123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422 |
- # ##### BEGIN GPL LICENSE BLOCK #####
- #
- # This program is free software; you can redistribute it and/or
- # modify it under the terms of the GNU General Public License
- # as published by the Free Software Foundation; either version 2
- # of the License, or (at your option) any later version.
- #
- # This program is distributed in the hope that it will be useful,
- # but WITHOUT ANY WARRANTY; without even the implied warranty of
- # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
- # GNU General Public License for more details.
- #
- # You should have received a copy of the GNU General Public License
- # along with this program; if not, write to the Free Software Foundation,
- # Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
- #
- # ##### END GPL LICENSE BLOCK #####
- # <pep8-80 compliant>
- """
- This module has a similar scope to os.path, containing utility
- functions for dealing with paths in Blender.
- """
- __all__ = (
- "abspath",
- "basename",
- "clean_name",
- "display_name",
- "display_name_to_filepath",
- "display_name_from_filepath",
- "ensure_ext",
- "extensions_image",
- "extensions_movie",
- "extensions_audio",
- "is_subdir",
- "module_names",
- "native_pathsep",
- "reduce_dirs",
- "relpath",
- "resolve_ncase",
- )
- import bpy as _bpy
- import os as _os
- from _bpy_path import (
- extensions_audio,
- extensions_movie,
- extensions_image,
- )
- def _getattr_bytes(var, attr):
- return var.path_resolve(attr, False).as_bytes()
- def abspath(path, start=None, library=None):
- """
- Returns the absolute path relative to the current blend file
- using the "//" prefix.
- :arg start: Relative to this path,
- when not set the current filename is used.
- :type start: string or bytes
- :arg library: The library this path is from. This is only included for
- convenience, when the library is not None its path replaces *start*.
- :type library: :class:`bpy.types.Library`
- """
- if isinstance(path, bytes):
- if path.startswith(b"//"):
- if library:
- start = _os.path.dirname(
- abspath(_getattr_bytes(library, "filepath")))
- return _os.path.join(
- _os.path.dirname(_getattr_bytes(_bpy.data, "filepath"))
- if start is None else start,
- path[2:],
- )
- else:
- if path.startswith("//"):
- if library:
- start = _os.path.dirname(
- abspath(library.filepath))
- return _os.path.join(
- _os.path.dirname(_bpy.data.filepath)
- if start is None else start,
- path[2:],
- )
- return path
- def relpath(path, start=None):
- """
- Returns the path relative to the current blend file using the "//" prefix.
- :arg path: An absolute path.
- :type path: string or bytes
- :arg start: Relative to this path,
- when not set the current filename is used.
- :type start: string or bytes
- """
- if isinstance(path, bytes):
- if not path.startswith(b"//"):
- if start is None:
- start = _os.path.dirname(_getattr_bytes(_bpy.data, "filepath"))
- return b"//" + _os.path.relpath(path, start)
- else:
- if not path.startswith("//"):
- if start is None:
- start = _os.path.dirname(_bpy.data.filepath)
- return "//" + _os.path.relpath(path, start)
- return path
- def is_subdir(path, directory):
- """
- Returns true if *path* in a subdirectory of *directory*.
- Both paths must be absolute.
- :arg path: An absolute path.
- :type path: string or bytes
- """
- from os.path import normpath, normcase, sep
- path = normpath(normcase(path))
- directory = normpath(normcase(directory))
- if len(path) > len(directory):
- sep = sep.encode('ascii') if isinstance(directory, bytes) else sep
- if path.startswith(directory.rstrip(sep) + sep):
- return True
- return False
- def clean_name(name, replace="_"):
- """
- Returns a name with characters replaced that
- may cause problems under various circumstances,
- such as writing to a file.
- All characters besides A-Z/a-z, 0-9 are replaced with "_"
- or the *replace* argument if defined.
