diff options
Diffstat (limited to 'jam-files/boost-build/util/path.py')
-rw-r--r-- | jam-files/boost-build/util/path.py | 904 |
1 files changed, 904 insertions, 0 deletions
diff --git a/jam-files/boost-build/util/path.py b/jam-files/boost-build/util/path.py new file mode 100644 index 00000000..222b96bf --- /dev/null +++ b/jam-files/boost-build/util/path.py @@ -0,0 +1,904 @@ +# Status: this module is ported on demand by however needs something +# from it. Functionality that is not needed by Python port will +# be dropped. + +# Copyright (C) Vladimir Prus 2002. Permission to copy, use, modify, sell and +# distribute this software is granted provided this copyright notice appears in +# all copies. This software is provided "as is" without express or implied +# warranty, and with no claim as to its suitability for any purpose. + +# Performs various path manipulations. Path are always in a 'normilized' +# representation. In it, a path may be either: +# +# - '.', or +# +# - ['/'] [ ( '..' '/' )* (token '/')* token ] +# +# In plain english, path can be rooted, '..' elements are allowed only +# at the beginning, and it never ends in slash, except for path consisting +# of slash only. + +import os.path +from utility import to_seq +from glob import glob as builtin_glob + +from b2.util import bjam_signature + +@bjam_signature((["path", "root"],)) +def root (path, root): + """ If 'path' is relative, it is rooted at 'root'. Otherwise, it's unchanged. + """ + if os.path.isabs (path): + return path + else: + return os.path.join (root, path) + +@bjam_signature((["native"],)) +def make (native): + """ Converts the native path into normalized form. + """ + # TODO: make os selection here. + return make_UNIX (native) + +def make_UNIX (native): + + # VP: I have no idea now 'native' can be empty here! But it can! + assert (native) + + return os.path.normpath (native) + +@bjam_signature((["path"],)) +def native (path): + """ Builds a native representation of the path. + """ + # TODO: make os selection here. + return native_UNIX (path) + +def native_UNIX (path): + return path + + +def pwd (): + """ Returns the current working directory. + # TODO: is it a good idea to use the current dir? Some use-cases + may not allow us to depend on the current dir. + """ + return make (os.getcwd ()) + +def is_rooted (path): + """ Tests if a path is rooted. + """ + return path and path [0] == '/' + + +################################################################### +# Still to port. +# Original lines are prefixed with "# " +# +# # Copyright (C) Vladimir Prus 2002. Permission to copy, use, modify, sell and +# # distribute this software is granted provided this copyright notice appears in +# # all copies. This software is provided "as is" without express or implied +# # warranty, and with no claim as to its suitability for any purpose. +# +# # Performs various path manipulations. Path are always in a 'normilized' +# # representation. In it, a path may be either: +# # +# # - '.', or +# # +# # - ['/'] [ ( '..' '/' )* (token '/')* token ] +# # +# # In plain english, path can be rooted, '..' elements are allowed only +# # at the beginning, and it never ends in slash, except for path consisting +# # of slash only. +# +# import modules ; +# import sequence ; +# import regex ; +# import errors : error ; +# +# +# os = [ modules.peek : OS ] ; +# if [ modules.peek : UNIX ] +# { +# local uname = [ modules.peek : JAMUNAME ] ; +# switch $(uname) +# { +# case CYGWIN* : +# os = CYGWIN ; +# +# case * : +# os = UNIX ; +# } +# } +# +# # +# # Tests if a path is rooted. +# # +# rule is-rooted ( path ) +# { +# return [ MATCH "^(/)" : $(path) ] ; +# } +# +# # +# # Tests if a path has a parent. +# # +# rule has-parent ( path ) +# { +# if $(path) != / { +# return 1 ; +# } else { +# return ; +# } +# } +# +# # +# # Returns the path without any directory components. +# # +# rule basename ( path ) +# { +# return [ MATCH "([^/]+)$" : $(path) ] ; +# } +# +# # +# # Returns parent directory of the path. If no parent exists, error is issued. +# # +# rule parent ( path ) +# { +# if [ has-parent $(path) ] { +# +# if $(path) = . { +# return .. ; +# } else { +# +# # Strip everything at the end of path up to and including +# # the last slash +# local result = [ regex.match "((.*)/)?([^/]+)" : $(path) : 2 3 ] ; +# +# # Did we strip what we shouldn't? +# if $(result[2]) = ".." { +# return $(path)/.. ; +# } else { +# if ! $(result[1]) { +# if [ is-rooted $(path) ] { +# result = / ; +# } else { +# result = . ; +# } +# } +# return $(result[1]) ; +# } +# } +# } else { +# error "Path '$(path)' has no parent" ; +# } +# } +# +# # +# # Returns path2 such that "[ join path path2 ] = .". +# # The path may not contain ".." element or be rooted. +# # +# rule reverse ( path ) +# { +# if $(path) = . +# { +# return $(path) ; +# } +# else +# { +# local tokens = [ regex.split $(path) "/" ] ; +# local tokens2 ; +# for local i in $(tokens) { +# tokens2 += .. ; +# } +# return [ sequence.join $(tokens2) : "/" ] ; +# } +# } +# +# # +# # Auxillary rule: does all the semantic of 'join', except for error cheching. +# # The error checking is separated because this rule is recursive, and I don't +# # like the idea of checking the same input over and over. +# # +# local rule join-imp ( elements + ) +# { +# return [ NORMALIZE_PATH $(elements:J="/") ] ; +# } +# +# # +# # Contanenates the passed path elements. Generates an error if +# # any element other than the first one is rooted. +# # +# rule join ( elements + ) +# { +# if ! $(elements[2]) +# { +# return $(elements[1]) ; +# } +# else +# { +# for local e in $(elements[2-]) +# { +# if [ is-rooted $(e) ] +# { +# error only first element may be rooted ; +# } +# } +# return [ join-imp $(elements) ] ; +# } +# } + + +def glob (dirs, patterns): + """ Returns the list of files matching the given pattern in the + specified directory. Both directories and patterns are + supplied as portable paths. Each pattern should be non-absolute + path, and can't contain "." or ".." elements. Each slash separated + element of pattern can contain the following special characters: + - '?', which match any character + - '*', which matches arbitrary number of characters. + A file $(d)/e1/e2/e3 (where 'd' is in $(dirs)) matches pattern p1/p2/p3 + if and only if e1 matches p1, e2 matches p2 and so on. + + For example: + [ glob . : *.cpp ] + [ glob . : */build/Jamfile ] + """ +# { +# local result ; +# if $(patterns:D) +# { +# # When a pattern has a directory element, we first glob for +# # directory, and then glob for file name is the found directories. +# for local p in $(patterns) +# { +# # First glob for directory part. +# local globbed-dirs = [ glob $(dirs) : $(p:D) ] ; +# result += [ glob $(globbed-dirs) : $(p:D="") ] ; +# } +# } +# else +# { +# # When a pattern has not directory, we glob directly. +# # Take care of special ".." value. The "GLOB" rule simply ignores +# # the ".." element (and ".") element in directory listings. This is +# # needed so that +# # +# # [ glob libs/*/Jamfile ] +# # +# # don't return +# # +# # libs/../Jamfile (which is the same as ./Jamfile) +# # +# # On the other hand, when ".." is explicitly present in the pattern +# # we need to return it. +# # +# for local dir in $(dirs) +# { +# for local p in $(patterns) +# { +# if $(p) != ".." +# { +# result += [ sequence.transform make +# : [ GLOB [ native $(dir) ] : $(p) ] ] ; +# } +# else +# { +# result += [ path.join $(dir) .. ] ; +# } +# } +# } +# } +# return $(result) ; +# } +# + +# TODO: (PF) I replaced the code above by this. I think it should work but needs to be tested. + result = [] + dirs = to_seq (dirs) + patterns = to_seq (patterns) + + splitdirs = [] + for dir in dirs: + splitdirs += dir.split (os.pathsep) + + for dir in splitdirs: + for pattern in patterns: + p = os.path.join (dir, pattern) + import glob + result.extend (glob.glob (p)) + return result + +# +# Find out the absolute name of path and returns the list of all the parents, +# starting with the immediate one. Parents are returned as relative names. +# If 'upper_limit' is specified, directories above it will be pruned. +# +def all_parents(path, upper_limit=None, cwd=None): + + if not cwd: + cwd = os.getcwd() + + path_abs = os.path.join(cwd, path) + + if upper_limit: + upper_limit = os.path.join(cwd, upper_limit) + + result = [] + while path_abs and path_abs != upper_limit: + (head, tail) = os.path.split(path) + path = os.path.join(path, "..") + result.append(path) + path_abs = head + + if upper_limit and path_abs != upper_limit: + raise BaseException("'%s' is not a prefix of '%s'" % (upper_limit, path)) + + return result + +# Search for 'pattern' in parent directories of 'dir', up till and including +# 'upper_limit', if it is specified, or till the filesystem root otherwise. +# +def glob_in_parents(dir, patterns, upper_limit=None): + + result = [] + parent_dirs = all_parents(dir, upper_limit) + + for p in parent_dirs: + result = glob(p, patterns) + if result: break + + return result + +# +# # +# # Assuming 'child' is a subdirectory of 'parent', return the relative +# # path from 'parent' to 'child' +# # +# rule relative ( child parent ) +# { +# if $(parent) = "." +# { +# return $(child) ; +# } +# else +# { +# local split1 = [ regex.split $(parent) / ] ; +# local split2 = [ regex.split $(child) / ] ; +# +# while $(split1) +# { +# if $(split1[1]) = $(split2[1]) +# { +# split1 = $(split1[2-]) ; +# split2 = $(split2[2-]) ; +# } +# else +# { +# errors.error $(child) is not a subdir of $(parent) ; +# } +# } +# return [ join $(split2) ] ; +# } +# } +# +# # Returns the minimal path to path2 that is relative path1. +# # +# rule relative-to ( path1 path2 ) +# { +# local root_1 = [ regex.split [ reverse $(path1) ] / ] ; +# local split1 = [ regex.split $(path1) / ] ; +# local split2 = [ regex.split $(path2) / ] ; +# +# while $(split1) && $(root_1) +# { +# if $(split1[1]) = $(split2[1]) +# { +# root_1 = $(root_1[2-]) ; +# split1 = $(split1[2-]) ; +# split2 = $(split2[2-]) ; +# } +# else +# { +# split1 = ; +# } +# } +# return [ join . $(root_1) $(split2) ] ; +# } + +# Returns the list of paths which are used by the operating system +# for looking up programs +def programs_path (): + raw = [] + names = ['PATH', 'Path', 'path'] + + for name in names: + raw.append(os.environ.get (name, '')) + + result = [] + for elem in raw: + if elem: + for p in elem.split(os.path.pathsep): + result.append(make(p)) + + return result + +# rule make-NT ( native ) +# { +# local tokens = [ regex.split $(native) "[/\\]" ] ; +# local result ; +# +# # Handle paths ending with slashes +# if $(tokens[-1]) = "" +# { +# tokens = $(tokens[1--2]) ; # discard the empty element +# } +# +# result = [ path.join $(tokens) ] ; +# +# if [ regex.match "(^.:)" : $(native) ] +# { +# result = /$(result) ; +# } +# +# if $(native) = "" +# { +# result = "." ; +# } +# +# return $(result) ; +# } +# +# rule native-NT ( path ) +# { +# local result = [ MATCH "^/?(.*)" : $(path) ] ; +# result = [ sequence.join [ regex.split $(result) "/" ] : "\\" ] ; +# return $(result) ; +# } +# +# rule make-CYGWIN ( path ) +# { +# return [ make-NT $(path) ] ; +# } +# +# rule native-CYGWIN ( path ) +# { +# local result = $(path) ; +# if [ regex.match "(^/.:)" : $(path) ] # win absolute +# { +# result = [ MATCH "^/?(.*)" : $(path) ] ; # remove leading '/' +# } +# return [ native-UNIX $(result) ] ; +# } +# +# # +# # split-VMS: splits input native path into +# # device dir file (each part is optional), +# # example: +# # +# # dev:[dir]file.c => dev: [dir] file.c +# # +# rule split-path-VMS ( native ) +# { +# local matches = [ MATCH ([a-zA-Z0-9_-]+:)?(\\[[^\]]*\\])?(.*)?$ : $(native) ] ; +# local device = $(matches[1]) ; +# local dir = $(matches[2]) ; +# local file = $(matches[3]) ; +# +# return $(device) $(dir) $(file) ; +# } +# +# # +# # Converts a native VMS path into a portable path spec. +# # +# # Does not handle current-device absolute paths such +# # as "[dir]File.c" as it is not clear how to represent +# # them in the portable path notation. +# # +# # Adds a trailing dot (".") to the file part if no extension +# # is present (helps when converting it back into native path). +# # +# rule make-VMS ( native ) +# { +# if [ MATCH ^(\\[[a-zA-Z0-9]) : $(native) ] +# { +# errors.error "Can't handle default-device absolute paths: " $(native) ; +# } +# +# local parts = [ split-path-VMS $(native) ] ; +# local device = $(parts[1]) ; +# local dir = $(parts[2]) ; +# local file = $(parts[3]) ; +# local elems ; +# +# if $(device) +# { +# # +# # rooted +# # +# elems = /$(device) ; +# } +# +# if $(dir) = "[]" +# { +# # +# # Special case: current directory +# # +# elems = $(elems) "." ; +# } +# else if $(dir) +# { +# dir = [ regex.replace $(dir) "\\[|\\]" "" ] ; +# local dir_parts = [ regex.split $(dir) \\. ] ; +# +# if $(dir_parts[1]) = "" +# { +# # +# # Relative path +# # +# dir_parts = $(dir_parts[2--1]) ; +# } +# +# # +# # replace "parent-directory" parts (- => ..) +# # +# dir_parts = [ regex.replace-list $(dir_parts) : - : .. ] ; +# +# elems = $(elems) $(dir_parts) ; +# } +# +# if $(file) +# { +# if ! [ MATCH (\\.) : $(file) ] +# { +# # +# # Always add "." to end of non-extension file +# # +# file = $(file). ; +# } +# elems = $(elems) $(file) ; +# } +# +# local portable = [ path.join $(elems) ] ; +# +# return $(portable) ; +# } +# +# # +# # Converts a portable path spec into a native VMS path. +# # +# # Relies on having at least one dot (".") included in the file +# # name to be able to differentiate it ftom the directory part. +# # +# rule native-VMS ( path ) +# { +# local device = "" ; +# local dir = $(path) ; +# local file = "" ; +# local native ; +# local split ; +# +# # +# # Has device ? +# # +# if [ is-rooted $(dir) ] +# { +# split = [ MATCH ^/([^:]+:)/?