UnicodeDecodeError help please?

[Paul Boddie]

Robin Haswell wrote:
> Okay I'm getting really frustrated with Python's Unicode handling, I'm
> trying everything I can think of an I can't escape Unicode(En|De)codeError
> no matter what I try.

If you follow a few relatively simple rules, the days of Unicode errors
will be over. Let's take a look!

> Could someone explain to me what I'm doing wrong here, so I can hope to
> throw light on the myriad of similar problems I'm having? Thanks :-)
>
> Python 2.4.1 (#2, May  6 2005, 11:22:24)
> [GCC 3.3.6 (Debian 1:3.3.6-2)] on linux2
> Type "help", "copyright", "credits" or "license" for more information.
> >>> import sys
> >>> sys.getdefaultencoding()
> 'utf-8'

Note that this only specifies the encoding assumed to be used in plain
strings when such strings are used to create Unicode objects. For some
applications this is sufficient, but where you may be dealing with many
different character sets (or encodings), having a default encoding will
not be sufficient. This has an impact below and in your wider problem.

> >>> import htmlentitydefs
> >>> char = htmlentitydefs.entitydefs["copy"] # this is an HTML © - a copyright symbol
> >>> print char
> ©

It's better here to use repr(char) to see exactly what kind of object
it is (or just give the name of the variable at the prompt). For me,
it's a plain string, despite htmlentitydefs defining the each name in
terms of its "Unicode codepoint". Moreover, for me the plain string
uses the "Latin-1" (or more correctly iso-8859-1) character set, and I
imagine that you get the same result.

> >>> str = u"Apple"
> >>> print str
> Apple
> >>> str + char
> Traceback (most recent call last):
>   File "<stdin>", line 1, in ?
> UnicodeDecodeError: 'utf8' codec can't decode byte 0xa9 in position 0: unexpected code byte

Here, Python attempts to make a Unicode object from char, using the
default encoding (which is utf-8) and finds that char is a plain string
containing non-utf-8 character values, specifically a single iso-8859-1
character value. It consequently complains. This is quite unfortunate
since you probably expected Python to give you the entity definition
either as a Unicode object or a plain string of your chosen encoding.

Having never used htmlentitydefs before, I can only imagine that it
provides plain strings containing iso-8859-1 values in order to support
"legacy" HTML processing (given that semi-modern HTML favours &#xx;
entities, and XHTML uses genuine character sequences in the stated
encoding), and that getting anything other than such strings might not
be particularly useful.

Anyway, what you'd do here is this:

str + unicode(char, "iso-8859-1)

Rule #1: if you have plain strings and you want them as Unicode, you
must somewhere state what encoding those strings are in, preferably as
you convert them to Unicode objects. Here, we can't rely on the default
encoding being correct and must explicitly state a different encoding.
Generally, stating the encoding is the right thing to do, rather than
assuming some default setting that may differ across environments.
Somehow, my default encoding is "ascii" not "utf-8", so your code would
fail on my system by relying on the default encoding.

[...]

> Basically my app is a search engine - I'm grabbing content from pages
> using HTMLParser and storing it in a database but I'm running in to these
> problems all over the shop (from decoding the entities to calling
> str.lower()) - I don't know what encoding my pages are coming in as, I'm
> just happy enough to accept that they're either UTF-8 or latin-1 with
> entities.

Rule #2: get your content as Unicode as soon as possible, then work
with it in Unicode. Once you've made your content Unicode, you
shouldn't get UnicodeDecodeError all over the place, and the only time
you then risk an UnicodeEncodeError is when you convert your content
back to plain strings, typically for serialisation purposes.

Rule #3: get acquainted with what kind of encodings apply to the
incoming data. If you are prepared to assume that the data is either
utf-8 or iso-8859-1, first try making Unicode objects from the data
stating that utf-8 is the encoding employed, and only if that fails
should you consider it as iso-8859-1, since an utf-8 string can quite
happily be interpreted (incorrectly) as a bunch of iso-8859-1
characters but not vice versa; thus, you have a primitive means of
validation.

> Any help would be great, I just hope that I have a brainwave over the
> weekend because I've lost two days to Unicode errors now. It's even worse
> that I've written the same app in PHP before with none of these problems -
> and PHP4 doesn't even support Unicode.

Perhaps that's why you never saw any such problems, but have you looked
at the quality of your data?

Paul