diff options
author | David Robillard <d@drobilla.net> | 2019-04-21 23:05:06 +0200 |
---|---|---|
committer | David Robillard <d@drobilla.net> | 2019-04-21 23:05:06 +0200 |
commit | 202873cf9bfdeb669365a6d76e54ad657c550fe2 (patch) | |
tree | 79db4dee998f1ef093c3e81f1d0ea708901ab96c /waflib/Tools/d_scan.py | |
parent | de6492738adf1794bf5fa39c1fe1ebbd167727ac (diff) | |
download | sratom-202873cf9bfdeb669365a6d76e54ad657c550fe2.tar.gz sratom-202873cf9bfdeb669365a6d76e54ad657c550fe2.tar.bz2 sratom-202873cf9bfdeb669365a6d76e54ad657c550fe2.zip |
Switch to using a submodule for autowaf
Diffstat (limited to 'waflib/Tools/d_scan.py')
m--------- | waflib | 0 | ||||
-rw-r--r-- | waflib/Tools/d_scan.py | 211 |
2 files changed, 0 insertions, 211 deletions
diff --git a/waflib b/waflib new file mode 160000 +Subproject 2314e236ca6e7d94a26c3c17091da0f25f5867f diff --git a/waflib/Tools/d_scan.py b/waflib/Tools/d_scan.py deleted file mode 100644 index 4e807a6..0000000 --- a/waflib/Tools/d_scan.py +++ /dev/null @@ -1,211 +0,0 @@ -#!/usr/bin/env python -# encoding: utf-8 -# Thomas Nagy, 2016-2018 (ita) - -""" -Provide a scanner for finding dependencies on d files -""" - -import re -from waflib import Utils - -def filter_comments(filename): - """ - :param filename: d file name - :type filename: string - :rtype: list - :return: a list of characters - """ - txt = Utils.readf(filename) - i = 0 - buf = [] - max = len(txt) - begin = 0 - while i < max: - c = txt[i] - if c == '"' or c == "'": # skip a string or character literal - buf.append(txt[begin:i]) - delim = c - i += 1 - while i < max: - c = txt[i] - if c == delim: - break - elif c == '\\': # skip the character following backslash - i += 1 - i += 1 - i += 1 - begin = i - elif c == '/': # try to replace a comment with whitespace - buf.append(txt[begin:i]) - i += 1 - if i == max: - break - c = txt[i] - if c == '+': # eat nesting /+ +/ comment - i += 1 - nesting = 1 - c = None - while i < max: - prev = c - c = txt[i] - if prev == '/' and c == '+': - nesting += 1 - c = None - elif prev == '+' and c == '/': - nesting -= 1 - if nesting == 0: - break - c = None - i += 1 - elif c == '*': # eat /* */ comment - i += 1 - c = None - while i < max: - prev = c - c = txt[i] - if prev == '*' and c == '/': - break - i += 1 - elif c == '/': # eat // comment - i += 1 - while i < max and txt[i] != '\n': - i += 1 - else: # no comment - begin = i - 1 - continue - i += 1 - begin = i - buf.append(' ') - else: - i += 1 - buf.append(txt[begin:]) - return buf - -class d_parser(object): - """ - Parser for d files - """ - def __init__(self, env, incpaths): - #self.code = '' - #self.module = '' - #self.imports = [] - - self.allnames = [] - - self.re_module = re.compile(r"module\s+([^;]+)") - self.re_import = re.compile(r"import\s+([^;]+)") - self.re_import_bindings = re.compile("([^:]+):(.*)") - self.re_import_alias = re.compile("[^=]+=(.+)") - - self.env = env - - self.nodes = [] - self.names = [] - - self.incpaths = incpaths - - def tryfind(self, filename): - """ - Search file a file matching an module/import directive - - :param filename: file to read - :type filename: string - """ - found = 0 - for n in self.incpaths: - found = n.find_resource(filename.replace('.', '/') + '.d') - if found: - self.nodes.append(found) - self.waiting.append(found) - break - if not found: - if not filename in self.names: - self.names.append(filename) - - def get_strings(self, code): - """ - :param code: d code to parse - :type code: string - :return: the modules that the code uses - :rtype: a list of match objects - """ - #self.imports = [] - self.module = '' - lst = [] - - # get the module name (if present) - - mod_name = self.re_module.search(code) - if mod_name: - self.module = re.sub(r'\s+', '', mod_name.group(1)) # strip all whitespaces - - # go through the code, have a look at all import occurrences - - # first, lets look at anything beginning with "import" and ending with ";" - import_iterator = self.re_import.finditer(code) - if import_iterator: - for import_match in import_iterator: - import_match_str = re.sub(r'\s+', '', import_match.group(1)) # strip all whitespaces - - # does this end with an import bindings declaration? - # (import bindings always terminate the list of imports) - bindings_match = self.re_import_bindings.match(import_match_str) - if bindings_match: - import_match_str = bindings_match.group(1) - # if so, extract the part before the ":" (since the module declaration(s) is/are located there) - - # split the matching string into a bunch of strings, separated by a comma - matches = import_match_str.split(',') - - for match in matches: - alias_match = self.re_import_alias.match(match) - if alias_match: - # is this an alias declaration? (alias = module name) if so, extract the module name - match = alias_match.group(1) - - lst.append(match) - return lst - - def start(self, node): - """ - The parsing starts here - - :param node: input file - :type node: :py:class:`waflib.Node.Node` - """ - self.waiting = [node] - # while the stack is not empty, add the dependencies - while self.waiting: - nd = self.waiting.pop(0) - self.iter(nd) - - def iter(self, node): - """ - Find all the modules that a file depends on, uses :py:meth:`waflib.Tools.d_scan.d_parser.tryfind` to process dependent files - - :param node: input file - :type node: :py:class:`waflib.Node.Node` - """ - path = node.abspath() # obtain the absolute path - code = "".join(filter_comments(path)) # read the file and filter the comments - names = self.get_strings(code) # obtain the import strings - for x in names: - # optimization - if x in self.allnames: - continue - self.allnames.append(x) - - # for each name, see if it is like a node or not - self.tryfind(x) - -def scan(self): - "look for .d/.di used by a d file" - env = self.env - gruik = d_parser(env, self.generator.includes_nodes) - node = self.inputs[0] - gruik.start(node) - nodes = gruik.nodes - names = gruik.names - return (nodes, names) - |