def _parse_file(self, contents: str, root: bblfsh.Node, path: str) -> \
Tuple[List[VirtualNode], Dict[int, bblfsh.Node]]:
"""
Parse a file into a sequence of `VirtuaNode`-s and a mapping from VirtualNode to parent.
Given the source text and the corresponding UAST this function compiles the list of
`VirtualNode`-s and the parents mapping. That list of nodes equals to the original
source text bit-to-bit after `"".join(n.value for n in nodes)`. `parents` map from
`id(node)` to its parent `bblfsh.Node`.
:param contents: source file text
:param root: UAST root node
:param path: path to the file, used for debugging
:return: list of `VirtualNode`-s and the parents.
"""
# build the line mapping
lines = contents.splitlines(keepends=True)
# Check if there is a newline in the end of file. Yes, you can just check
# lines[-1][-1] == "\n" but if someone decide to use weird '\u2028' unicode character for
# new line this condition gives wrong result.
eof_new_line = lines[-1].splitlines()[0] != lines[-1]
if eof_new_line:
# We add last line as empty one because it actually exists, but .splitlines() does not
# return it.
lines.append("")
line_offsets = numpy.zeros(len(lines) + 1, dtype=numpy.int32)
pos = 0
for i, line in enumerate(lines):
line_offsets[i] = pos
pos += len(line)
line_offsets[-1] = pos + 1
# walk the tree: collect nodes with assigned tokens and build the parents map
node_tokens = []
parents = {}
queue = [root]
while queue:
node = queue.pop()
if node.internal_type in self.node_fixtures:
node = self.node_fixtures[node.internal_type](node)
for child in node.children:
parents[id(child)] = node
queue.extend(node.children)
if (node.token or node.start_position and node.end_position
and node.start_position != node.end_position and not node.children):
node_tokens.append(node)
node_tokens.sort(key=lambda n: n.start_position.offset)
sentinel = bblfsh.Node()
sentinel.start_position.offset = len(contents)
sentinel.start_position.line = len(lines)
node_tokens.append(sentinel)
# scan `node_tokens` and fill the gaps with imaginary nodes
result = []
pos = 0
parser = self.tokens.PARSER
searchsorted = numpy.searchsorted
for node in node_tokens:
if node.start_position.offset < pos:
continue
if node.start_position.offset > pos:
sumlen = 0
diff = contents[pos:node.start_position.offset]
for match in parser.finditer(diff):
positions = []
for suboff in (match.start(), match.end()):
offset = pos + suboff
line = searchsorted(line_offsets, offset, side="right")
col = offset - line_offsets[line - 1] + 1
positions.append(Position(offset, line, col))
token = match.group()
sumlen += len(token)
result.append(VirtualNode(token, *positions, path=path))
assert sumlen == node.start_position.offset - pos, \
"missed some imaginary tokens: \"%s\"" % diff
if node is sentinel:
break
result.extend(VirtualNode.from_node(node, contents, path, self.token_unwrappers))
pos = node.end_position.offset
return result, parents
def _parse_file(self, file: AnnotationManager) -> None:
"""
Annotate source code with `RawTokenAnnotation`-s.
Given the source text and the corresponding UAST this function covers all code with a
`RawTokenAnnotation`-s.
:param file: Source code annotated with `UASTAnnotation`.
"""
# TODO(zurk): rename this function when the refactoring is finished.
contents = file.sequence
# build the line mapping
lines = contents.splitlines(keepends=True)
# Check if there is a newline in the end of file. Yes, you can just check
# lines[-1][-1] == "\n" but if someone decide to use weird '\u2028' unicode character for
# new line this condition gives wrong result.
eof_new_line = lines[-1].splitlines()[0] != lines[-1]
if eof_new_line:
