""" generate_pi.py -- Generate Python interface by inspecting a module at runtime Copyright (c) 2001-2008, Archaeopteryx Software, Inc. All rights reserved. Written by John P. Ehresman and Stephan R.A. Deibel Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. Simple utility to generate a python interface file from inspecting a module at run time. First argument is the name of the module. Subsequent arguments are name:expression pairs -- a class with given name will be created in the interface file with methods & attributes matching the methods & attributes of the object that results from the expression. The expression will be evaluated within the context of the module. The interface will be written to stdout. This contains some code specific to Python standard library code because it parses the docstring standards used there to determine information about return values. However, it works also with code that does not contain those type hints in docstrings (but return value type cannot be determined). """ # IMPORTANT: This code has to run under all Python versions! import sys import os import string import stat try: import inspect except: inspect = None kIronPython = (sys.version_info >= (1, 6) and sys.version.find('IronPython') != -1) try: ascii_letters = string.ascii_letters except: ascii_letters = string.letters version = ((sys.hexversion & 0xff000000) >> 24, (sys.hexversion & 0x00ff0000) >> 16) def string_split(s, sep=' '): return s.split(sep) def string_join(s, sep): return sep.join(s) def string_find(s, t): return s.find(t) def string_strip(s): return s.strip() def string_replace(s, f, t, count=-1): return s.replace(f, t, count) def string_lower(s): return s.lower() def string_rfind(s, x): return s.rfind(x) def has_key(o, key): if version >= (3, 0): return key in o else: return o.has_key(key) if version >= (3, 0): def callable(o): return hasattr(o, '__call__') if version[0] == 1: printable_chars = '0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ!"#$%&\'()*+,-./:;<=>?@[\\]^_`{|}~ \t\n\r\x0b\x0c' else: printable_chars = string.printable kOneLevelIndent = ' ' kNoValue = [] kStringTypes = [type('')] import types try: kStringTypes = types.StringTypes except: pass kLiteralTypes = [type(1), type(1.0), type(None)] try: kLiteralTypes.append(types.EllipsisType) kLiteralTypes.append(types.BooleanType) except: pass kStructureTypes = [type([]), type(()), type({})] kFileTypes = [type(sys.stdout)] if version >= (3, 0): kListType = type([]) kTupleType = type(()) kDictType = type({}) else: kListType = types.ListType kTupleType = types.TupleType kDictType = types.DictType # Property support contributed by Nigel Rowe, Aug 10, 2007 kPropertyTypes = [] try: kPropertyTypes.append(type(property())) except: pass import types try: # types.GetSetDescriptorType only exists in python versions >= 2.5 # and would otherwise need to be extracted from an extension module kPropertyTypes.append(types.GetSetDescriptorType) except: pass if kIronPython: import System kPropertyTypes.append(type(System.Version.Major)) del System # Types of type, old-style class -- use C api names to be a bit clearer PyType_Type = type(type('')) class _OldStyleClass: pass PyClass_Type = type(_OldStyleClass) del _OldStyleClass kClassLikeTypes = [PyType_Type, PyClass_Type] def WriteDocString(obj, file, indent): """ Writes doc string for object if there is one. """ quote = '"""' doc = getattr(obj, '__doc__', None) if doc == None: return doc = doc.replace('\r\n', '\n') # CHANGED: unicode docstrings seem to be broken # wouldn't the u prefix be incorrect for Py 3 anyway? if kIronPython or sys.version_info >= (3, 0): is_unicode = 0 else: try: is_unicode = isinstance(doc, unicode) except: