[Kimchi-devel] [PATCH V5] add a make check-local command to verify the i18n string formatting

[Zhou Zheng Sheng]

on 2014/04/08 21:18, shaohef at linux.vnet.ibm.com wrote:
> +PATTERN = re.compile("%\([^\)]*\)[#0\-\+ ]?(\d+|\*)?(\.(\d+|\*))?"
> +                     "[lLh]?[cdeEfFgGioursxX%]")
> +BAD_PATTERN = re.compile("%\([^\)]*\)")

There are some problems in the regular expressions.

The first problem is that "\" is a special character in Python string.
For example, we all know "\" gives a special meaning to "\n" and "\t".
If we want to use it in a regular expression, we have to escape itself,
for example re.compile("%\\(\\)"). This is tedious and error-prone.
Usually we just use a prefix "r" before the string literal to stop
Python from translating "\", for example, re.compile(r"%\(\)").

The second problem is that "+", "*", "(", ")" automatically lose their
special meaning inside "[ ]", we can use them directly in "[ ]".

The third problem is that we should comment each part of a complicated
regular expression.

As a result, I suggest the following the regular expression.

# Match all conversion specifier with mapping key
PATTERN = re.compile(r'''%\([^)]+\)  # Mapping key
                         [#0\-+]?  # Conversion flags (optional)
                         (\d+|\*)?  # Minimum field width (optional)
                         (\.(\d+|\*))?  # Precision (optional)
                         [lLh]?  # Length modifier (optional)
                         [cdeEfFgGioursxX%]  # Conversion type''',
                    re.VERBOSE)
BAD_PATTERN = re.compile(r"%\([^)]*?\)")

Notice I drop the space from the conversion flags, and leave only "#",
"0", "-" and "+". It is fully legal for someone to write the following,
  "%(k) done" % {"k": 100}
and the result is
  ' 100one'

This is because Python interpreters the space between "%(k)" and "done"
as the conversion flag, then Python eats "d" from "done" as the
conversion type. Though it's legal, it's error prone. So I think it's
better not to support this rare usage of space in the conversion flags,
and report it as error.


-- 
Thanks and best regards!

Zhou Zheng Sheng / 周征晟
E-mail: zhshzhou at linux.vnet.ibm.com
Telephone: 86-10-82454397