Skip to content


Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
Fetching contributors…

Cannot retrieve contributors at this time

731 lines (629 sloc) 29.298 kb
""" Track relationships between compiled extension functions & code fragments
catalog keeps track of which compiled(or even standard) functions are
related to which code fragments. It also stores these relationships
to disk so they are remembered between Python sessions. When
a = 1
compiler.inline('printf("printed from C: %d",a);',['a'] )
is called, inline() first looks to see if it has seen the code
'printf("printed from C");' before. If not, it calls
catalog.get_functions('printf("printed from C: %d", a);')
which returns a list of all the function objects that have been compiled
for the code fragment. Multiple functions can occur because the code
could be compiled for different types for 'a' (although not likely in
this case). The catalog first looks in its cache and quickly returns
a list of the functions if possible. If the cache lookup fails, it then
looks through possibly multiple catalog files on disk and fills its
cache with all the functions that match the code fragment.
In case where the code fragment hasn't been compiled, inline() compiles
the code and then adds it to the catalog:
function = <code to compile function>
catalog.add_function('printf("printed from C: %d", a);',function)
add_function() adds function to the front of the cache. function,
along with the path information to its module, are also stored in a
persistent catalog for future use by python sessions.
import os
import sys
import pickle
import socket
import tempfile
import dbhash
import shelve
dumb = 0
except ImportError:
import as shelve
dumb = 1
#For testing...
#import as shelve
#dumb = 1
#import shelve
#dumb = 0
def getmodule(object):
""" Discover the name of the module where object was defined.
This is an augmented version of inspect.getmodule that can discover
the parent module for extension functions.
import inspect
value = inspect.getmodule(object)
if value is None:
#walk trough all modules looking for function
for name,mod in sys.modules.items():
# try except used because of some comparison failures
# in wxPoint code. Need to review this
if mod and object in mod.__dict__.values():
value = mod
# if it is a built-in module, keep looking to see
# if a non-builtin also has it. Otherwise quit and
# consider the module found. (ain't perfect, but will
# have to do for now).
if str(mod) not in '(built-in)':
except (TypeError, KeyError, ImportError):
return value
def expr_to_filename(expr):
""" Convert an arbitrary expr string to a valid file name.
The name is based on the md5 check sum for the string and
Something that was a little more human readable would be
nice, but the computer doesn't seem to care.
import scipy.weave.md5_load as md5
base = 'sc_'
return base +
def unique_file(d,expr):
""" Generate a unqiue file name based on expr in directory d
This is meant for use with building extension modules, so
a file name is considered unique if none of the following
extension '.cpp','.o','.so','','.py', or '.pyd'
exists in directory d. The fully qualified path to the
new name is returned. You'll need to append your own
extension to it before creating files.
files = os.listdir(d)
#base = 'scipy_compile'
base = expr_to_filename(expr)
for i in xrange(1000000):
fname = base + `i`
if not (fname+'.cpp' in files or
fname+'.o' in files or
fname+'.so' in files or
fname+'' in files or
fname+'.py' in files or
fname+'.pyd' in files):
return os.path.join(d,fname)
def create_dir(p):
""" Create a directory and any necessary intermediate directories."""
if not os.path.exists(p):
except OSError:
# perhaps one or more intermediate path components don't exist
# try to create them
base,dir = os.path.split(p)
# don't enclose this one in try/except - we want the user to
# get failure info
def is_writable(dir):
"""Determine whether a given directory is writable in a portable manner.
dir : str
A string represeting a path to a directory on the filesystem.
res : bool
True or False.