is_unicode = 0 doc_parts = string_split(doc, '\n') for i in range(0, len(doc_parts)): line_repr = repr(doc_parts[i]) if string_lower(line_repr[:1]) == 'u': line_repr = line_repr[1:] line_repr = line_repr[1:-1] line_parts = string_split(line_repr, quote) line_repr = string_join(line_parts, '\\' + quote) if i != 0: line_repr = indent + line_repr doc_parts[i] = line_repr doc = string_join(doc_parts, '\n') if is_unicode: file.write(indent + 'u%s %s %s\n' % (quote, doc, quote)) else: file.write(indent + '%s %s %s\n' % (quote, doc, quote)) def NextNonBlank(doc, i): for j in range(i, len(doc)): if doc[j] not in ' \t\r\n': return j return -1 def FindFirstLine(doc): paren_count = 0 bracket_count = 0 brace_count = 0 first_line = '' seen_chars = 0 i = 0 while i < len(doc): c = doc[i] if c == '(': paren_count = paren_count + 1 elif c == '[': bracket_count = bracket_count + 1 next_pos = NextNonBlank(doc, i+1) if doc[next_pos] == ',': first_line = first_line + ', [' i = next_pos + 1 continue elif c == '{': brace_count = brace_count + 1 elif c == ')': paren_count = paren_count - 1 elif c == ']': bracket_count = bracket_count - 1 next_pos = NextNonBlank(doc, i+1) if doc[next_pos] == '[': first_line = first_line + '], [' i = next_pos + 1 bracket_count = bracket_count + 1 continue elif c == '}': brace_count = brace_count - 1 elif c in (' \t\r\n') and not seen_chars: i = i + 1 continue elif c == '\n': if paren_count + bracket_count + brace_count == 0: return string_join(string_split(first_line), ' ') else: i = i + 1 continue seen_chars = 1 first_line = first_line + c i = i + 1 return string_join(string_split(first_line), ' ') def ValidArgName(n): if len(n) == 0: return 0 if n[0] not in '_' + ascii_letters: return 0 for c in n[1:]: if c not in ascii_letters + string.digits + '_': return 0 return 1 def GetCallableSignatureIPy(val, obj, name): import_line = None return_line = 'pass' doc = getattr(val, '__doc__', None) if doc is None: return '', None, 'pass' import clr try: clr_type = clr.GetClrType(obj) except TypeError: # XXXXX happens when we get namespace tracker objects # Annoying ???? print >> sys.stderr, '@@@@@ Class %s is not a type. (Method = %s)' % (obj, name) a, b = GetCallableSignature(val) return a, None, b methods = [m for m in clr_type.GetMethods() if m.Name == name] # Very occasionally (e.g. Microsoft.CSharp.CSharpCodeProvider.CreateProvider) # We fail to find the method. Also happens for some IronPython methods # Like ReferenceEquals and ToObject (inherited methods?) if not methods: print >> sys.stderr, 'Unlikely as it may sound, class %s has no method %s' % (obj, name) a, b = GetCallableSignature(val) return a, None, b # XXXX if len(methods) > 1 we have multiple overloads # what to do? Just use the first. method = methods[0] return_type = method.ReturnType if return_type is not None: ns = return_type.Namespace import_line = 'import %s' % ns return_line = 'return %s.%s()' % (ns, return_type.Name) parameter_info = method.GetParameters() params = [p.Name for p in parameter_info] if not method.IsStatic: params = ['self'] + params return string_join(params, ', '), import_line, return_line def GetCallableSignature(obj): doc = getattr(obj, '__doc__', None) if doc is None: return '', 'pass' if kIronPython: return _IronPythonGetCallableSignature(doc) else: return _GetCallableSignature(doc) def _GetCallableSignature(doc): try: return __GetCallableSignature(doc) except: import traceback sys.stderr.write("Error parsing call signature from docstring:\n") sys.stderr.write(doc) etype, evalue, etb = sys.exc_info() sys.stderr.write(string_join(traceback.format_exception(etype, evalue, etb), '\n')) return '', 'pass' # CHANGED: exclude the indexing of return types import re ret_val_re = re.compile(r'^(?: )?