[SOLVED] Least-lossy string.encode to us-ascii?

[Vlastimil Brom]

2012/9/14 Tim Chase <python.list at tim.thechases.com>:
> On 09/13/12 16:44, Vlastimil Brom wrote:
>> >>> import unicodedata
>> >>> unicodedata.normalize("NFD", u"serviço móvil").encode("ascii", "ignore").decode("ascii")
>> u'servico movil'
>
> Works well for all the test-cases I threw at it.  Thanks!
>
> -tkc
>
>

Hi,
I am glad, it works, but I agree with the other comments, that it
would be preferable to keep the original accented text, if at all
possible in the whole processing.
The above works by decomposing the accented characters into "basic"
characters and the bare accents (combining diacritics) using
normalize() and just striping anything outside ascii in encode("...",
"ignore")
This works for "combinable" accents, and most of the Portuguese
characters outside of ascii appear to fall into this category, but
there are others as well.
E.g. according to
http://tlt.its.psu.edu/suggestions/international/bylanguage/portuguese.html
there are at least ºª«»€, which would be lost completely in such conversion.
ª (dec.: 170)  (hex.: 0xaa) # FEMININE ORDINAL INDICATOR
º (dec.: 186)  (hex.: 0xba) # MASCULINE ORDINAL INDICATOR

You can preprocess such cases as appropriate before doing the
conversion, e.g. just:

>>> u"ºª«»€".replace(u"º", u"o").replace(u"ª", u"a").replace(u"«", u'"').replace(u"»", u'"').replace(u"€", u"EUR")
u'oa""EUR'
>>>
or using a more elegant function and the replacement lists (eventually
handling other cases as well).

regards,
   vbr