python: Handle errors when loading chained modules more properly.
[ashd.git] / python3 / ashd / wsgidir.py
CommitLineData
173e0e9e
FT
1"""WSGI handler for serving chained WSGI modules from physical files
2
7ed9b82b
FT
3The WSGI handler in this module ensures that the SCRIPT_FILENAME
4variable is properly set in every request and points out a file that
5exists and is readable. It then dispatches the request in one of two
6ways: If the header X-Ash-Python-Handler is set in the request, its
7value is used as the name of a handler object to dispatch the request
8to; otherwise, the file extension of the SCRIPT_FILENAME is used to
9determine the handler object.
10
11The name of a handler object is specified as a string, which is split
12along its last constituent dot. The part left of the dot is the name
13of a module, which is imported; and the part right of the dot is the
14name of an object in that module, which should be a callable adhering
15to the WSGI specification. Alternatively, the module part may be
16omitted (such that the name is a string with no dots), in which case
17the handler object is looked up from this module.
18
19By default, this module will handle files with the extensions `.wsgi'
2d4ab435 20or `.wsgi3' using the `chain' handler, which chainloads such files and
7ed9b82b
FT
21runs them as independent WSGI applications. See its documentation for
22details.
173e0e9e
FT
23
24This module itself contains both an `application' and a `wmain'
25object. If this module is used by ashd-wsgi(1) or scgi-wsgi(1) so that
26its wmain function is called, arguments can be specified to it to
27install handlers for other file extensions. Such arguments take the
7ed9b82b
FT
28form `.EXT=HANDLER', where EXT is the file extension to be handled,
29and HANDLER is a handler name, as described above. For example, the
30argument `.fpy=my.module.foohandler' can be given to pass requests for
31`.fpy' files to the function `foohandler' in the module `my.module'
32(which must, of course, be importable). When writing such handler
33functions, you may want to use the getmod() function in this module.
173e0e9e
FT
34"""
35
58ee5c4a 36import sys, os, threading, types, logging, importlib, getopt
173e0e9e
FT
37from . import wsgiutil
38
5c1a2105 39__all__ = ["application", "wmain", "getmod", "cachedmod", "chain"]
173e0e9e 40
58ee5c4a
FT
41log = logging.getLogger("wsgidir")
42
173e0e9e
FT
43class cachedmod(object):
44 """Cache entry for modules loaded by getmod()
45
46 Instances of this class are returned by the getmod()
47 function. They contain three data attributes:
48 * mod - The loaded module
49 * lock - A threading.Lock object, which can be used for
50 manipulating this instance in a thread-safe manner
51 * mtime - The time the file was last modified
52
53 Additional data attributes can be arbitrarily added for recording
54 any meta-data about the module.
55 """
7fe08a6f 56 def __init__(self, mod = None, mtime = -1):
173e0e9e
FT
57 self.lock = threading.Lock()
58 self.mod = mod
59 self.mtime = mtime
60
173e0e9e
FT
61modcache = {}
62cachelock = threading.Lock()
63
64def mangle(path):
65 ret = ""
66 for c in path:
67 if c.isalnum():
68 ret += c
69 else:
70 ret += "_"
71 return ret
72
73def getmod(path):
74 """Load the given file as a module, caching it appropriately
75
76 The given file is loaded and compiled into a Python module. The
77 compiled module is cached and returned upon subsequent requests
78 for the same file, unless the file has changed (as determined by
79 its mtime), in which case the cached module is discarded and the
80 new file contents are reloaded in its place.
81
82 The return value is an instance of the cachedmod class, which can
83 be used for locking purposes and for storing arbitrary meta-data
84 about the module. See its documentation for details.
85 """
86 sb = os.stat(path)
2037cee2 87 with cachelock:
173e0e9e
FT
88 if path in modcache:
89 entry = modcache[path]
7fe08a6f 90 else:
b8d56e8f 91 entry = [threading.Lock(), None]
7fe08a6f 92 modcache[path] = entry
b8d56e8f
FT
93 with entry[0]:
94 if entry[1] is None or sb.st_mtime > entry[1].mtime:
7fe08a6f
FT
95 with open(path, "rb") as f:
96 text = f.read()
97 code = compile(text, path, "exec")
98 mod = types.ModuleType(mangle(path))
99 mod.__file__ = path
100 exec(code, mod.__dict__)
b8d56e8f
FT
101 entry[1] = cachedmod(mod, sb.st_mtime)
102 return entry[1]
173e0e9e 103
e9817fee
FT
104class handler(object):
105 def __init__(self):
106 self.lock = threading.Lock()
107 self.handlers = {}
108 self.exts = {}
109 self.addext("wsgi", "chain")
110 self.addext("wsgi3", "chain")
111
112 def resolve(self, name):
113 with self.lock:
114 if name in self.handlers:
115 return self.handlers[name]
116 p = name.rfind('.')