# Do NOT use a hardcoded name here due to the danger from race conditions
# on NFS when multiple processes are accessing the same base directory in
# parallel. We use both hostname and pocess id for the prefix in an
# attempt to ensure that there can really be no name collisions (tempfile
# appends 6 random chars to this prefix).
prefix = 'dummy_%s_%s_' % (socket.gethostname(),os.getpid())
tmp = tempfile.TemporaryFile(prefix=prefix,dir=dir)
except OSError:
return False
# The underlying file is destroyed upon closing the file object (under
# *nix, it was unlinked at creation time)
return True
def whoami():
"""return a string identifying the user."""
return os.environ.get("USER") or os.environ.get("USERNAME") or "unknown"
def default_dir():
""" Return a default location to store compiled files and catalogs.
XX is the Python version number in all paths listed below
On windows, the default location is the temporary directory
returned by gettempdir()/pythonXX.
On Unix, ~/.pythonXX_compiled is the default location. If it doesn't
exist, it is created. The directory is marked rwx------.
If for some reason it isn't possible to build a default directory
in the user's home, /tmp/<uid>_pythonXX_compiled is used. If it
doesn't exist, it is created. The directory is marked rwx------
to try and keep people from being able to sneak a bad module
in on you.
# Use a cached value for fast return if possible
if hasattr(default_dir,"cached_path") and \
os.path.exists(default_dir.cached_path) and \
os.access(default_dir.cached_path, os.W_OK):
return default_dir.cached_path
python_name = "python%d%d_compiled" % tuple(sys.version_info[:2])
path_candidates = []
if sys.platform != 'win32':
'.' + python_name))
except KeyError:
temp_dir = `os.getuid()` + '_' + python_name
path_candidates.append(os.path.join(tempfile.gettempdir(), temp_dir))
"%s" % whoami(), python_name))
writable = False
for path in path_candidates:
if not os.path.exists(path):
os.chmod(path, 0700) # make it only accessible by this user.
if is_writable(path):
writable = True
if not writable:
print 'warning: default directory is not write accessible.'
print 'default:', path
# Cache the default dir path so that this function returns quickly after
# being called once (nothing in it should change after the first call)
default_dir.cached_path = path
return path
def intermediate_dir():
""" Location in temp dir for storing .cpp and .o files during
python_name = "python%d%d_intermediate" % tuple(sys.version_info[:2])
path = os.path.join(tempfile.gettempdir(),"%s"%whoami(),python_name)
if not os.path.exists(path):
return path
def default_temp_dir():
path = os.path.join(default_dir(),'temp')
if not os.path.exists(path):
os.chmod(path,0700) # make it only accessible by this user.
if not is_writable(path):
print 'warning: default directory is not write accessible.'
print 'default:', path
return path
def os_dependent_catalog_name():
""" Generate catalog name dependent on OS and Python version being used.
This allows multiple platforms to have catalog files in the
same directory without stepping on each other. For now, it
bases the name of the value returned by sys.platform and the
version of python being run. If this isn't enough to descriminate
on some platforms, we can try to add other info. It has
occurred to me that if we get fancy enough to optimize for different
architectures, then chip type might be added to the catalog name also.
version = '%d%d' % sys.version_info[:2]
return sys.platform+version+'compiled_catalog'
def catalog_path(module_path):
""" Return the full path name for the catalog file in the given directory.
module_path can either be a file name or a path name. If it is a
file name, the catalog file name in its parent directory is returned.
If it is a directory, the catalog file in that directory is returned.
If module_path doesn't exist, None is returned. Note though, that the
catalog file does *not* have to exist, only its parent. '~', shell
variables, and relative ('.' and '..') paths are all acceptable.
catalog file names are os dependent (based on sys.platform), so this
should support multiple platforms sharing the same disk space
(NFS mounts). See os_dependent_catalog_name() for more info.
module_path = os.path.expanduser(module_path)
module_path = os.path.expandvars(module_path)
module_path = os.path.abspath(module_path)
if not os.path.exists(module_path):
catalog_file = None
elif not os.path.isdir(module_path):
module_path,dummy = os.path.split(module_path)
catalog_file = os.path.join(module_path,os_dependent_catalog_name())
catalog_file = os.path.join(module_path,os_dependent_catalog_name())
return catalog_file
def get_catalog(module_path,mode='r'):
""" Return a function catalog (shelve object) from the path module_path
If module_path is a directory, the function catalog returned is
from that directory. If module_path is an actual module_name,
then the function catalog returned is from its parent directory.
mode uses the standard 'c' = create, 'n' = new, 'r' = read,
'w' = write file open modes available for anydbm databases.
