SCons.Scanner.Fortran

79 80 # cache the includes list in node so we only scan it once: 81 if node.includes is not None: 82 mods_and_includes = node.includes 83 else: 84 # retrieve all included filenames 85 includes = self.cre_incl.findall(node.get_text_contents()) 86 # retrieve all USE'd module names 87 modules = self.cre_use.findall(node.get_text_contents()) 88 # retrieve all defined module names 89 defmodules = self.cre_def.findall(node.get_text_contents()) 90 91 # Remove all USE'd module names that are defined in the same file 92 # (case-insensitively) 93 d = {} 94 for m in defmodules: 95 d[m.lower()] = 1 96 modules = [m for m in modules if m.lower() not in d] 97 98 # Convert module name to a .mod filename 99 suffix = env.subst('$FORTRANMODSUFFIX') 100 modules = [x.lower() + suffix for x in modules] 101 # Remove unique items from the list 102 mods_and_includes = SCons.Util.unique(includes+modules) 103 node.includes = mods_and_includes 104 105 # This is a hand-coded DSU (decorate-sort-undecorate, or 106 # Schwartzian transform) pattern. The sort key is the raw name 107 # of the file as specifed on the USE or INCLUDE line, which lets 108 # us keep the sort order constant regardless of whether the file 109 # is actually found in a Repository or locally. 110 nodes = [] 111 source_dir = node.get_dir() 112 if callable(path): 113 path = path() 114 for dep in mods_and_includes: 115 n, i = self.find_include(dep, source_dir, path) 116 117 if n is None: 118 SCons.Warnings.warn(SCons.Warnings.DependencyWarning, 119 "No dependency generated for file: %s (referenced by: %s) -- file not found" % (i, node)) 120 else: 121 sortkey = self.sort_key(dep) 122 nodes.append((sortkey, n)) 123 124 return [pair[1] for pair in sorted(nodes)]