(.*) : $(dir) ] ; +# device = $(split[1]) ; +# dir = $(split[2]) ; +# } +# +# # +# # Has file ? +# # +# # This is no exact science, just guess work: +# # +# # If the last part of the current path spec +# # includes some chars, followed by a dot, +# # optionally followed by more chars - +# # then it is a file (keep your fingers crossed). +# # +# split = [ regex.split $(dir) / ] ; +# local maybe_file = $(split[-1]) ; +# +# if [ MATCH ^([^.]+\\..*) : $(maybe_file) ] +# { +# file = $(maybe_file) ; +# dir = [ sequence.join $(split[1--2]) : / ] ; +# } +# +# # +# # Has dir spec ? +# # +# if $(dir) = "." +# { +# dir = "[]" ; +# } +# else if $(dir) +# { +# dir = [ regex.replace $(dir) \\.\\. - ] ; +# dir = [ regex.replace $(dir) / . ] ; +# +# if $(device) = "" +# { +# # +# # Relative directory +# # +# dir = "."$(dir) ; +# } +# dir = "["$(dir)"]" ; +# } +# +# native = [ sequence.join $(device) $(dir) $(file) ] ; +# +# return $(native) ; +# } +# +# +# rule __test__ ( ) { +# +# import assert ; +# import errors : try catch ; +# +# assert.true is-rooted "/" ; +# assert.true is-rooted "/foo" ; +# assert.true is-rooted "/foo/bar" ; +# assert.result : is-rooted "." ; +# assert.result : is-rooted "foo" ; +# assert.result : is-rooted "foo/bar" ; +# +# assert.true has-parent "foo" ; +# assert.true has-parent "foo/bar" ; +# assert.true has-parent "." ; +# assert.result : has-parent "/" ; +# +# assert.result "." : basename "." ; +# assert.result ".." : basename ".." ; +# assert.result "foo" : basename "foo" ; +# assert.result "foo" : basename "bar/foo" ; +# assert.result "foo" : basename "gaz/bar/foo" ; +# assert.result "foo" : basename "/gaz/bar/foo" ; +# +# assert.result "." : parent "foo" ; +# assert.result "/" : parent "/foo" ; +# assert.result "foo/bar" : parent "foo/bar/giz" ; +# assert.result ".." : parent "." ; +# assert.result ".." : parent "../foo" ; +# assert.result "../../foo" : parent "../../foo/bar" ; +# +# +# assert.result "." : reverse "." ; +# assert.result ".." : reverse "foo" ; +# assert.result "../../.." : reverse "foo/bar/giz" ; +# +# assert.result "foo" : join "foo" ; +# assert.result "/foo" : join "/" "foo" ; +# assert.result "foo/bar" : join "foo" "bar" ; +# assert.result "foo/bar" : join "foo/giz" "../bar" ; +# assert.result "foo/giz" : join "foo/bar/baz" "../../giz" ; +# assert.result ".." : join "." ".." ; +# assert.result ".." : join "foo" "../.." ; +# assert.result "../.." : join "../foo" "../.." ; +# assert.result "/foo" : join "/bar" "../foo" ; +# assert.result "foo/giz" : join "foo/giz" "." ; +# assert.result "." : join lib2 ".." ; +# assert.result "/" : join "/a" ".." ; +# +# assert.result /a/b : join /a/b/c .. ; +# +# assert.result "foo/bar/giz" : join "foo" "bar" "giz" ; +# assert.result "giz" : join "foo" ".." "giz" ; +# assert.result "foo/giz" : join "foo" "." "giz" ; +# +# try ; +# { +# join "a" "/b" ; +# } +# catch only first element may be rooted ; +# +# local CWD = "/home/ghost/build" ; +# assert.result : all-parents . : . : $(CWD) ; +# assert.result . .. ../.. ../../.. : all-parents "Jamfile" : "" : $(CWD) ; +# assert.result foo . .. ../.. ../../.. : all-parents "foo/Jamfile" : "" : $(CWD) ; +# assert.result ../Work .. ../.. ../../.. : all-parents "../Work/Jamfile" : "" : $(CWD) ; +# +# local CWD = "/home/ghost" ; +# assert.result . .. : all-parents "Jamfile" : "/home" : $(CWD) ; +# assert.result . : all-parents "Jamfile" : "/home/ghost" : $(CWD) ; +# +# assert.result "c/d" : relative "a/b/c/d" "a/b" ; +# assert.result "foo" : relative "foo" "." ; +# +# local save-os = [ modules.peek path : os ] ; +# modules.poke path : os : NT ; +# +# assert.result "foo/bar/giz" : make "foo/bar/giz" ; +# assert.result "foo/bar/giz" : make "foo\\bar\\giz" ; +# assert.result "foo" : make "foo/." ; +# assert.result "foo" : make "foo/bar/.." ; +# assert.result "/D:/My Documents" : make "D:\\My Documents" ; +# assert.result "/c:/boost/tools/build/new/project.jam" : make "c:\\boost\\tools\\build\\test\\..\\new\\project.jam" ; +# +# assert.result "foo\\bar\\giz" : native "foo/bar/giz" ; +# assert.result "foo" : native "foo" ; +# assert.result "D:\\My Documents\\Work" : native "/D:/My Documents/Work" ; +# +# modules.poke path : os : UNIX ; +# +# assert.result "foo/bar/giz" : make "foo/bar/giz" ; +# assert.result "/sub1" : make "/sub1/." ; +# assert.result "/sub1" : make "/sub1/sub2/.." ; +# assert.result "sub1" : make "sub1/." ; +# assert.result "sub1" : make "sub1/sub2/.." ; +# assert.result "/foo/bar" : native "/foo/bar" ; +# +# modules.poke path : os : VMS ; +# +# # +# # Don't really need to poke os before these +# # +# assert.result "disk:" "[dir]" "file" : split-path-VMS "disk:[dir]file" ; +# assert.result "disk:" "[dir]" "" : split-path-VMS "disk:[dir]" ; +# assert.result "disk:" "" "" : split-path-VMS "disk:" ; +# assert.result "disk:" "" "file" : split-path-VMS "disk:file" ; +# assert.result "" "[dir]" "file" : split-path-VMS "[dir]file" ; +# assert.result "" "[dir]" "" : split-path-VMS "[dir]" ; +# assert.result "" "" "file" : split-path-VMS "file" ; +# assert.result "" "" "" : split-path-VMS "" ; +# +# # +# # Special case: current directory +# # +# assert.result "" "[]" "" : split-path-VMS "[]" ; +# assert.result "disk:" "[]" "" : split-path-VMS "disk:[]" ; +# assert.result "" "[]" "file" : split-path-VMS "[]file" ; +# assert.result "disk:" "[]" "file" : split-path-VMS "disk:[]file" ; +# +# # +# # Make portable paths +# # +# assert.result "/disk:" : make "disk:" ; +# assert.result "foo/bar/giz" : make "[.foo.bar.giz]" ; +# assert.result "foo" : make "[.foo]" ; +# assert.result "foo" : make "[.foo.bar.-]" ; +# assert.result ".." : make "[.-]" ; +# assert.result ".." : make "[-]" ; +# assert.result "." : make "[]" ; +# assert.result "giz.h" : make "giz.h" ; +# assert.result "foo/bar/giz.h" : make "[.foo.bar]giz.h" ; +# assert.result "/disk:/my_docs" : make "disk:[my_docs]" ; +# assert.result "/disk:/boost/tools/build/new/project.jam" : make "disk:[boost.tools.build.test.-.new]project.jam" ; +# +# # +# # Special case (adds '.' to end of file w/o extension to +# # disambiguate from directory in portable path spec). +# # +# assert.result "Jamfile." : make "Jamfile" ; +# assert.result "dir/Jamfile." : make "[.dir]Jamfile" ; +# assert.result "/disk:/dir/Jamfile." : make "disk:[dir]Jamfile" ; +# +# # +# # Make native paths +# # +# assert.result "disk:" : native "/disk:" ; +# assert.result "[.foo.bar.giz]" : native "foo/bar/giz" ; +# assert.result "[.foo]" : native "foo" ; +# assert.result "[.-]" : native ".." ; +# assert.result "[.foo.