Re: [xml] Attribute indenting

["Nicolas Mailhot" <nicolas mailhot laposte net>]

On Jeu 28 avril 2005 13:33, Daniel Veillard a écrit :
> On Mon, Apr 25, 2005 at 05:54:55PM +0200, Nicolas Mailhot wrote:
> > Hi,
> >
> > Is there any good reason not to indent properly attributes when xmllint
> > is
> > invoked with --format ? When I say indenting I think about constructions
> > like:
> >
> >  <foo bar1="..."
> >       bar2="..."
> >       bar3="...">
> >
> > which you see in all XML textbooks but can not achieve with most current
> > XML formaters.
>
>   because for document with a lot of deep, lots of attributes and little
> CDATA you could end up with document serialization many multiplied by a
> factor
> or 2 or 3 easilly for no practical gain.

Well I assume when you ask document reformating you're ready to pay the
formating cost.

Or are you telling me some processes reformat on th fly without any real
need to ?

> > I ask this because someone told me last week that using more than one
> > attribute per element should be avoided because the way tools indent
> > them
> > (inline) a human can not parse them easily (SIC) so one can not write
> > readable XML with attributes. And indeed I see that any XML file that
> > uses
>
>   I see no valid argument there.

Well, I'm afraid the other person was right when arguing that tools
(including libxml) seem to consider attributes second-class citizens that
should not be prettified alogside elements.

> > namespace declarations (big fat attributes) is very difficult to read
> > once
> > reformated by xmllint - the output is fine for elements but not much
> > else.
>
>   There is so many axis to the "formatting" of markup, than trying to
> provide customization for each an every axis would just be insanely
> complex
> and confusing APIs. The in-memory format is public, write your own
> serializer
> if you really need something very custom.

It's not about doing something very custom. Since --format is useful only
when a human reads the output (tools do not need the pretty indenting) I
assumed its aim was to produce code as human-readable as possible, ie use
the same tricks as in XML reference books.

I don't see the point of indenting a document at all if a large part of
its contents (attributes) is still lumped together. It does not take a lot
of attributes to get line wrapping mess - it's a general case.

Regards,

-- 
Nicolas Mailhot