# We add last line as empty one because it actually exists, but .splitlines() does not
# return it.
lines.append("")
line_offsets = numpy.zeros(len(lines) + 1, dtype=numpy.int32)
pos = 0
for i, line in enumerate(lines):
line_offsets[i] = pos
pos += len(line)
line_offsets[-1] = pos + 1
# walk the tree: collect nodes with assigned tokens
node_tokens = []
queue = [file.get(UASTAnnotation).uast]
while queue:
node = queue.pop()
if node.internal_type in self.node_fixtures:
self.node_fixtures[node.internal_type](node)
queue.extend(node.children)
if (node.token or node.start_position and node.end_position
and node.start_position != node.end_position
and not node.children):
node_tokens.append(node)
node_tokens.sort(key=lambda n: n.start_position.offset)
sentinel = bblfsh.Node()
sentinel.start_position.offset = len(contents)
sentinel.start_position.line = len(lines)
node_tokens.append(sentinel)
# scan `node_tokens` and fill the gaps with imaginary nodes
pos = 0
parser = self.tokens.PARSER
for node in node_tokens:
if node.start_position.offset < pos:
continue
if node.start_position.offset > pos:
sumlen = 0
diff = contents[pos:node.start_position.offset]
for match in parser.finditer(diff):
offsets = []
for suboff in (match.start(), match.end()):
offsets.append(pos + suboff)
token = match.group()
sumlen += len(token)
file.add(RawTokenAnnotation(*offsets))
assert sumlen == node.start_position.offset - pos, \
"missed some imaginary tokens: \"%s\"" % diff
if node is sentinel:
break
uast_node_annot = list(
VirtualNode.from_node(node, contents, self.token_unwrappers))
file.add(*uast_node_annot)
pos = node.end_position.offset
def analyze_uast(path: str, root: bblfsh.Node, roles: set, reserved: set):
contents = Path(path).read_text()
# walk the tree: collect nodes with assigned tokens and build the parents map
node_tokens = []
parents = {}
queue = [root]
while queue:
node = queue.pop()
for child in node.children:
parents[id(child)] = node
queue.extend(node.children)
if node.token or node.start_position and node.end_position and not node.children:
node_tokens.append(node)
node_tokens.sort(key=lambda n: n.start_position.offset)
sentinel = bblfsh.Node()
sentinel.start_position.offset = len(contents)
sentinel.start_position.line = contents.count("\n")
node_tokens.append(sentinel)
# scan `node_tokens` and analyze the gaps and the token prefixes and suffixes
pos = 0
ws = re.compile("\s+")
alpha = re.compile("[a-zA-Z]+")
IDENTIFIER = bblfsh.role_id("IDENTIFIER")
log = logging.getLogger("analyze_uast")
def ccheck(char: str) -> bool:
return not char.isspace() and not char.isalnum(
) and not ord(char) >= 128
for node in node_tokens:
token = node.token if node.token else \
contents[node.start_position.offset:node.end_position.offset]
if node.start_position.offset > pos:
diff = contents[pos:node.start_position.offset]
parts = ws.split(diff)
for part in parts:
if len(part) >= 8:
continue
# for keyword in alpha.finditer(part):
# reserved.add(keyword.group())
for nonalpha in alpha.split(part):
for char in nonalpha:
if ccheck(char):
reserved.add(char)
if node is sentinel:
break
pos = node.end_position.offset
if IDENTIFIER not in node.roles:
continue
outer = contents[node.start_position.offset:node.end_position.offset]
if outer == token:
continue
pos = outer.find(token)
if pos < 0:
log.warning(
"skipped %s, token offset corruption \"%s\" vs. \"%s\"", path,
token, outer)
break
if pos > 0:
for char in outer[:pos]:
if ccheck(char):
reserved.add(char)
if pos + len(token) < len(outer):
for char in outer[pos + len(token):]:
if ccheck(char):
reserved.add(char)
def tokenize_code(self, path: str, contents: str = None, root: bblfsh.Node = None) -> \
Tuple[List[VirtualNode], Dict[int, bblfsh.Node]]:
"""
Parse a file into a sequence of `VirtuaNode`-s and a mapping from VirtualNode to parent.
Given the source text and the corresponding UAST this function compiles the list of
`VirtualNode`-s and the parents mapping. That list of nodes equals to the original
source text bit-to-bit after `"".join(n.value for n in nodes)`. `parents` map from
`id(node)` to its parent `bblfsh.Node`.
:param path: path to the file.
:param contents: source file text, if not provided - path will be used to read content.
:param root: UAST root node. If None - the file will be parsed using bblfsh client.
:return: list of `VirtualNode`-s, the parents and root.
"""
if contents is None:
contents = self.client._get_contents(contents=contents, filename=path)
if root is None:
root = self.client.parse(filename=path, contents=contents).uast
# build the line mapping
contents = contents.decode()
lines = contents.split("\n")
line_offsets = numpy.zeros(len(lines) + 1, dtype=numpy.int32)
pos = 0
for i, line in enumerate(lines):
line_offsets[i] = pos
pos += len(line) + 1
line_offsets[-1] = pos
# walk the tree: collect nodes with assigned tokens and build the parents map
node_tokens = []
parents = {}
queue = [root]
while queue:
node = queue.pop()
if node.internal_type in self.node_fixtures:
node = self.node_fixtures[node.internal_type](node)
for child in node.children:
parents[id(child)] = node
queue.extend(node.children)
if (node.token or node.start_position and node.end_position
and node.start_position != node.end_position and not node.children):
node_tokens.append(node)
node_tokens.sort(key=lambda n: n.start_position.offset)
sentinel = bblfsh.Node()
sentinel.start_position.offset = len(contents)
sentinel.start_position.line = len(lines)
node_tokens.append(sentinel)
# scan `node_tokens` and fill the gaps with imaginary nodes
result = []
pos = 0
parser = self.tokens.PARSER
searchsorted = numpy.searchsorted
for node in node_tokens:
if node.start_position.offset < pos:
continue
if node.start_position.offset > pos:
sumlen = 0
diff = contents[pos:node.start_position.offset]
for match in parser.finditer(diff):
positions = []
for suboff in (match.start(), match.end()):
offset = pos + suboff
line = searchsorted(line_offsets, offset, side="right")
col = offset - line_offsets[line - 1] + 1
positions.append(Position(offset, line, col))
token = match.group()
sumlen += len(token)
result.append(VirtualNode(token, *positions, path=path))
assert sumlen == node.start_position.offset - pos, \
"missed some imaginary tokens: \"%s\"" % diff
if node is sentinel:
break
result.extend(VirtualNode.from_node(node, contents, path, self.token_unwrappers))
pos = node.end_position.offset
return result, parents, root
def analyze_uast(path: str, content: str, root: bblfsh.Node,
internal_types: dict, roles: dict, reserved: set):
"""
Fill internal types, roles and reserved dictionaries with statistics computed from an UAST.
