def write(types, string_dict, ty, tree, out):
'''Compresses ast and writes it to a byte stream.
Note, this may modify tree. ShiftAST produces trees with numbers in
double fields as ints. Our type-directed encoder coerces them to
doubles. This updates the input tree, in place, with this change.
Args:
types: idl.TypeResolver
string_dict: list of strings stored in external file.
ty: the type of 'tree'.
tree: the AST to encode.
out: byte-oriented stream to write content to.
'''
# Rewrite ints in float position to floats.
tycheck.FloatFixer(types).rewrite(ty, tree)
# Check the AST conforms to the IDL.
tycheck.TypeChecker(types).check_any(ty, tree)
out.write(SIGNATURE)
# Collect the local strings and write the string table
local_strings = strings.StringCollector(types)
local_strings.visit(ty, tree)
local_strings.strings -= set(string_dict)
local_strings = list(sorted(local_strings.strings))
string_dict = local_strings + string_dict
strings.write_dict(out, local_strings, with_signature=False)
# Build probability models of the AST and serialize it.
m = model.model_tree(types, ty, tree)
model_writer = encode.ModelWriter(types, string_dict, out)
model_writer.write(ty, m)
# Now write the file content.
def write_piece(ty, node, out):
lazy_parts = lazy.LazyMemberExtractor(types)
node = lazy_parts.replace(ty, node)
encode.encode(types, m, out, ty, node)
# Encode the lazy parts in memory
lazy_encoded = []
for _, attr, part in lazy_parts.lazies:
buf = io.BytesIO()
lazy_encoded.append(buf)
write_piece(attr.resolved_ty, part, buf)
# Write the dictionary of lazy parts, then the lazy parts
bits.write_varint(out, len(lazy_encoded))
for encoded_part in lazy_encoded:
bits.write_varint(out, encoded_part.tell())
for encoded_part in lazy_encoded:
out.write(encoded_part.getbuffer())
write_piece(ty, tree, out)
def make_dict(in_files, out_file):
types = idl.parse_es6_idl()
ty_script = types.interfaces['Script']
sources = []
for in_file in in_files:
proggy = json.loads(in_file.read())
tycheck.FloatFixer(types).rewrite(ty_script, proggy)
tycheck.TypeChecker(types).check_any(ty_script, proggy)
sources.append((ty_script, proggy))
string_dict = strings.prepare_dict(types, sources)
strings.write_dict(out_file, string_dict, with_signature=True)
def type_check(in_files):
types = idl.parse_es6_idl()
ty_script = types.interfaces['Script']
for in_file in in_files:
proggy = json.loads(in_file.read())
print(f'checking {in_file.name}... ', end='')
try:
tycheck.TypeChecker(types).check_any(ty_script, proggy)
except Exception as e:
# FIXME: Make this a specific exception type, do error recovery, etc.
print(f'NG, {e!s}')
continue
print('OK')
def read(types, string_dict, ty, inp):
'''Decompresses ast from a byte stream and returns an AST.
>>> import json
>>> import ast, idl, strings
>>> types = idl.parse_es6_idl()
>>> ty_script = types.interfaces['Script']
>>> tree_in = ast.load_test_ast('y5R7cnYctJv.js.dump')
>>> #tree_in = ast.load_test_ast('three.min.js.dump')
>>> string_dict = strings.prepare_dict(types, [(ty_script, tree_in)])
>>> buf = io.BytesIO()
>>> write(types, string_dict, ty_script, tree_in, buf)
>>> buf.tell()
1884
>>> buf.seek(0)
0
>>> tree_out = read(types, string_dict, ty_script, buf)
>>> #assert json.dumps(tree_in) == json.dumps(tree_out)
>>> s_in = json.dumps(tree_in, indent=1).split('\\n')
>>> s_out = json.dumps(tree_out, indent=1).split('\\n')
>>> for i, (l_in, l_out) in enumerate(zip(s_in, s_out)):
... if l_in != l_out:
... print(f'{i:3d} {l_in}')
... print(f' {l_out}')
... print('mismatch')
... break
Now try to round-trip something which uses laziness:
>>> import opt
>>> tree_in = opt.optimize(tree_in)
lazified 1 functions
>>> buf = io.BytesIO()
>>> write(types, string_dict, ty_script, tree_in, buf)
>>> buf.tell()
1898
>>> buf.seek(0)
0
>>> tree_out = read(types, string_dict, ty_script, buf)
>>> assert json.dumps(tree_in) == json.dumps(tree_out)
Now try to round-trip something which misses the dictionary:
>>> del string_dict[-10:-3]
>>> buf = io.BytesIO()
>>> write(types, string_dict, ty_script, tree_in, buf)
>>> buf.tell()
1934
>>> buf.seek(0)
0
>>> tree_out = read(types, string_dict, ty_script, buf)
>>> assert json.dumps(tree_in) == json.dumps(tree_out)
'''
# Read the local string table
local_strings = strings.read_dict(inp, with_signature=False)
string_dict = local_strings + string_dict
# Read the probability models
model_reader = encode.ModelReader(types, string_dict, inp)
m = model_reader.read(ty)
def read_piece(ty):
tree = encode.decode(types, m, ty, inp)
# Read the dictionary of lazy parts
# TODO: We don't need this; it is implicit in the tree we just read.
num_lazy_parts = bits.read_varint(inp)
lazy_offsets = [0]
for _ in range(num_lazy_parts):
lazy_size = bits.read_varint(inp)
lazy_offsets.append(lazy_offsets[-1] + lazy_size)
lazy_offsets = list(
map(lambda offset: offset + inp.tell(), lazy_offsets))
def restore_lazy_part(ty, attr, index):
inp.seek(lazy_offsets[index])
part = read_piece(attr.resolved_ty)
assert inp.tell() == lazy_offsets[
index + 1], f'{inp.tell()}, {lazy_offsets[index + 1]}'
return part
restorer = lazy.LazyMemberRestorer(types, restore_lazy_part)
tree = restorer.replace(ty, tree)
inp.seek(lazy_offsets[-1])
return tree
tree = read_piece(ty)
type_checker = tycheck.TypeChecker(types)
type_checker.check_any(ty, tree)
return tree
# This file transforms ASTs in ways outside the scope of file format
# and compression. There are no BinAST minifiers yet, so for now it's
# convenient to integrate it into the encoder/decoder tool.
import doctest
import ast
import idl
import types
import tycheck
# Test helpers.
resolver = idl.parse_es6_idl()
resolver.resolve_types()
checker = tycheck.TypeChecker(resolver)
def check_ty(ty_name, obj):
'''Checks whether obj matches the type named ty_name.'''
ty = resolver.resolve(idl.TyRef(ty_name))
checker.check_any(ty, obj)
def make_iffe():
'''Makes an AST fragment which is an immediately-invoked function expression.
>>> check_ty('Expression', make_iffe())
'''
return {
'type': 'CallExpression',