([A-Za-z][A-Za-z0-9_\-]*)(?: |\[.*\] ).*\(.*\)') # Yuck!!! def _IronPythonGetCallableSignature(doc): a, b = _GetCallableSignature(doc) if b != 'pass': return a, b mat = ret_val_re.match(doc) if mat is None: return a, 'pass' ret_val = 'return ' + mat.groups()[0] + '()' print >>sys.stderr, '#### Generating return value of', ret_val return a, ret_val def __GetCallableSignature(doc): doc = string_replace(doc, '\r\n', '\n') doc = string_replace(doc, '\r', '\n') # Find the argument and return value spec in the docstring first_line = FindFirstLine(doc) #sys.stdout.write("FIRSTLINE %s" % str(first_line)) args = None # Sometimes in form "spec -- comment" but also sometimes use --> or <==> instead of -> first_line = string_replace(first_line, '-->', '->') first_line = string_replace(first_line, '<==>', '->') if string_find(first_line, '--') >= 0: if string_find(first_line, '->') > 0: first_line = first_line[:string_find(first_line, '--')] else: first_line = string_replace(first_line, '--', '->', 1) if string_find(first_line, '->') >= 0: parts = string_split(first_line, '->') if len(parts) == 2: args, ret = parts elif string_find(first_line, '=') > 0: _parts = string_split(first_line, '=') parts = [] for part in _parts: parts.append(string_strip(part)) if len(parts) > 0 and string_find(parts[0], ' ') == -1: ret = parts[0] args = string_join(parts[1:], '=') lparen = string_find(args, '(') rparen = string_find(args, ')') if len(args) > 2 and lparen >= 0 and rparen > lparen: args = args[lparen+1:rparen] elif string_find(first_line, '(') >= 0: args = first_line ret = '' if args is None: lines = string_split(doc, '\n') if len(lines) == 1: return '', 'pass' else: return _GetCallableSignature(string_join(lines[1:], '\n')) # Extract arg spec from parenthesis lparen = string_find(args, '(') rparen = string_rfind(args, ')') if lparen >= 0 and rparen > lparen: args = args[lparen+1:rparen] # Parse each argument arglist = string_split(args, ',') parsed_args = [] arg_count = 0 seen_defaults = 0 for i in range(0, len(arglist)): arg = string_strip(arglist[i]) if string_find(arg, '=') > 0: seen_defaults = 1 # Remove and count []'s indicating optional args fixed_arg = '' for c in arg: if c == '[': seen_defaults = 1 elif c == ']': pass else: fixed_arg = fixed_arg + c arg = string_strip(fixed_arg) arg = string_replace(arg, '-', '_') # Remove surrounding quote seen in some cases for no apparent reason if len(arg) > 2 and arg[0] in ('"', "'") and arg[-1] == arg[0]: arg = arg[1:-1] # Sometimes there's an extra comma; treat as anonymous arg if len(arg) == 0: arg = 'arg' # An optional argument of some type with default value or * or ** if seen_defaults: # Rest of args if arg == '...': arg = '*args' # Verbal description: Create appropriate generic arg elif string_find(arg, ' ') >= 0: if string_find(arg, 'optional') == 0 and i == len(arglist) - 1: if seen_defaults: arg = '**argv' else: arg = '*argv' else: arg = 'arg%i=None' % arg_count seen_defaults = 1 arg_count = arg_count + 1 # Actual arg name -- add default if appropriate elif string_find(arg, '=') == -1 and arg[0] != '*': arg = arg + '=None' seen_defaults = 1 elif arg == '...': seen_defaults = 1 arg = '*args' # Use generic arg name if specified but not parseable elif len(arg) > 0 and not ValidArgName(arg): arg = 'arg%i' % arg_count arg_count = arg_count + 1 if len(arg) > 0: parsed_args.append(arg) # Build return specification retval = BuildReturnSpec(ret) if retval == '': retval = 'pass' else: retval = 'return ' + retval return string_join(parsed_args, ', '), retval def BuildReturnSpec(ret): ret = string_strip(ret) if len(ret) > 2 and ((ret[0] == '(' and