Re: [Docutils-develop] spurious target?

[David G. <go...@py...>]

On Wed, Jun 3, 2009 at 10:28, Guenter Milde <mi...@us...> wrote:
> On 2009-06-02, David Goodger wrote:
>> On Tue, Jun 2, 2009 at 06:57, Guenter Milde <mi...@us...> wrote:
>
>>> The following snipped from the functional test's standard.txt ::
>
>>>  .. _Another Target:
>
>>>  Targets
>>>  -------
>
>>> is converted to the following pseudo-xml::
>
>>>      <target refid="another-target">
>>>  <section dupnames="targets"
>             ids="targets another-target"
>             names="another\ target">
>>>      <title auto="1" refid="id51">
>>>          <generated classes="sectnum">
>>>              2.13
>>>          Targets
>
>>> You see the "another-target" as both, refid of a <target> and id of the
>>> following node.
>
> But is it really needed/sensible to have both?
> This is why I asked if this is a doubling already in the doctree.
>
>> "refid" is a reference to an ID. If you look at the doctree before the
>> transforms are applied (use tools/quicktest.py), you'll see that
>> initially it's <target id="another-target" ...>, and the ID is
>> transferred to the following element, the section. There is only one
>> "another-target" ID. The <target refid="..."> node is left behind by
>> the transform and should be ignored by the writer.
>
> But why is it left behind, if it should be ignored?

The transform that handles this is the
docutils.transforms.references.PropagateTargets. See its docstring &
code. I can't recall exactly why <target refid="..."> nodes are left
behind (perhaps simply to keep track for debugging purposes), but I
don't have time to audit the entire codebase to see if these nodes can
be removed without ill-effects.

>> There should be no doubling. Take a look at the logic in
>> docutils/writers/html4css1/__init__.py::HTMLTranslator.visit_target.
>
> Wouldn't it be easier to have the logic just once in the tranlator?

Every writer needs to process targets anyway (for non-ref attributes), so no.

It seems to me that it would be easier to fix the LaTeX writer.
Everything else works, and if it works, don't fix it.

-- 
David Goodger <http://python.net/~goodger>