- """
- if replace != "_":
- if len(replace) != 1 or ord(replace) > 255:
- raise ValueError("Value must be a single ascii character")
- def maketrans_init():
- trans_cache = clean_name._trans_cache
- trans = trans_cache.get(replace)
- if trans is None:
- bad_chars = (
- 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
- 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f,
- 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17,
- 0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f,
- 0x20, 0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27,
- 0x28, 0x29, 0x2a, 0x2b, 0x2c, 0x2e, 0x2f, 0x3a,
- 0x3b, 0x3c, 0x3d, 0x3e, 0x3f, 0x40, 0x5b, 0x5c,
- 0x5d, 0x5e, 0x60, 0x7b, 0x7c, 0x7d, 0x7e, 0x7f,
- 0x80, 0x81, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87,
- 0x88, 0x89, 0x8a, 0x8b, 0x8c, 0x8d, 0x8e, 0x8f,
- 0x90, 0x91, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97,
- 0x98, 0x99, 0x9a, 0x9b, 0x9c, 0x9d, 0x9e, 0x9f,
- 0xa0, 0xa1, 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7,
- 0xa8, 0xa9, 0xaa, 0xab, 0xac, 0xad, 0xae, 0xaf,
- 0xb0, 0xb1, 0xb2, 0xb3, 0xb4, 0xb5, 0xb6, 0xb7,
- 0xb8, 0xb9, 0xba, 0xbb, 0xbc, 0xbd, 0xbe, 0xbf,
- 0xc0, 0xc1, 0xc2, 0xc3, 0xc4, 0xc5, 0xc6, 0xc7,
- 0xc8, 0xc9, 0xca, 0xcb, 0xcc, 0xcd, 0xce, 0xcf,
- 0xd0, 0xd1, 0xd2, 0xd3, 0xd4, 0xd5, 0xd6, 0xd7,
- 0xd8, 0xd9, 0xda, 0xdb, 0xdc, 0xdd, 0xde, 0xdf,
- 0xe0, 0xe1, 0xe2, 0xe3, 0xe4, 0xe5, 0xe6, 0xe7,
- 0xe8, 0xe9, 0xea, 0xeb, 0xec, 0xed, 0xee, 0xef,
- 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7,
- 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe,
- )
- trans = str.maketrans({char: replace for char in bad_chars})
- trans_cache[replace] = trans
- return trans
- trans = maketrans_init()
- return name.translate(trans)
- clean_name._trans_cache = {}
- def _clean_utf8(name):
- if type(name) == bytes:
- return name.decode("utf8", "replace")
- else:
- return name.encode("utf8", "replace").decode("utf8")
- _display_name_literals = {
- ":": "_colon_",
- "+": "_plus_",
- }
- def display_name(name, *, has_ext=True):
- """
- Creates a display string from name to be used menus and the user interface.
- Capitalize the first letter in all lowercase names,
- mixed case names are kept as is. Intended for use with
- filenames and module names.
- """
- if has_ext:
- name = _os.path.splitext(basename(name))[0]
- # string replacements
- for disp_value, file_value in _display_name_literals.items():
- name = name.replace(file_value, disp_value)
- # strip to allow underscore prefix
- # (when paths can't start with numbers for eg).
- name = name.replace("_", " ").lstrip(" ")
- if name.islower():
- name = name.lower().title()
- name = _clean_utf8(name)
- return name
- def display_name_to_filepath(name):
- """
- Performs the reverse of display_name using literal versions of characters
- which aren't supported in a filepath.
- """
- for disp_value, file_value in _display_name_literals.items():
- name = name.replace(disp_value, file_value)
- return name
- def display_name_from_filepath(name):
- """
- Returns the path stripped of directory and extension,
- ensured to be utf8 compatible.
- """
- name = _os.path.splitext(basename(name))[0]
- name = _clean_utf8(name)
- return name
- def resolve_ncase(path):
- """
- Resolve a case insensitive path on a case sensitive system,
- returning a string with the path if found else return the original path.
- """
- def _ncase_path_found(path):
- if not path or _os.path.exists(path):
- return path, True
- # filename may be a directory or a file
- filename = _os.path.basename(path)
- dirpath = _os.path.dirname(path)
- suffix = path[:0] # "" but ensure byte/str match
- if not filename: # dir ends with a slash?