ret[-1] == ')') or (ret[0] == '[' and ret[-1] == ']') or (ret[0] == '{' and ret[-1] == '}')): if string_find(ret, '...') >= 0: retval = ret[0] + ret[-1] else: ret_items = string_split(ret[1:-1], ',') fixed_items = [] for item in ret_items: item = string_replace(string_strip(item), ' ', '_') item = string_replace(item, '-', '_') fixed_items.append(BuildReturnSpec(item)) ret = ret[0] + string_join(fixed_items, ', ') + ret[-1] retval = ret elif string_find(ret, 'tuple') >= 0: retval = '()' elif string_find(ret, 'list') >= 0: retval = '[]' elif string_find(ret, 'dict') >= 0: retval = '{}' elif string_find(ret, 'str') >= 0: retval = '""' elif string_find(ret, 'file') >= 0: if version[0] == 1 or (version[0] == 2 and version[1] < 2): retval = '__FileObject()' else: retval = 'file(__file__)' elif string_find(ret, 'socket') >= 0: retval = 'SocketType()' elif string_find(ret, 'int') >= 0: retval = '1' elif string_find(ret, 'float') >= 0: retval = '1.0' elif len(ret) > 0: retval = 'None' else: retval = '' return retval kOmitTopLevel = ('__doc__', '__class__', '__new__', '__file__', '__name__', '__module__', '__builtins__') def ValidName(scope, name): if len(name) < 5: return 1 if len(scope) == 1 and name in kOmitTopLevel: return 0 if name[:2] == '__' and name[-2:] == '__': name = name[2:-2] if len(scope) > 1: return name == 'init' else: return 1 else: return 1 def GetValue(obj, slot, default): if type(obj) == type({}): return obj.get(slot, default) else: return getattr(obj, slot, default) def HasValue(obj, slot): if type(obj) == type({}): return has_key(obj, slot) else: return hasattr(obj, slot) def ValueItems(obj): if type(obj) == type({}): return obj.keys() elif type(obj) in kClassLikeTypes or type(obj) == types.ModuleType: return dir(obj) else: return () def IsClassLike(val): for base in kClassLikeTypes: if isinstance(val, base): return 1 return 0 def IsFunctionOrMethod(val): return not IsClassLike(val) and callable(val) def SafeDir(o): try: return dir(o) except: return [] def GenForObject(scope, overrides, overrides_file, obj, file, seen_vals, indent = '', use_inspect=0): # Use __all__ attrib if obj is a module name_list = None if isinstance(obj, type(sys)): name_list = getattr(obj, '__all__', None) elif type(obj) == type({}): name_list = obj.keys() print >> sys.stderr, '**** generating', repr(obj) if name_list == None: name_list = SafeDir(obj) last_val = kNoValue num_written = 0 for name in name_list: if type(obj) == type({}): val = obj[name] else: # XXX Probably use kIronPython branch for CPython as well if kIronPython: val = obj.__dict__.get(name, None) if val is None: val = getattr(obj, name, None) else: val = getattr(obj, name, None) if kIronPython and name == 'None': name = 'None_' # skip nested namespaces if kIronPython: import System if isinstance(val, type(System)): continue if ValidName(scope, name) and (not hasattr(val, '__objclass__') or val.__objclass__ == obj): # Value is a class object if IsClassLike(val): # Skip overridden values if the override is not also a class object if HasValue(overrides, name): oval = GetValue(overrides, name, kNoValue) if oval is not kNoValue and type(oval) not in kClassLikeTypes: continue val_name = getattr(val, '__name__', None) if val_name != name and val_name in name_list: if last_val is not kNoValue and not callable(last_val): file.write('\n') file.write(indent + '%s = %s\n' % (name, val_name)) elif not has_key(seen_vals, val): try: seen_vals[val] = 1 except TypeError: sys.stderr.write("Failed to register %s (may cause duplicates)\n" % str(val)) if last_val is not kNoValue and not callable(last_val): file.write('\n') GenPClassForType(scope + [name], GetValue(overrides, name, {}), overrides_file, name, val, file, seen_vals, indent, use_inspect) num_written = num_written + 1 # Try to use inspect module if requested, to place source directly (used # to write overrides) elif use_inspect and inspect is not None and IsFunctionOrMethod(val): try: src = inspect.getsource(val) except: #sys.stderr.write("NO SOURCE FOR NAME=%s TYPE=%s VAL=%s\n" % (name, str(type(val)), str(val))) pass else: if src: #sys.stderr.write("FOUND SOURCE FOR NAME=%s TYPE=%s VAL=%s\n" % (name, str(type(val)), str(val))) file.write(src + '\n') num_written = num_written + 1 last_val = val continue #else: #sys.stderr.write("EMPTY SOURCE FOR NAME=%s TYPE=%s VAL=%s\n" % (name, str(type(val)), str(val))) # Skip functions, methods, or constants modified in overrides elif HasValue(overrides, name): continue # Value is a data-descriptor # Contributed by Nigel Rowe, Aug 10, 2007 elif type(val) in kPropertyTypes: indent_to_paren = indent + ' ' * len(name) + ' ' * 12 file.write(indent + '%s = property(None, None, None,\n' % (name,)) WriteDocString(val, file, indent_to_paren) file.write(indent_to_paren + ')\n\n') # Value is a function or method elif callable(val): try: seen_vals[val] = 1 except TypeError: sys.stderr.write("Failed to register %s (may cause duplicates)\n" % str(val)) if last_val is not kNoValue and not callable(last_val): file.write('\n') import_line = None # Trick to get init docs from class level doc where it usually is if name == '__init__': doc1 = getattr(val, '__doc__', '') if doc1 is None: doc1 = '' doc2 = getattr(obj, '__doc__', '') if doc2 is None: doc2 = '' args, retval = _GetCallableSignature(doc2 + '\n' + doc1) elif kIronPython: args, import_line, retval = GetCallableSignatureIPy(val, obj, name) else: args, retval = GetCallableSignature(val) # Try to fall back on inspect, tho this only works if the module # is not an extension module or contains some methods defined in Python if args == '' and inspect is not None: fval = getattr(val, 'im_func', val) try: args = inspect.formatargspec(inspect.getargspec(fval)[0]) args = args[1:-1] except: args = '' if type(obj) == PyType_Type or hasattr(val, 'im_func'): if len(args) > 0 and string_find(args, 'self') != 0: args = 'self, ' + args elif string_find(args, 'self') < 0: args = 'self' # Write the definition file.write(indent + 'def %s(%s):\n' % (name, args)) WriteDocString(val, file, indent + ' ') if import_line is not None: file.write(indent + ' %s\n' % import_line) file.write(indent + ' %s\n\n' % retval) num_written = num_written + 1 # Value is a constant elif type(val) in kStringTypes: use_triple = (string_find(val, '\n') > 0 and string_find(val, '"""') == -1) val = val.replace('\r\n', '\n') if use_triple: for c in val: if c not in printable_chars: use_triple = 0 break if use_triple: val = '"""%s"""' % str(val) else: val = repr(val) file.write(indent + '%s = %s\n' % (name, val)) num_written = num_written + 1 elif type(val) in kLiteralTypes: file.write(indent + '%s = %s\n' % (name, str(val))) num_written = num_written + 1 elif type(val) in kStructureTypes: if type(val) == kListType: s = '[]' elif type(val) == kTupleType: s = '()' elif type(val) == kDictType: s = '{}' file.write(indent + '%s = %s\n' % (name, s)) num_written = num_written + 1 elif type(val) in kFileTypes: file.write(indent + '%s = file(__file__)\n' % name) num_written = num_written + 1 else: file.write(indent + '%s = None\n' % name) num_written = num_written + 1 last_val = val # Add in appropriate overrides additions = {} for oname in ValueItems(overrides): if oname in kOmitTopLevel: continue oval = GetValue(overrides, oname, None) # Class overrides already processed as addition to existing class # are not shown again if type(oval) in (kListType, kDictType) or