def bracket_strings(start, end, b_brack, e_brack):
sep = "tzvlw"
t = basicutils.CompileTextFromRange(start, end, sep)
tokens = [tk.lower() for tk in t.split(sep)]
b = []
for tk in tokens:
tk = tk.strip()
if tk.startswith(b_brack):
b_contents = tk[1:tk.find(e_brack)]
#Hack to get rid of [-],[+],[*] - could also try to remove non alpha
if (len(b_contents) > 3):
#Hack for debug prints that started with [0x%x]
if (b_contents != "0x%x"):
b.append(tk[1:tk.find(e_brack)])
print "bracket_strings tokens:"
print tokens
print b
u_gram = ""
u_gram_score = 0
if (len(b) > 0):
f = nltk.FreqDist(b)
u_gram = f.most_common(1)[0][0]
u_gram_score = f.most_common(1)[0][1]
return (u_gram, u_gram_score)
def string_range_tokenize(start, end, sep):
# get all string references in this range concatenated into a single string
t = basicutils.CompileTextFromRange(start, end, sep)
#Enable this if you already have a bunch of function names and want to include that in the mix
#t+= basicutils.CompileFuncNamesFromRangeAsText(start,end,sep)
#print "string_range_tokenize: raw text:"
#print t
#remove printf/sprintf format strings
tc = re.sub("%[0-9A-Za-z]+", " ", t)
#convert dash to underscore
tc = re.sub("-", "_", tc)
#replace _ and / with space - may want to turn this off sometimes
#this will break up snake case and paths
#problem is that if you have a path that is used throughout the binary it will probably dominate results
tc = re.sub("_", " ", tc)
#replace / and \\ with a space
tc = re.sub("[/\\\\]", " ", tc)
#remove anything except alphanumeric, spaces, . (for .c, .cpp, etc) and _
tc = re.sub("[^A-Za-z0-9_\.\s]", " ", tc)
#lowercase it - and store this as the original set of tokens to work with
tokens = [tk.lower() for tk in tc.split()]
#remove English stop words
#this is the list from the MIT *bow project
eng_stopw = {
"about", "all", "am", "an", "and", "are", "as", "at", "be", "been",
"but", "by", "can", "cannot", "did", "do", "does", "doing", "done",
"for", "from", "had", "has", "have", "having", "if", "in", "is", "it",
"its", "of", "on", "that", "the", "these", "they", "this", "those",
"to", "too", "want", "wants", "was", "what", "which", "will", "with",
"would"
}
#remove "code" stop words
#e.g. common words in debugging strings
code_sw = {
"error", "err", "errlog", "log", "return", "returned", "byte", "bytes",
"status", "len", "length", "size", "ok", "0x", "warning", "fail",
"failed", "failure", "invalid", "illegal", "param", "parameter",
"done", "complete", "assert", "assertion", "cant", "didnt", "class",
"foundation", "cdecl", "stdcall", "thiscall"
}
stopw = eng_stopw.union(code_sw)
c = 0
tokens_f = []
for t in tokens:
if t not in stopw:
tokens_f.append(t)
return tokens_f
def source_file_strings(start, end):
sep = "tzvlw"
t = basicutils.CompileTextFromRange(start, end, sep)
#normally would do lower here to normalize but we lose camel case that way
tokens = [tk for tk in t.split(sep)]
#for each string, remove quotes and commas, then tokenize based on spaces to generate the final list
tokens2 = []
for tk in tokens:
tk = tk.strip()
#strip punctuation, need to leave in _ for filenames and / and \ for paths
tk = re.sub("[\"\'\,]", " ", tk)
for tk2 in tk.split(" "):
tokens2.append(tk2)
b = []
for tk in tokens2:
tk = tk.strip()
if tk.endswith(".c") or tk.endswith(".cpp") or tk.endswith(".cc"):
#If there's a dir path, only use the end filename
#This could be tweaked if the directory structure is part of the software architecture
#e.g. if there are multiple source directories with meaningful names
if tk.rfind("/") != -1:
ntk = tk[tk.rfind("/") + 1:]
elif tk.rfind("\\") != -1:
ntk = tk[tk.rfind("\\") + 1:]
else:
ntk = tk
b.append(ntk)
print "source_file_strings tokens:"
#print tokens
print b
#a better way to do this (if there are multiple)
#would be to sort, uniquify, and then make the name foo.c_and_bar.c
u_gram = ""
u_gram_score = 0
if (len(b) > 0):
f = nltk.FreqDist(b)
u_gram = f.most_common(1)[0][0]
u_gram_score = f.most_common(1)[0][1]
return (u_gram, u_gram_score)
