = True
def serve(path=localFile, port=8080, root=None):
if coverage is None:
raise ImportError("The coverage module could not be imported.")
coverage.cache_default = path
import cherrypy
cherrypy.config.update({'server.socket_port': int(port),
'server.thread_pool': 10,
'environment': "production",
})
cherrypy.quickstart(CoverStats(root))
if __name__ == "__main__":
serve(*tuple(sys.argv[1:]))