117 if p < 0:
118 return globals()[name]
119 mname = name[:p]
120 hname = name[p + 1:]
121 mod = importlib.import_module(mname)
122 ret = getattr(mod, hname)
123 self.handlers[name] = ret
124 return ret
125
126 def addext(self, ext, handler):
127 self.exts[ext] = self.resolve(handler)
128
129 def handle(self, env, startreq):
130 if not "SCRIPT_FILENAME" in env:
131 return wsgiutil.simpleerror(env, startreq, 500, "Internal Error", "The server is erroneously configured.")
132 path = env["SCRIPT_FILENAME"]
7ed9b82b 133 if not os.access(path, os.R_OK):
e9817fee 134 return wsgiutil.simpleerror(env, startreq, 500, "Internal Error", "The server is erroneously configured.")
7ed9b82b
FT
135 if "HTTP_X_ASH_PYTHON_HANDLER" in env:
136 handler = self.resolve(env["HTTP_X_ASH_PYTHON_HANDLER"])
137 else:
138 base = os.path.basename(path)
139 p = base.rfind('.')
140 if p < 0:
141 return wsgiutil.simpleerror(env, startreq, 500, "Internal Error", "The server is erroneously configured.")
142 ext = base[p + 1:]
143 if not ext in self.exts:
144 return wsgiutil.simpleerror(env, startreq, 500, "Internal Error", "The server is erroneously configured.")
145 handler = self.exts[ext]
146 return handler(env, startreq)
e9817fee
FT
147
148def wmain(*argv):
149 """Main function for ashd(7)-compatible WSGI handlers
150
151 Returns the `application' function. If any arguments are given,
152 they are parsed according to the module documentation.
153 """
6085469b
FT
154 hnd = handler()
155 ret = hnd.handle
156
157 opts, args = getopt.getopt(argv, "-V")
158 for o, a in opts:
159 if o == "-V":
160 import wsgiref.validate
161 ret = wsgiref.validate.validator(ret)
162
163 for arg in args:
e9817fee
FT
164 if arg[0] == '.':
165 p = arg.index('=')
6085469b
FT
166 hnd.addext(arg[1:p], arg[p + 1:])
167 return ret
e9817fee 168
173e0e9e 169def chain(env, startreq):
7ed9b82b
FT
170 """Chain-loading WSGI handler
171
172 This handler loads requested files, compiles them and loads them
173 into their own modules. The compiled modules are cached and reused
174 until the file is modified, in which case the previous module is
175 discarded and the new file contents are loaded into a new module
176 in its place. When chaining such modules, an object named `wmain'
177 is first looked for and called with no arguments if found. The
178 object it returns is then used as the WSGI application object for
179 that module, which is reused until the module is reloaded. If
180 `wmain' is not found, an object named `application' is looked for
181 instead. If found, it is used directly as the WSGI application
182 object.
183 """
173e0e9e 184 path = env["SCRIPT_FILENAME"]
58ee5c4a
FT
185 try:
186 mod = getmod(path)
187 except Exception:
188 log.error("Exception occurred when loading %s" % path, exc_info=sys.exc_info())
189 return wsgiutil.simpleerror(env, startreq, 500, "Internal Error", "Could not load WSGI handler.")
173e0e9e
FT
190 entry = None
191 if mod is not None:
2037cee2 192 with mod.lock:
173e0e9e
FT
193 if hasattr(mod, "entry"):
194 entry = mod.entry
195 else:
196 if hasattr(mod.mod, "wmain"):
197 entry = mod.mod.wmain()
198 elif hasattr(mod.mod, "application"):
199 entry = mod.mod.application
200 mod.entry = entry
173e0e9e
FT
201 if entry is not None:
202 return entry(env, startreq)
203 return wsgiutil.simpleerror(env, startreq, 500, "Internal Error", "Invalid WSGI handler.")
173e0e9e 204
e9817fee 205application = handler().handle