scantools.py

reduced python source for embedded apps

"""\

Tools for scanning header files in search of function prototypes.

Often, the function prototypes in header files contain enough information
to automatically generate (or reverse-engineer) interface specifications
from them.  The conventions used are very vendor specific, but once you've
figured out what they are they are often a great help, and it sure beats
manually entering the interface specifications.  (These are needed to generate
the glue used to access the functions from Python.)

In order to make this class useful, almost every component can be overridden.
The defaults are (currently) tuned to scanning Apple Macintosh header files,
although most Mac specific details are contained in header-specific subclasses.
"""

import regex
import regsub
import string
import sys
import os
import fnmatch
from types import *
try:
	import MacOS
except ImportError:
	MacOS = None

from bgenlocations import CREATOR, INCLUDEDIR

Error = "scantools.Error"

class Scanner:

	def __init__(self, input = None, output = None, defsoutput = None):
		self.initsilent()
		self.initblacklists()
		self.initrepairinstructions()
		self.initpaths()
		self.initfiles()
		self.initpatterns()
		self.compilepatterns()
		self.initosspecifics()
		self.initusedtypes()
		if output:
			self.setoutput(output, defsoutput)
		if input:
			self.setinput(input)
	
	def initusedtypes(self):
		self.usedtypes = {}
	
	def typeused(self, type, mode):
		if not self.usedtypes.has_key(type):
			self.usedtypes[type] = {}
		self.usedtypes[type][mode] = None
	
	def reportusedtypes(self):
		types = self.usedtypes.keys()
		types.sort()
		for type in types:
			modes = self.usedtypes[type].keys()
			modes.sort()
			print type, string.join(modes)
			
	def gentypetest(self, file):
		fp = open(file, "w")
		fp.write("types=[\n")
		types = self.usedtypes.keys()
		types.sort()
		for type in types:
			fp.write("\t'%s',\n"%type)
		fp.write("]\n")
		fp.write("""missing=0
for t in types:
	try:
		tt = eval(t)
	except NameError:
		print "** Missing type:", t
		missing = 1
if missing: raise "Missing Types"
""")
		fp.close()

	def initsilent(self):
		self.silent = 0

	def error(self, format, *args):
		if self.silent >= 0:
			print format%args

	def report(self, format, *args):
		if not self.silent:
			print format%args

	def writeinitialdefs(self):
		pass
		
	def initblacklists(self):
		self.blacklistnames = self.makeblacklistnames()
		self.blacklisttypes = ["unknown", "-"] + self.makeblacklisttypes()

	def makeblacklistnames(self):
		return []

	def makeblacklisttypes(self):
		return []

	def initrepairinstructions(self):
		self.repairinstructions = self.makerepairinstructions()

	def makerepairinstructions(self):
		"""Parse the repair file into repair instructions.
		
		The file format is simple:
		1) use \ to split a long logical line in multiple physical lines
		2) everything after the first # on a line is ignored (as comment)
		3) empty lines are ignored
		4) remaining lines must have exactly 3 colon-separated fields:
		   functionpattern : argumentspattern : argumentsreplacement
		5) all patterns use shell style pattern matching
		6) an empty functionpattern means the same as *
		7) the other two fields are each comma-separated lists of triples
		8) a triple is a space-separated list of 1-3 words
		9) a triple with less than 3 words is padded at the end with "*" words
		10) when used as a pattern, a triple matches the type, name, and mode
		    of an argument, respectively
		11) when used as a replacement, the words of a triple specify
		    replacements for the corresponding words of the argument,
		    with "*" as a word by itself meaning leave the original word
		    (no other uses of "*" is allowed)
		12) the replacement need not have the same number of triples
		    as the pattern
		"""
		f = self.openrepairfile()
		if not f: return []
		print "Reading repair file", `f.name`, "..."
		list = []
		lineno = 0
		while 1:
			line = f.readline()
			if not line: break
			lineno = lineno + 1
			startlineno = lineno
			while line[-2:] == '\\\n':
				line = line[:-2] + ' ' + f.readline()
				lineno = lineno + 1
			i = string.find(line, '#')
			if i >= 0: line = line[:i]
			words = map(string.strip, string.splitfields(line, ':'))
			if words == ['']: continue
			if len(words) <> 3:
				print "Line", startlineno,
				print ": bad line (not 3 colon-separated fields)"
				print `line`
				continue
			[fpat, pat, rep] = words
			if not fpat: fpat = "*"
			if not pat:
				print "Line", startlineno,
				print "Empty pattern"
				print `line`
				continue
			patparts = map(string.strip, string.splitfields(pat, ','))
			repparts = map(string.strip, string.splitfields(rep, ','))
			patterns = []
			for p in patparts:
				if not p:
					print "Line", startlineno,
					print "Empty pattern part"
					print `line`
					continue
				pattern = string.split(p)
				if len(pattern) > 3:
					print "Line", startlineno,
					print "Pattern part has > 3 words"
					print `line`
					pattern = pattern[:3]
				else:
					while len(pattern) < 3:
						pattern.append("*")
				patterns.append(pattern)
			replacements = []
			for p in repparts:
				if not p:
					print "Line", startlineno,
					print "Empty replacement part"
					print `line`
					continue
				replacement = string.split(p)
				if len(replacement) > 3:
					print "Line", startlineno,
					print "Pattern part has > 3 words"
					print `line`
					replacement = replacement[:3]
				else:
					while len(replacement) < 3:
						replacement.append("*")
				replacements.append(replacement)
			list.append((fpat, patterns, replacements))
		return list
	