- if len(dirpath) < len(path):
- suffix = path[:len(path) - len(dirpath)]
- filename = _os.path.basename(dirpath)
- dirpath = _os.path.dirname(dirpath)
- if not _os.path.exists(dirpath):
- if dirpath == path:
- return path, False
- dirpath, found = _ncase_path_found(dirpath)
- if not found:
- return path, False
- # at this point, the directory exists but not the file
- # we are expecting 'dirpath' to be a directory, but it could be a file
- if _os.path.isdir(dirpath):
- try:
- files = _os.listdir(dirpath)
- except PermissionError:
- # We might not have the permission to list dirpath...
- return path, False
- else:
- return path, False
- filename_low = filename.lower()
- f_iter_nocase = None
- for f_iter in files:
- if f_iter.lower() == filename_low:
- f_iter_nocase = f_iter
- break
- if f_iter_nocase:
- return _os.path.join(dirpath, f_iter_nocase) + suffix, True
- else:
- # can't find the right one, just return the path as is.
- return path, False
- ncase_path, found = _ncase_path_found(path)
- return ncase_path if found else path
- def ensure_ext(filepath, ext, case_sensitive=False):
- """
- Return the path with the extension added if it is not already set.
- :arg ext: The extension to check for, can be a compound extension. Should
- start with a dot, such as '.blend' or '.tar.gz'.
- :type ext: string
- :arg case_sensitive: Check for matching case when comparing extensions.
- :type case_sensitive: bool
- """
- if case_sensitive:
- if filepath.endswith(ext):
- return filepath
- else:
- if filepath[-len(ext):].lower().endswith(ext.lower()):
- return filepath
- return filepath + ext
- def module_names(path, recursive=False):
- """
- Return a list of modules which can be imported from *path*.
- :arg path: a directory to scan.
- :type path: string
- :arg recursive: Also return submodule names for packages.
- :type recursive: bool
- :return: a list of string pairs (module_name, module_file).
- :rtype: list
- """
- from os.path import join, isfile
- modules = []
- for filename in sorted(_os.listdir(path)):
- if filename == "modules":
- pass # XXX, hard coded exception.
- elif filename.endswith(".py") and filename != "__init__.py":
- fullpath = join(path, filename)
- modules.append((filename[0:-3], fullpath))
- elif "." not in filename:
- directory = join(path, filename)
- fullpath = join(directory, "__init__.py")
- if isfile(fullpath):
- modules.append((filename, fullpath))
- if recursive:
- for mod_name, mod_path in module_names(directory, True):
- modules.append(("%s.%s" % (filename, mod_name),
- mod_path,
- ))
- return modules
- def basename(path):
- """
- Equivalent to os.path.basename, but skips a "//" prefix.
- Use for Windows compatibility.
- """
- return _os.path.basename(path[2:] if path[:2] in {"//", b"//"} else path)
- def native_pathsep(path):
- """
- Replace the path separator with the systems native ``os.sep``.
- """
- if type(path) is str:
- if _os.sep == "/":
- return path.replace("\\", "/")
- else:
- if path.startswith("//"):
- return "//" + path[2:].replace("/", "\\")
- else:
- return path.replace("/", "\\")
- else: # bytes
- if _os.sep == "/":
- return path.replace(b"\\", b"/")
- else:
- if path.startswith(b"//"):
- return b"//" + path[2:].replace(b"/", b"\\")
- else:
- return path.replace(b"/", b"\\")
- def reduce_dirs(dirs):
- """
- Given a sequence of directories, remove duplicates and
- any directories nested in one of the other paths.
- (Useful for recursive path searching).
- :arg dirs: Sequence of directory paths.
- :type dirs: sequence
- :return: A unique list of paths.
- :rtype: list
- """
- dirs = list({_os.path.normpath(_os.path.abspath(d)) for d in dirs})
- dirs.sort(key=lambda d: len(d))
- for i in range(len(dirs) - 1, -1, -1):
- for j in range(i):
- print(i, j)
- if len(dirs[i]) == len(dirs[j]):
- break
- elif is_subdir(dirs[i], dirs[j]):
- del dirs[i]
- break
- return dirs
|