Well... it should be. Stuck with dumbdbm for now and the modes
almost don't matter. We do some checking for 'r' mode, but that
is about it.
See catalog_path() for more information on module_path.
if mode not in ['c','r','w','n']:
msg = " mode must be 'c', 'n', 'r', or 'w'. See anydbm for more info"
raise ValueError(msg)
catalog_file = catalog_path(module_path)
if (catalog_file is not None) \
and ((dumb and os.path.exists(catalog_file+'.dat')) \
or os.path.exists(catalog_file)):
sh =,mode)
if mode=='r':
sh = None
sh =,mode)
return sh
class catalog(object):
""" Stores information about compiled functions both in cache and on disk.
catalog stores (code, list_of_function) pairs so that all the functions
that have been compiled for code are available for calling (usually in
inline or blitz).
catalog keeps a dictionary of previously accessed code values cached
for quick access. It also handles the looking up of functions compiled
in previously called Python sessions on disk in function catalogs.
catalog searches the directories in the PYTHONCOMPILED environment
variable in order loading functions that correspond to the given code
fragment. A default directory is also searched for catalog functions.
On unix, the default directory is usually '~/.pythonxx_compiled' where
xx is the version of Python used. On windows, it is the directory
returned by temfile.gettempdir(). Functions closer to the front are of
the variable list are guaranteed to be closer to the front of the
function list so that they will be called first. See
get_cataloged_functions() for more info on how the search order is
Catalog also handles storing information about compiled functions to
a catalog. When writing this information, the first writable catalog
file in PYTHONCOMPILED path is used. If a writable catalog is not
found, it is written to the catalog in the default directory. This
directory should always be writable.
def __init__(self,user_path_list=None):
""" Create a catalog for storing/searching for compiled functions.
user_path_list contains directories that should be searched
first for function catalogs. They will come before the path
entries in the PYTHONCOMPILED environment varilable.
if isinstance(user_path_list, str):
self.user_path_list = [user_path_list]
elif user_path_list:
self.user_path_list = user_path_list
self.user_path_list = []
self.cache = {}
self.module_dir = None
self.paths_added = 0
# unconditionally append the default dir for auto-generated compiled
# extension modules, so that pickle.load()s don't fail.
def set_module_directory(self,module_dir):
""" Set the path that will replace 'MODULE' in catalog searches.
You should call clear_module_directory() when your finished
working with it.
self.module_dir = module_dir
def get_module_directory(self):
""" Return the path used to replace the 'MODULE' in searches.
return self.module_dir
def clear_module_directory(self):
""" Reset 'MODULE' path to None so that it is ignored in searches.
self.module_dir = None
def get_environ_path(self):
""" Return list of paths from 'PYTHONCOMPILED' environment variable.
On Unix the path in PYTHONCOMPILED is a ':' separated list of
directories. On Windows, a ';' separated list is used.
paths = []
if 'PYTHONCOMPILED' in os.environ:
path_string = os.environ['PYTHONCOMPILED']
paths = path_string.split(os.path.pathsep)
return paths
def build_search_order(self):
""" Returns a list of paths that are searched for catalogs.
Values specified in the catalog constructor are searched first,
then values found in the PYTHONCOMPILED environment variable.
The directory returned by default_dir() is always returned at
the end of the list.
There is a 'magic' path name called 'MODULE' that is replaced
by the directory defined by set_module_directory(). If the
module directory hasn't been set, 'MODULE' is ignored.
paths = self.user_path_list + self.get_environ_path()
search_order = []
for path in paths:
if path == 'MODULE':
if self.module_dir:
return search_order
def get_catalog_files(self):
""" Returns catalog file list in correct search order.