	def openrepairfile(self, filename = "REPAIR"):
		try:
			return open(filename, "r")
		except IOError, msg:
			print `filename`, ":", msg
			print "Cannot open repair file -- assume no repair needed"
			return None

	def initfiles(self):
		self.specmine = 0
		self.defsmine = 0
		self.scanmine = 0
		self.specfile = sys.stdout
		self.defsfile = None
		self.scanfile = sys.stdin
		self.lineno = 0
		self.line = ""

	def initpaths(self):
		self.includepath = [':', INCLUDEDIR]

	def initpatterns(self):
#		self.head_pat = "^extern pascal[ \t]+" # XXX Mac specific!
		self.head_pat = "^EXTERN_API[^_]"
		self.tail_pat = "[;={}]"
#		self.type_pat = "pascal[ \t\n]+\(<type>[a-zA-Z0-9_ \t]*[a-zA-Z0-9_]\)[ \t\n]+"
		self.type_pat = "EXTERN_API" + \
						"[ \t\n]*([ \t\n]*" + \
						"\(<type>[a-zA-Z0-9_* \t]*[a-zA-Z0-9_*]\)" + \
						"[ \t\n]*)[ \t\n]*"
		self.name_pat = "\(<name>[a-zA-Z0-9_]+\)[ \t\n]*"
		self.args_pat = "(\(<args>\([^(;=)]+\|([^(;=)]*)\)*\))"
		self.whole_pat = self.type_pat + self.name_pat + self.args_pat
#		self.sym_pat = "^[ \t]*\(<name>[a-zA-Z0-9_]+\)[ \t]*=" + \
#		               "[ \t]*\(<defn>[-0-9'\"(][^\t\n,;}]*\),?"
		self.sym_pat = "^[ \t]*\(<name>[a-zA-Z0-9_]+\)[ \t]*=" + \
		               "[ \t]*\(<defn>[-0-9_a-zA-Z'\"(][^\t\n,;}]*\),?"
		self.asplit_pat = "^\(<type>.*[^a-zA-Z0-9_]\)\(<name>[a-zA-Z0-9_]+\)$"
		self.comment1_pat = "\(<rest>.*\)//.*"
		# note that the next pattern only removes comments that are wholly within one line
		self.comment2_pat = "\(<rest>.*\)/\*.*\*/"

	def compilepatterns(self):
		for name in dir(self):
			if name[-4:] == "_pat":
				pat = getattr(self, name)
				prog = regex.symcomp(pat)
				setattr(self, name[:-4], prog)

	def initosspecifics(self):
		if MacOS:
			self.filetype = 'TEXT'
			self.filecreator = CREATOR
		else:
			self.filetype = self.filecreator = None

	def setfiletype(self, filename):
		if MacOS and (self.filecreator or self.filetype):
			creator, type = MacOS.GetCreatorAndType(filename)
			if self.filecreator: creator = self.filecreator
			if self.filetype: type = self.filetype
			MacOS.SetCreatorAndType(filename, creator, type)

	def close(self):
		self.closefiles()

	def closefiles(self):
		self.closespec()
		self.closedefs()
		self.closescan()

	def closespec(self):
		tmp = self.specmine and self.specfile
		self.specfile = None
		if tmp: tmp.close()

	def closedefs(self):
		tmp = self.defsmine and self.defsfile
		self.defsfile = None
		if tmp: tmp.close()

	def closescan(self):