:param path: Path of the analyzed file.
:param content: Content of the analyzed file.
:param root: UAST of the analyzed file.
:param internal_types: Dictionary containing the internal types statistics.
:param roles: Dictionary containing the roles statistics.
:param reserved: Dictionary containing the reserved (or tokens) statistics.
"""
# walk the tree: collect nodes with assigned tokens and build the parents map
node_tokens = []
parents = {}
queue = [root]
while queue:
node = queue.pop()
internal_types[node.internal_type] += 1
for role in node.roles:
roles[role] += 1
for child in node.children:
parents[id(child)] = node
queue.extend(node.children)
if node.token or node.start_position and node.end_position and not node.children:
node_tokens.append(node)
node_tokens.sort(key=lambda n: n.start_position.offset)
sentinel = bblfsh.Node()
sentinel.start_position.offset = len(content)
sentinel.start_position.line = content.count("\n")
node_tokens.append(sentinel)
# scan `node_tokens` and analyze the gaps and the token prefixes and suffixes
pos = 0
ws = re.compile("\s+")
alpha = re.compile("[a-zA-Z]+")
IDENTIFIER = bblfsh.role_id("IDENTIFIER")
log = logging.getLogger("analyze_uast")
def ccheck(char: str) -> bool:
return not char.isspace() and not char.isalnum(
) and not ord(char) >= 128
for node in node_tokens:
token = node.token if node.token else \
content[node.start_position.offset:node.end_position.offset]
if node.start_position.offset > pos:
diff = content[pos:node.start_position.offset]
parts = ws.split(diff)
for part in parts:
if len(part) >= 8:
log.debug("Skipping weird part in code: %s. Path: %s",
diff, path)
continue
for nonalpha in alpha.split(part):
for char in nonalpha:
if ccheck(char):
reserved.add(char)
if node is sentinel:
break
pos = node.end_position.offset
if IDENTIFIER not in node.roles:
continue
outer = content[node.start_position.offset:node.end_position.offset]
if outer == token:
continue
pos = outer.find(token)
if pos < 0:
log.warning(
"skipped %s, token offset corruption \"%s\" vs. \"%s\"", path,
token, outer)
break
if pos > 0:
for char in outer[:pos]:
if ccheck(char):
reserved.add(char)
if pos + len(token) < len(outer):
for char in outer[pos + len(token):]:
if ccheck(char):
reserved.add(char)
def _parse_file(self, contents: str, root: bblfsh.Node, path: str) -> \
Tuple[List[VirtualNode], Dict[int, bblfsh.Node]]:
"""
Given the source text and the corresponding UAST this function compiles the list of
`VirtualNode`-s and the parents mapping. That list of nodes equals to the original
source text bit-to-bit after `"".join(n.value for n in nodes)`. `parents` map from
`id(node)` to its parent `bblfsh.Node`.
:param contents: source file text
:param root: UAST root node
:return: list of `VirtualNode`-s and the parents.
"""
# build the line mapping
lines = contents.split("\n")
line_offsets = numpy.zeros(len(lines) + 1, dtype=numpy.int32)
pos = 0
for i, line in enumerate(lines):
line_offsets[i] = pos
pos += len(line) + 1
line_offsets[-1] = pos
# walk the tree: collect nodes with assigned tokens and build the parents map
node_tokens = []
parents = {}
queue = [root]
while queue:
node = queue.pop()
for child in node.children:
parents[id(child)] = node
queue.extend(node.children)
if node.token or node.start_position and node.end_position and not node.children:
node_tokens.append(node)
node_tokens.sort(key=lambda n: n.start_position.offset)
sentinel = bblfsh.Node()
sentinel.start_position.offset = len(contents)
sentinel.start_position.line = len(lines)
node_tokens.append(sentinel)
# scan `node_tokens` and fill the gaps with imaginary nodes
result = []
pos = 0
parser = self.tokens.PARSER
searchsorted = numpy.searchsorted
for node in node_tokens:
if node.start_position.offset > pos:
sumlen = 0
diff = contents[pos:node.start_position.offset]
for match in parser.finditer(diff):
positions = []
for suboff in (match.start(), match.end()):
offset = pos + suboff
line = searchsorted(line_offsets, offset, side="right")
col = offset - line_offsets[line - 1] + 1
positions.append(Position(offset, line, col))
token = match.group()
sumlen += len(token)
result.append(VirtualNode(token, *positions, path=path))
assert sumlen == node.start_position.offset - pos, \
"missed some imaginary tokens: \"%s\"" % diff
if node is sentinel:
break
result.extend(VirtualNode.from_node(node, contents, path))
pos = node.end_position.offset
return result, parents
def SerializeToString(self):
return bblfsh.Node().SerializeToString()
