1
0
forked from VimPlug/jedi
Files
jedi-fork/jedi/common.py
Dave Halter 989e4bac89 Speed up splitlines.
We use the python function again with the modifications we need.
I ran it with:

    python3 -m timeit  -n 10000 -s 'from jedi.common import splitlines; x = open("test_regression.py").read()'

The speed differences are quite remarkable, it's ~3 times faster:

    10000 loops, best of 3: 52.1 usec per loop

vs. the old:

    10000 loops, best of 3: 148 usec per loop

We might need to speedup splitlines with  as well. It's probably
also a factor 2-3 slower than it should be.
2017-03-09 08:58:57 +01:00

198 lines
6.2 KiB
Python

""" A universal module with functions / classes without dependencies. """
import sys
import contextlib
import functools
import re
from ast import literal_eval
from jedi._compatibility import unicode, reraise
from jedi import settings
class UncaughtAttributeError(Exception):
"""
Important, because `__getattr__` and `hasattr` catch AttributeErrors
implicitly. This is really evil (mainly because of `__getattr__`).
`hasattr` in Python 2 is even more evil, because it catches ALL exceptions.
Therefore this class originally had to be derived from `BaseException`
instead of `Exception`. But because I removed relevant `hasattr` from
the code base, we can now switch back to `Exception`.
:param base: return values of sys.exc_info().
"""
def safe_property(func):
return property(reraise_uncaught(func))
def reraise_uncaught(func):
"""
Re-throw uncaught `AttributeError`.
Usage: Put ``@rethrow_uncaught`` in front of the function
which does **not** suppose to raise `AttributeError`.
AttributeError is easily get caught by `hasattr` and another
``except AttributeError`` clause. This becomes problem when you use
a lot of "dynamic" attributes (e.g., using ``@property``) because you
can't distinguish if the property does not exist for real or some code
inside of the "dynamic" attribute through that error. In a well
written code, such error should not exist but getting there is very
difficult. This decorator is to help us getting there by changing
`AttributeError` to `UncaughtAttributeError` to avoid unexpected catch.
This helps us noticing bugs earlier and facilitates debugging.
.. note:: Treating StopIteration here is easy.
Add that feature when needed.
"""
@functools.wraps(func)
def wrapper(*args, **kwds):
try:
return func(*args, **kwds)
except AttributeError:
exc_info = sys.exc_info()
reraise(UncaughtAttributeError(exc_info[1]), exc_info[2])
return wrapper
class PushBackIterator(object):
def __init__(self, iterator):
self.pushes = []
self.iterator = iterator
self.current = None
def push_back(self, value):
self.pushes.append(value)
def __iter__(self):
return self
def next(self):
""" Python 2 Compatibility """
return self.__next__()
def __next__(self):
if self.pushes:
self.current = self.pushes.pop()
else:
self.current = next(self.iterator)
return self.current
@contextlib.contextmanager
def scale_speed_settings(factor):
a = settings.max_executions
b = settings.max_until_execution_unique
settings.max_executions *= factor
settings.max_until_execution_unique *= factor
try:
yield
finally:
settings.max_executions = a
settings.max_until_execution_unique = b
def indent_block(text, indention=' '):
"""This function indents a text block with a default of four spaces."""
temp = ''
while text and text[-1] == '\n':
temp += text[-1]
text = text[:-1]
lines = text.split('\n')
return '\n'.join(map(lambda s: indention + s, lines)) + temp
@contextlib.contextmanager
def ignored(*exceptions):
"""
Context manager that ignores all of the specified exceptions. This will
be in the standard library starting with Python 3.4.
"""
try:
yield
except exceptions:
pass
def source_to_unicode(source, encoding=None):
def detect_encoding():
"""
For the implementation of encoding definitions in Python, look at:
- http://www.python.org/dev/peps/pep-0263/
- http://docs.python.org/2/reference/lexical_analysis.html#encoding-declarations
"""
byte_mark = literal_eval(r"b'\xef\xbb\xbf'")
if source.startswith(byte_mark):
# UTF-8 byte-order mark
return 'utf-8'
first_two_lines = re.match(br'(?:[^\n]*\n){0,2}', source).group(0)
possible_encoding = re.search(br"coding[=:]\s*([-\w.]+)",
first_two_lines)
if possible_encoding:
return possible_encoding.group(1)
else:
# the default if nothing else has been set -> PEP 263
return encoding if encoding is not None else 'utf-8'
if isinstance(source, unicode):
# only cast str/bytes
return source
encoding = detect_encoding()
if not isinstance(encoding, unicode):
encoding = unicode(encoding, 'utf-8', 'replace')
# cast to unicode by default
return unicode(source, encoding, 'replace')
def splitlines(string, keepends=False):
"""
A splitlines for Python code. In contrast to Python's ``str.splitlines``,
looks at form feeds and other special characters as normal text. Just
splits ``\n`` and ``\r\n``.
Also different: Returns ``['']`` for an empty string input.
In Python 2.7 form feeds are used as normal characters when using
str.splitlines. However in Python 3 somewhere there was a decision to split
also on form feeds.
"""
if keepends:
lst = string.splitlines(keepends=True)
# We have to merge lines that were broken by form feed characters.
merge = []
for i, line in enumerate(lst):
if line.endswith('\f'):
merge.append(i)
for index in reversed(merge):
try:
lst[index] = lst[index] + lst[index + 1]
del lst[index + 1]
except IndexError:
# index + 1 can be empty and therefore there's no need to
# merge.
pass
# The stdlib's implementation of the end is inconsistent when calling
# it with/without keepends. One time there's an empty string in the
# end, one time there's none.
if string.endswith('\n') or string == '':
lst.append('')
return lst
else:
return re.split('\n|\r\n', string)
def unite(iterable):
"""Turns a two dimensional array into a one dimensional."""
return set(typ for types in iterable for typ in types)
def to_list(func):
def wrapper(*args, **kwargs):
return list(func(*args, **kwargs))
return wrapper