Some of the catalog files may not currently exists.
However, all will be valid locations for a catalog
to be created (if you have write permission).
files = map(catalog_path,self.build_search_order())
files = filter(lambda x: x is not None,files)
return files
def get_existing_files(self):
""" Returns all existing catalog file list in correct search order.
files = self.get_catalog_files()
# open every stinking file to check if it exists.
# This is because anydbm doesn't provide a consistent naming
# convention across platforms for its files
existing_files = []
for file in files:
cat = get_catalog(os.path.dirname(file),'r')
if cat is not None:
# This is the non-portable (and much faster) old code
#existing_files = filter(os.path.exists,files)
return existing_files
def get_writable_file(self,existing_only=0):
""" Return the name of the first writable catalog file.
Its parent directory must also be writable. This is so that
compiled modules can be written to the same directory.
# note: both file and its parent directory must be writeable
if existing_only:
files = self.get_existing_files()
files = self.get_catalog_files()
# filter for (file exists and is writable) OR directory is writable
def file_test(x):
from os import access, F_OK, W_OK
return (access(x,F_OK) and access(x,W_OK) or
writable = filter(file_test,files)
if writable:
file = writable[0]
file = None
return file
def get_writable_dir(self):
""" Return the parent directory of first writable catalog file.
The returned directory has write access.
return os.path.dirname(self.get_writable_file())
def unique_module_name(self,code,module_dir=None):
""" Return full path to unique file name that in writable location.
The directory for the file is the first writable directory in
the catalog search path. The unique file name is derived from
the code fragment. If, module_dir is specified, it is used
to replace 'MODULE' in the search path.
if module_dir is not None:
d = self.get_writable_dir()
if module_dir is not None:
return unique_file(d, code)
def path_key(self,code):
""" Return key for path information for functions associated with code.
return '__path__' + code
def configure_path(self,cat,code):
""" Add the python path for the given code to the sys.path
unconfigure_path() should be called as soon as possible after
imports associated with code are finished so that sys.path
is restored to normal.
paths = cat[self.path_key(code)]
self.paths_added = len(paths)
sys.path = paths + sys.path
self.paths_added = 0
def unconfigure_path(self):
""" Restores sys.path to normal after calls to configure_path()
Remove the previously added paths from sys.path
sys.path = sys.path[self.paths_added:]
self.paths_added = 0
def get_cataloged_functions(self,code):
""" Load all functions associated with code from catalog search path.
Sometimes there can be trouble loading a function listed in a
catalog file because the actual module that holds the function
has been moved or deleted. When this happens, that catalog file
is "repaired", meaning the entire entry for this function is
removed from the file. This only affects the catalog file that
has problems -- not the others in the search path.
The "repair" behavior may not be needed, but I'll keep it for now.
mode = 'r'
cat = None
function_list = []
for path in self.build_search_order():
cat = get_catalog(path,mode)
if cat is not None and code in cat:
# set up the python path so that modules for this
# function can be loaded.
function_list += cat[code]
except: #SystemError and ImportError so far seen
# problems loading a function from the catalog. Try to
# repair the cause.
if cat is not None:
# ensure that the catalog is properly closed
return function_list
def repair_catalog(self,catalog_path,code):
""" Remove entry for code from catalog_path
Occasionally catalog entries could get corrupted. An example
would be when a module that had functions in the catalog was
deleted or moved on the disk. The best current repair method is
just to trash the entire catalog entry for this piece of code.
This may loose function entries that are valid, but thats life.
catalog_path must be writable for repair. If it isn't, the
function exists with a warning.
writable_cat = None
if (catalog_path is not None) and (not os.path.exists(catalog_path)):
writable_cat = get_catalog(catalog_path,'w')
print 'warning: unable to repair catalog entry\n %s\n in\n %s' % \
# shelve doesn't guarantee flushing, so it's safest to explicitly
# close the catalog
if code in writable_cat:
print 'repairing catalog by removing key'
del writable_cat[code]
# it is possible that the path key doesn't exist (if the function
# registered was a built-in function), so we have to check if the path
# exists before arbitrarily deleting it.
path_key = self.path_key(code)
if path_key in writable_cat:
del writable_cat[path_key]
def get_functions_fast(self,code):
""" Return list of functions for code from the cache.
