Question

I’m writing a simple regular expression parser for the output of the sensors utility on Ubuntu. Here’s an example of a line of text I’m parsing:

temp1:        +31.0°C  (crit = +107.0°C)

And here’s the regex I’m using to match that (in Python):

temp_re = re.compile(r'(temp1:)\s+(\+|-)(\d+\.\d+)\W\WC\s+' 
                     r'\(crit\s+=\s+(\+|-)(\d+\.\d+)\W\WC\).*')

This code works as expected and matches the example text I’ve given above. The only bits I’m really interested in are the numbers, so this bit:

(\+|-)(\d+\.\d+)\W\WC

which starts by matching the + or - sign and ends by matching the °C.

My question is, why does it take two \W (non-alphanumeric) characters to match ° rather than one? Will the code break on systems where Unicode is represented differently to mine? If so, how can I make it portable?

Answer 1

Possible portable solution:

Convert input data to unicode, and use re.UNICODE flag in regular expressions.

#!/usr/bin/env python
# -*- coding: utf-8 -*-
import re


data = u'temp1:        +31.0°C  (crit = +107.0°C)'
temp_re = re.compile(ur'(temp1:)\s+(\+|-)(\d+\.\d+)°C\s+' 
                     ur'\(crit\s+=\s+(\+|-)(\d+\.\d+)°C\).*', flags=re.UNICODE)

print temp_re.findall(data)

Output

[(u'temp1:', u'+', u'31.0', u'+', u'107.0')]

EDIT

@netvope allready pointed this out in comments for question.

Update

Notes from J.F. Sebastian comments about input encoding:

check_output() returns binary data that sometimes can be text (that should have a known character encoding in this case and you can convert it to Unicode). Anyway ord(u’°’) == 176 so it can not be encoded using ASCII encoding.

So, to decode input data to unicode, basically* you should use encoding from system locale using locale.getpreferredencoding() e.g.:

data = subprocess.check_output(...).decode(locale.getpreferredencoding())

With data encoded correctly:

you’ll get the same output without re.UNICODE in this case.

---

Why basically? Because on Russian Win7 with cp1251 as preferredencoding if we have for example script.py which decodes it’s output to utf-8:

#!/usr/bin/env python
# -*- coding: utf8 -*-

print u'temp1: +31.0°C  (crit = +107.0°C)'.encode('utf-8')

And wee need to parse it’s output:

subprocess.check_output(['python', 
                         'script.py']).decode(locale.getpreferredencoding())

will produce wrong results: 'В°' instead °.

So you need to know encoding of input data, in some cases.