-]" : native "foo/.." ; +# assert.result "[]" : native "." ; +# assert.result "disk:[my_docs.work]" : native "/disk:/my_docs/work" ; +# assert.result "giz.h" : native "giz.h" ; +# assert.result "disk:Jamfile." : native "/disk:Jamfile." ; +# assert.result "disk:[my_docs.work]Jamfile." : native "/disk:/my_docs/work/Jamfile." ; +# +# modules.poke path : os : $(save-os) ; +# +# } + +# + + +#def glob(dir, patterns): +# result = [] +# for pattern in patterns: +# result.extend(builtin_glob(os.path.join(dir, pattern))) +# return result + +def glob(dirs, patterns, exclude_patterns=None): + """Returns the list of files matching the given pattern in the + specified directory. Both directories and patterns are + supplied as portable paths. Each pattern should be non-absolute + path, and can't contain '.' or '..' elements. Each slash separated + element of pattern can contain the following special characters: + - '?', which match any character + - '*', which matches arbitrary number of characters. + A file $(d)/e1/e2/e3 (where 'd' is in $(dirs)) matches pattern p1/p2/p3 + if and only if e1 matches p1, e2 matches p2 and so on. + For example: + [ glob . : *.cpp ] + [ glob . : */build/Jamfile ] + """ + + assert(isinstance(patterns, list)) + assert(isinstance(dirs, list)) + + if not exclude_patterns: + exclude_patterns = [] + else: + assert(isinstance(exclude_patterns, list)) + + real_patterns = [os.path.join(d, p) for p in patterns for d in dirs] + real_exclude_patterns = [os.path.join(d, p) for p in exclude_patterns + for d in dirs] + + inc = [os.path.normpath(name) for p in real_patterns + for name in builtin_glob(p)] + exc = [os.path.normpath(name) for p in real_exclude_patterns + for name in builtin_glob(p)] + return [x for x in inc if x not in exc] + +def glob_tree(roots, patterns, exclude_patterns=None): + """Recursive version of GLOB. Builds the glob of files while + also searching in the subdirectories of the given roots. An + optional set of exclusion patterns will filter out the + matching entries from the result. The exclusions also apply + to the subdirectory scanning, such that directories that + match the exclusion patterns will not be searched.""" + + if not exclude_patterns: + exclude_patterns = [] + + result = glob(roots, patterns, exclude_patterns) + subdirs = [s for s in glob(roots, ["*"]) if s != "." and s != ".." and os.path.isdir(s)] + if subdirs: + result.extend(glob_tree(subdirs, patterns, exclude_patterns)) + + return result + +def glob_in_parents(dir, patterns, upper_limit=None): + """Recursive version of GLOB which glob sall parent directories + of dir until the first match is found. Returns an empty result if no match + is found""" + + assert(isinstance(dir, str)) + assert(isinstance(patterns, list)) + + result = [] + + absolute_dir = os.path.join(os.getcwd(), dir) + absolute_dir = os.path.normpath(absolute_dir) + while absolute_dir: + new_dir = os.path.split(absolute_dir)[0] + if new_dir == absolute_dir: + break + result = glob([new_dir], patterns) + if result: + break + absolute_dir = new_dir + + return result + + +# The relpath functionality is written by +# Cimarron Taylor +def split(p, rest=[]): + (h,t) = os.path.split(p) + if len(h) < 1: return [t]+rest + if len(t) < 1: return [h]+rest + return split(h,[t]+rest) + +def commonpath(l1, l2, common=[]): + if len(l1) < 1: return (common, l1, l2) + if len(l2) < 1: return (common, l1, l2) + if l1[0] != l2[0]: return (common, l1, l2) + return commonpath(l1[1:], l2[1:], common+[l1[0]]) + +def relpath(p1, p2): + (common,l1,l2) = commonpath(split(p1), split(p2)) + p = [] + if len(l1) > 0: + p = [ '../' * len(l1) ] + p = p + l2 + if p: + return os.path.join( *p ) + else: + return "." |