def Scan(self, path, matcher):
"""Scans given file searching for occurrences of given pattern.
Args:
path: A path to the file that needs to be searched.
matcher: A matcher object specifying a pattern to search for.
Yields:
`BufferReference` objects pointing to file parts with matching content.
"""
streamer = streaming.FileStreamer(chunk_size=self.CHUNK_SIZE,
overlap_size=self.OVERLAP_SIZE)
offset = self.params.start_offset
amount = self.params.length
for chunk in streamer.StreamFilePath(path,
offset=offset,
amount=amount):
for span in chunk.Scan(matcher):
ctx_begin = max(span.begin - self.params.bytes_before, 0)
ctx_end = min(span.end + self.params.bytes_after,
len(chunk.data))
ctx_data = chunk.data[ctx_begin:ctx_end]
yield rdf_client.BufferReference(offset=chunk.offset +
ctx_begin,
length=len(ctx_data),
data=ctx_data)
if self.params.mode == self.params.Mode.FIRST_HIT:
return
def testSingleChunk(self):
with open(self.temp_filepath, "wb") as fd:
fd.write("abcdef")
streamer = streaming.FileStreamer(chunk_size=8, overlap_size=2)
chunks = list(streamer.StreamFilePath(self.temp_filepath, amount=7))
self.assertEqual(len(chunks), 1)
self.assertEqual(chunks[0].data, "abcdef")
self.assertEqual(chunks[0].offset, 0)
self.assertEqual(chunks[0].overlap, 0)
def testSmallAmount(self):
with open(self.temp_filepath, "wb") as fd:
fd.write("abc")
streamer = streaming.FileStreamer(chunk_size=1, overlap_size=0)
chunks = list(streamer.StreamFilePath(self.temp_filepath, amount=2))
self.assertEqual(len(chunks), 2)
self.assertEqual(chunks[0].data, "a")
self.assertEqual(chunks[1].data, "b")
self.assertEqual(chunks[0].offset, 0)
self.assertEqual(chunks[1].offset, 1)
self.assertEqual(chunks[0].overlap, 0)
self.assertEqual(chunks[1].overlap, 0)
def testOneByteOverlap(self):
with open(self.temp_filepath, "wb") as fd:
fd.write("abcdef")
streamer = streaming.FileStreamer(chunk_size=3, overlap_size=1)
chunks = list(streamer.StreamFilePath(self.temp_filepath, amount=8))
self.assertEqual(len(chunks), 3)
self.assertEqual(chunks[0].data, "abc")
self.assertEqual(chunks[1].data, "cde")
self.assertEqual(chunks[2].data, "ef")
self.assertEqual(chunks[0].offset, 0)
self.assertEqual(chunks[1].offset, 2)
self.assertEqual(chunks[2].offset, 4)
self.assertEqual(chunks[0].overlap, 0)
self.assertEqual(chunks[1].overlap, 1)
self.assertEqual(chunks[2].overlap, 1)
def testShorterOverlap(self):
with open(self.temp_filepath, "wb") as fd:
fd.write("abcdefg")
streamer = streaming.FileStreamer(chunk_size=4, overlap_size=2)
chunks = list(streamer.StreamFilePath(self.temp_filepath, amount=1024))
self.assertEqual(len(chunks), 3)
self.assertEqual(chunks[0].data, "abcd")
self.assertEqual(chunks[1].data, "cdef")
self.assertEqual(chunks[2].data, "efg")
self.assertEqual(chunks[0].offset, 0)
self.assertEqual(chunks[1].offset, 2)
self.assertEqual(chunks[2].offset, 4)
self.assertEqual(chunks[0].overlap, 0)
self.assertEqual(chunks[1].overlap, 2)
self.assertEqual(chunks[2].overlap, 2)
def testOffset(self):
with open(self.temp_filepath, "wb") as fd:
fd.write("abcdefghi")
streamer = streaming.FileStreamer(chunk_size=3, overlap_size=2)
stream = streamer.StreamFilePath(self.temp_filepath,
offset=4,
amount=108)
chunks = list(stream)
self.assertEqual(len(chunks), 3)
self.assertEqual(chunks[0].data, "efg")
self.assertEqual(chunks[1].data, "fgh")
self.assertEqual(chunks[2].data, "ghi")
self.assertEqual(chunks[0].offset, 4)
self.assertEqual(chunks[1].offset, 5)
self.assertEqual(chunks[2].offset, 6)
self.assertEqual(chunks[0].overlap, 0)
self.assertEqual(chunks[1].overlap, 2)
self.assertEqual(chunks[2].overlap, 2)
def testNoData(self):
streamer = streaming.FileStreamer(chunk_size=3, overlap_size=1)
chunks = list(streamer.StreamFilePath(self.temp_filepath, amount=5))
self.assertEqual(len(chunks), 0)