127 """Return a prototype Scanner instance for scanning source files 128 for Fortran USE & INCLUDE statements""" 129 130 # The USE statement regex matches the following: 131 # 132 # USE module_name 133 # USE :: module_name 134 # USE, INTRINSIC :: module_name 135 # USE, NON_INTRINSIC :: module_name 136 # 137 # Limitations 138 # 139 # -- While the regex can handle multiple USE statements on one line, 140 # it cannot properly handle them if they are commented out. 141 # In either of the following cases: 142 # 143 # ! USE mod_a ; USE mod_b [entire line is commented out] 144 # USE mod_a ! ; USE mod_b [in-line comment of second USE statement] 145 # 146 # the second module name (mod_b) will be picked up as a dependency 147 # even though it should be ignored. The only way I can see 148 # to rectify this would be to modify the scanner to eliminate 149 # the call to re.findall, read in the contents of the file, 150 # treating the comment character as an end-of-line character 151 # in addition to the normal linefeed, loop over each line, 152 # weeding out the comments, and looking for the USE statements. 153 # One advantage to this is that the regex passed to the scanner 154 # would no longer need to match a semicolon. 155 # 156 # -- I question whether or not we need to detect dependencies to 157 # INTRINSIC modules because these are built-in to the compiler. 158 # If we consider them a dependency, will SCons look for them, not 159 # find them, and kill the build? Or will we there be standard 160 # compiler-specific directories we will need to point to so the 161 # compiler and SCons can locate the proper object and mod files? 162 163 # Here is a breakdown of the regex: 164 # 165 # (?i) : regex is case insensitive 166 # ^ : start of line 167 # (?: : group a collection of regex symbols without saving the match as a "group" 168 # ^|; : matches either the start of the line or a semicolon - semicolon 169 # ) : end the unsaved grouping 170 # \s* : any amount of white space 171 # USE : match the string USE, case insensitive 172 # (?: : group a collection of regex symbols without saving the match as a "group" 173 # \s+| : match one or more whitespace OR .... (the next entire grouped set of regex symbols) 174 # (?: : group a collection of regex symbols without saving the match as a "group" 175 # (?: : establish another unsaved grouping of regex symbols 176 # \s* : any amount of white space 177 # , : match a comma 178 # \s* : any amount of white space 179 # (?:NON_)? : optionally match the prefix NON_, case insensitive 180 # INTRINSIC : match the string INTRINSIC, case insensitive 181 # )? : optionally match the ", INTRINSIC/NON_INTRINSIC" grouped expression 182 # \s* : any amount of white space 183 # :: : match a double colon that must appear after the INTRINSIC/NON_INTRINSIC attribute 184 # ) : end the unsaved grouping 185 # ) : end the unsaved grouping 186 # \s* : match any amount of white space 187 # (\w+) : match the module name that is being USE'd 188 # 189 # 190 use_regex = "(?i)(?:^|;)\s*USE(?:\s+|(?:(?:\s*,\s*(?:NON_)?INTRINSIC)?\s*::))\s*(\w+)" 191 192 193 # The INCLUDE statement regex matches the following: 194 # 195 # INCLUDE 'some_Text' 196 # INCLUDE "some_Text" 197 # INCLUDE "some_Text" ; INCLUDE "some_Text" 198 # INCLUDE kind_"some_Text" 199 # INCLUDE kind_'some_Text" 200 # 201 # where some_Text can include any alphanumeric and/or special character 202 # as defined by the Fortran 2003 standard. 203 # 204 # Limitations: 205 # 206 # -- The Fortran standard dictates that a " or ' in the INCLUDE'd 207 # string must be represented as a "" or '', if the quotes that wrap 208 # the entire string are either a ' or ", respectively. While the 209 # regular expression below can detect the ' or " characters just fine, 210 # the scanning logic, presently is unable to detect them and reduce 211 # them to a single instance. This probably isn't an issue since, 212 # in practice, ' or " are not generally used in filenames. 213 # 214 # -- This regex will not properly deal with multiple INCLUDE statements 215 # when the entire line has been commented out, ala 216 # 217 # ! INCLUDE 'some_file' ; INCLUDE 'some_file' 218 # 219 # In such cases, it will properly ignore the first INCLUDE file, 220 # but will actually still pick up the second. Interestingly enough, 221 # the regex will properly deal with these cases: 222 # 223 # INCLUDE 'some_file' 224 # INCLUDE 'some_file' !; INCLUDE 'some_file' 225 # 226 # To get around the above limitation, the FORTRAN programmer could 227 # simply comment each INCLUDE statement separately, like this 228 # 229 # ! INCLUDE 'some_file' !; INCLUDE 'some_file' 230 # 231 # The way I see it, the only way to get around this limitation would 232 # be to modify the scanning logic to replace the calls to re.findall 233 # with a custom loop that processes each line separately, throwing 234 # away fully commented out lines before attempting to match against 235 # the INCLUDE syntax. 236 # 237 # Here is a breakdown of the regex: 238 # 239 # (?i) : regex is case insensitive 240 # (?: : begin a non-saving group that matches the following: 241 # ^ : either the start of the line 242 # | : or 243 # ['">]\s*; : a semicolon that follows a single quote, 244 # double quote or greater than symbol (with any 245 # amount of whitespace in between). This will 246 # allow the regex to match multiple INCLUDE 247 # statements per line (although it also requires 248 # the positive lookahead assertion that is 249 # used below). It will even properly deal with 250 # (i.e. ignore) cases in which the additional 251 # INCLUDES are part of an in-line comment, ala 252 # " INCLUDE 'someFile' ! ; INCLUDE 'someFile2' " 253 # ) : end of non-saving group 254 # \s* : any amount of white space 255 # INCLUDE : match the string INCLUDE, case insensitive 256 # \s+ : match one or more white space characters 257 # (?\w+_)? : match the optional "kind-param _" prefix allowed by the standard 258 # [<"'] : match the include delimiter - an apostrophe, double quote, or less than symbol 259 # (.+?) : match one or more characters that make up 260 # the included path and file name and save it 261 # in a group. The Fortran standard allows for 262 # any non-control character to be used. The dot 263 # operator will pick up any character, including 264 # control codes, but I can't conceive of anyone 265 # putting control codes in their file names. 266 # The question mark indicates it is non-greedy so 267 # that regex will match only up to the next quote, 268 # double quote, or greater than symbol 269 # (?=["'>]) : positive lookahead assertion to match the include 270 # delimiter - an apostrophe, double quote, or 271 # greater than symbol. This level of complexity 272 # is required so that the include delimiter is 273 # not consumed by the match, thus allowing the 274 # sub-regex discussed above to uniquely match a 275 # set of semicolon-separated INCLUDE statements 276 # (as allowed by the F2003 standard) 277 278 include_regex = """(?i)(?:^|['">]\s*;)\s*INCLUDE\s+(?:\w+_)?[<"'](.+?)(?=["'>])""" 279 280 # The MODULE statement regex finds module definitions by matching 281 # the following: 282 # 283 # MODULE module_name 284 # 285 # but *not* the following: 286 # 287 # MODULE PROCEDURE procedure_name 288 # 289 # Here is a breakdown of the regex: 290 # 291 # (?i) : regex is case insensitive 292 # ^\s* : any amount of white space 293 # MODULE : match the string MODULE, case insensitive 294 # \s+ : match one or more white space characters 295 # (?!PROCEDURE) : but *don't* match if the next word matches 296 # PROCEDURE (negative lookahead assertion), 297 # case insensitive 298 # (\w+) : match one or more alphanumeric characters 299 # that make up the defined module name and 300 # save it in a group 301 302 def_regex = """(?i)^\s*MODULE\s+(?!PROCEDURE)(\w+)""" 303 304 scanner = F90Scanner("FortranScan", 305 "$FORTRANSUFFIXES", 306 path_variable, 307 use_regex, 308 include_regex, 309 def_regex) 310 return scanner

Source Code for Module SCons.Scanner.Fortran