Return an empty list if the code entry is not found.
return self.cache.get(code,[])
def get_functions(self,code,module_dir=None):
""" Return the list of functions associated with this code fragment.
The cache is first searched for the function. If an entry
in the cache is not found, then catalog files on disk are
searched for the entry. This is slooooow, but only happens
once per code object. All the functions found in catalog files
on a cache miss are loaded into the cache to speed up future calls.
The search order is as follows:
1. user specified path (from catalog initialization)
2. directories from the PYTHONCOMPILED environment variable
3. The temporary directory on your platform.
The path specified by module_dir will replace the 'MODULE'
place holder in the catalog search path. See build_search_order()
for more info on the search path.
# Fast!! try cache first.
if code in self.cache:
return self.cache[code]
# 2. Slow!! read previously compiled functions from disk.
function_list = self.get_cataloged_functions(code)
# put function_list in cache to save future lookups.
if function_list:
self.cache[code] = function_list
# return function_list, empty or otherwise.
return function_list
def add_function(self,code,function,module_dir=None):
""" Adds a function to the catalog.
The function is added to the cache as well as the first
writable file catalog found in the search path. If no
code entry exists in the cache, the on disk catalogs
are loaded into the cache and function is added to the
beginning of the function list.
The path specified by module_dir will replace the 'MODULE'
place holder in the catalog search path. See build_search_order()
for more info on the search path.
# 1. put it in the cache.
if code in self.cache:
if function not in self.cache[code]:
# if it is in the cache, then it is also
# been persisted
# Load functions and put this one up front
self.cache[code] = self.get_functions(code)
# 2. Store the function entry to disk.
def add_function_persistent(self,code,function):
""" Store the code->function relationship to disk.
Two pieces of information are needed for loading functions
from disk -- the function pickle (which conveniently stores
the module name, etc.) and the path to its module's directory.
The latter is needed so that the function can be loaded no
matter what the user's Python path is.
# add function to data in first writable catalog
mode = 'c' # create if doesn't exist, otherwise, use existing
cat_dir = self.get_writable_dir()
cat = get_catalog(cat_dir,mode)
if cat is None:
cat_dir = default_dir()
cat = get_catalog(cat_dir,mode)
if cat is None:
cat_dir = default_dir()
cat_file = catalog_path(cat_dir)
print 'problems with default catalog -- removing'
import glob
files = glob.glob(cat_file+'*')
for f in files:
cat = get_catalog(cat_dir,mode)
if cat is None:
raise ValueError('Failed to access a catalog for storing functions')
# Prabhu was getting some corrupt catalog errors. I'll put a try/except
# to protect against this, but should really try and track down the issue.
function_list = [function]
function_list = function_list + cat.get(code,[])
except pickle.UnpicklingError:
cat[code] = function_list
# now add needed path information for loading function
module = getmodule(function)
# built in modules don't have the __file__ extension, so this
# will fail. Just pass in this case since path additions aren't
# needed for built-in modules.
pkey = self.path_key(code)
cat[pkey] = [mod_path] + cat.get(pkey,[])
def fast_cache(self,code,function):
""" Move function to the front of the cache entry for code
If future calls to the function have the same type signature,
this will speed up access significantly because the first
function call is correct.
Note: The cache added to the inline_tools module is significantly
faster than always calling get_functions, so this isn't
as necessary as it used to be. Still, it's probably worth
if self.cache[code][0] == function:
except: # KeyError, IndexError
except ValueError:
# put new function at the beginning of the list to search.
Jump to Line
Something went wrong with that request. Please try again.