not has_key(seen_vals, oval): additions[oname] = oval last_val = oval if len(additions) > 0: try: seen_vals[overrides] = 1 except TypeError: sys.stderr.write("Failed to register %s (may cause duplicates)\n" % str(overrides)) file.write(indent + '# BEGIN MANUAL OVERRIDES FROM %s\n' % overrides_file) GenForObject(scope, GetValue(overrides, name, {}), overrides_file, additions, file, seen_vals, indent, use_inspect=1) file.write(indent + '# END MANUAL OVERRIDES\n') if last_val is not kNoValue and not callable(last_val): file.write('\n') return num_written def GenPClassForType(scope, overrides, overrides_file, name, obj, file, seen_vals, indent = '', use_inspect=0): bases = "" if inspect is not None and hasattr(inspect, 'getmro'): baselist = inspect.getmro(obj) if baselist[0] is obj and len(baselist) > 1: bases = "(%s)" % baselist[1].__name__ file.write(indent + 'class %s%s:\n' % (name, bases)) WriteDocString(obj, file, indent + kOneLevelIndent) file.write('\n') num_written = GenForObject(scope, overrides, overrides_file, obj, file, seen_vals, indent + kOneLevelIndent, use_inspect) if num_written == 0: file.write(indent + ' ' + 'pass\n\n') def ProcessModule(mod_name, magic_code=None, metadata={}, overrides={}, file=sys.stdout): if string_find(mod_name, '-') >= 0: return seen_vals = {} if type(overrides) == types.ModuleType: overrides_file = overrides.__file__ overrides_file = string_join(string_split(overrides_file, os.sep)[-3:], os.sep) if overrides_file[-1] in ('o', 'c'): overrides_file = overrides_file[:-1] else: overrides_file = None namespace = {} if magic_code is not None: try: exec('import %s' % mod_name, namespace) except: exec(magic_code, namespace) if not has_key(namespace, mod_name): exec('import %s' % mod_name, namespace) else: # Work-around for bug in 2.1's warning module if version == (2, 1) and mod_name == 'regex': namespace['__name__'] = mod_name try: exec('import %s' % mod_name, namespace) except ImportError: # This case sometimes succeeds when above fails but adding '.' to # path blindly seems like a bad idea so do it only as a fall-back sys.path.append('.') exec('import %s' % mod_name, namespace) print >>sys.stderr, "Now imported", namespace.keys() file.write('# AUTO-GENERATED FILE -- DO NOT EDIT\n') if overrides_file is not None: file.write('# OVERRIDES FILE: %s\n\n' % overrides_file) else: file.write('\n') components = mod_name.split('.') mod_name = components[0] components = components[1:] mod = namespace.get(mod_name) while components: mod = getattr(mod, components[0]) del components[0] WriteDocString(mod, file, '') file.write('\n') GenForObject(['__toplevel__'], overrides, overrides_file, mod, file, seen_vals) fn = getattr(mod, '__file__', None) if fn is not None: try: fn = os.path.abspath(fn) except: metadata['file'] = repr(fn) else: metadata['file'] = fn try: if os.path.isfile(fn): s = os.stat(fn) metadata['modtime'] = s[stat.ST_MTIME] except: pass def DecodeArg(a): """ If a starts with +, treat any subsequent + as indicator that the 4 characters following are the hex representation of a code point which will replace the + and the following 4 chars. Any / will be translated into os.sep. On win32, always convert to unicode if possible. """ if os.sep != '/': a = string_replace(a, '/', os.sep) if a == '' or a[0] != '+': if sys.platform == 'win32': try: return unicode(a) except: pass return a try: a = unicode(a) except: pass part_list = string_split(a[1:], '+') fragments = [] for i, part in enumerate(part_list): if i == 0: frag = part else: try: frag = unichr(string.atoi(part[:4], 16)) except: frag = chr(string.atoi(part[:4], 16)) frag = frag + part[4:] if frag != '': fragments.append(frag) try: decoded = string_join(fragments, unicode('')) except: decoded = string_join(fragments, '') return decoded def _PrintUsage(): sys.stdout.write("Usage: %s [options] module-name output-file [python-path] [ld_library_path]\n") sys.stdout.write("output-file may be '-' to write to stdout. The paths, if given, should be\n") sys.stdout.write("string separated by os.pathsep (':' or ';' depending on OS). The output-file\n") sys.stdout.write("is treated as an encoded name if the name begins with a + and any + after\n") sys.stdout.write("the first is to be followed by a 4 digit hexadecimal number for the code\n") sys.stdout.write("point at that position in the string\n") sys.stdout.write("\n") sys.stdout.write("Valid options:\n") sys.stdout.write(" --meta-data -- Write meta data file also -- the file name is output-file\n") sys.stdout.write(" plus '.meta'. This arg is ignored if output-file is '-'\n") sys.stdout.write(" --magic-code [code] -- Execute given code before attempting to import the \n") sys.stdout.write(" extension module if importing it alone fails\n") sys.stdout.write("\n") sys.stdout.write("The meta data file, if output, contains a cPickle.dump()'ed Python dictionary\n") sys.stdout.write("with the following fields:\n") sys.stdout.write(" file -- The module file name\n") sys.stdout.write(" modtime -- The value of os.stat(file)[stat.M_TIME]\n") sys.stdout.write("\n") ############################################################## # IronPython specifics class CIronPythonNameSpaceFinder: def __init__(self): self.fNamespaces = {} self.fInspected = set() import System self.fNamespaceType = type(System) def Visit(self, obj, mod_name): if obj in self.fInspected: return self.fInspected.add(obj) # XXXX was skipping top level namespaces!!! self.fNamespaces[mod_name] = obj for name in SafeDir(obj): try: val = getattr(obj, name) except: continue if isinstance(val, self.fNamespaceType): dotted_name = '%s.%s' % (mod_name, name) self.fNamespaces[dotted_name] = val self.Visit(val, dotted_name) def GenerateForIronPythonNameSpaces(namespace_list, output_dirname, clr_refs): import clr for name in clr_refs: clr.AddReference(name) ns_finder = CIronPythonNameSpaceFinder() for ns_name in namespace_list: mod = __import__(ns_name) ns_finder.Visit(mod, ns_name) print ns_finder.fNamespaces.keys() print len(ns_finder.fNamespaces) expanded_list = sorted(ns_finder.fNamespaces.keys()) for ns_name in expanded_list: kPIPackagesAreBroken = True if not kPIPackagesAreBroken: # if Wing worked like it should dirname = os.path.join(output_dirname, *ns_name.split('.')) if not os.path.exists(dirname): os.makedirs(dirname) filename = os.path.join(dirname, '__init__.pi') else: filename = os.path.join(output_dirname, ns_name + '.pi') f = open(filename, 'w') ProcessModule(ns_name, file=f) f.close() ######################################################################## if __name__ == '__main__': if '--ironpython' in sys.argv: # CHANGE: Hardwire assembly names to add references and pre-do imports to # make nested namespaces available as attributes on top level namespace ref_names = ['System', 'System.Data', 'System.Windows.Forms', 'System.Drawing', 'System.Xml'] output_dir = os.path.abspath('.') GenerateForIronPythonNameSpaces(['System', 'Microsoft', 'clr'], output_dir, ref_names) sys.exit(0) if len(sys.argv) < 3: _PrintUsage() sys.exit(1) write_metadata = 0 if '--meta-data' in sys.argv: sys.argv.remove('--meta-data') write_metadata = 1 magic_code = None if '--magic-code' in sys.argv: mpos = sys.argv.index('--magic-code') if len(sys.argv) < mpos + 2: _PrintUsage() sys.exit(1) magic_code = sys.argv[sys.argv.index('--magic-code')+1] sys.argv.remove('--magic-code') sys.argv.remove(magic_code) modname = sys.argv[1] outfile = sys.argv[2] if len(sys.argv) >= 4: for p in string_split(sys.argv[3], os.pathsep): sys.path.append(os.path.abspath(p)) if len(sys.argv) == 5: os.environ['LD_LIBRARY_PATH'] = sys.argv[4] cmd = '"%s" %s "%s" "%s" "%s"' % (sys.executable, sys.argv[0], sys.argv[1], sys.argv[2], sys.argv[3]) sys.exit(os.system(cmd)) if outfile != '-': outfile = DecodeArg(outfile) f = open(outfile, 'w') sys.stdout = f else: f = sys.stdout metadata = {} ProcessModule(modname, magic_code, metadata, file=f) if outfile != '-': sys.stdout = sys.__stdout__ f.close() # Avoid leaving around lots of empty trash files if os.stat(outfile)[stat.ST_SIZE] == 0: try: sys.stdout.write("generate_pi.py: Removing empty %s\n" % outfile) os.unlink(outfile) except: pass if write_metadata and outfile != '-': f = open(outfile + '.meta', 'w') import cPickle cPickle.dump(metadata, f) f.close() """ I changed the references / assemblies to include Microsoft, System.Xml and clr. Still need command line argument handling for these. Unicode docstrings seem to be broken in pi files (don't appear in the source-assistant). I've switched them back to non-Unicode. Changed the regular expression to exclude the indexing from return types. (Array[int]() becomes Array().) For .NET types with public constructors an __init__ method could be constructed from the type docstring. Currently they all show in the source assistant as *args. Namespace finder was skipping the top-level namespaces. Fixed. I see your note about PI file packages. It will be nice to have that working. :-) Not covered attribute types (all set to None): * field descriptors (enumerations) In .NET enumerations behave much more like instances than classes. The field descriptor docstrings are usually not helpful but I'm also sure that None is not helpful as a value. It should really be a property returning an instance of the enumeration. The enumaration fireld has an underlying value (accessed as .value__) but I don't think setting it to that would be useful. * events (very common and there may be multiple types of events) * indexers (Item) * attributes that are classes or enumerations. The docstrings are useful. e.g. System.ActivationContext.ContextForm: >>> ActivationContext.__dict__['ContextForm'].__doc__ '\r\n\r\nenum ContextForm, values: Loose (0), StoreBounded (1)\r\n\r\n' Return type inferencing from properties would be very useful. (Feature request!!) :-) Cross namespace returns now do: import System.Drawing return System.Drawing.Point() Something similar will need to be done with base-classes. Many classes inherit from types in different namespaces - so in the PI file the class they inherit from will probably need to be imported (and preferably also excluded from completion). If I add the import at the start of the module - with a corresponding list of deletes at the end will this work? For most .NET methods the parameter names are determined using reflection. This doesn't work for some - particularly methods like ReferenceEquals, ToObject and so on. (Are these inherited methods? Needs investigating.) For these we fallback to __GetCallableSignature which doesn't get the argument names from .NET methods (not really a big deal). This is because they are of the form: int Add(self, CodeCatchClause value) __GetCallableSignature also fails to handle indexing in type signature for methods (use of square brackets) is interpreted as optional arguments. E.g.: System.ModuleHandle.ResolveFieldHandle(self, arg0, arg1=None, arg2=None) from: RuntimeFieldHandle ResolveFieldHandle(self, int fieldToken, Array[RuntimeTypeHandle] typeInstantiationContext, Array[RuntimeTypeHandle] methodInstantiationContext) """