jonathan.slend...@gmail.com wrote:
>> Can you expand a bit on how array("u") helps here? Are the matches in the
>> gigabyte range?
>
> I have a string of unicode characters, e.g.:
>
> data = array.array('u', u'x' * 1000000000)
>
> Then I need to change some data in the middle of this string, for
> instance:
>
> data[500000] = 'y'
>
> Then I want to use re to search in this text:
>
> re.search('y', data)
>
> This has to be fast. I really don't want to split and concatenate strings.
> Re should be able to process it and the expressions can be much more
> complex than this. (I think it should be anything that implements the
> buffer protocol).
>
> So, this works perfectly fine and fast. But it scares me that it's
> deprecated and Python 4 will not support it anymore.
Hm, this doesn't even work with Python 3:
>>> data = array.array("u", u"x"*1000)
>>> data[100] = "y"
>>> re.search("y", data)
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/usr/lib/python3.4/re.py", line 166, in search
return _compile(pattern, flags).search(string)
TypeError: can't use a string pattern on a bytes-like object
You can search for bytes
>>> re.search(b"y", data)
<_sre.SRE_Match object; span=(400, 401), match=b'y'>
>>> data[101] = "z"
>>> re.search(b"y", data)
<_sre.SRE_Match object; span=(400, 401), match=b'y'>
>>> re.search(b"yz", data)
>>> re.search(b"y\0\0\0z", data)
<_sre.SRE_Match object; span=(400, 405), match=b'y\x00\x00\x00z'>
but if that is good enough you can use a bytearray in the first place.
--
https://mail.python.org/mailman/listinfo/python-list
