Re: [graphviz-interest] Using HTML to name a node

Ron Savage <[email protected]> Thu, 18 Sep 2014 10:05:45 +1000
Newsgroups gmane.comp.video.graphviz
Message-ID <[email protected]>
Hi Emden

On 18/09/14 03:38, Emden R. Gansner wrote:
> On 9/16/14 9:49 PM, Ron Savage wrote:
>> This is for dot - graphviz version 2.38.0 (20140413.2041).
>>
>> I've found an effect which I regard as an anomaly when a HTML label is
>> used to name a node.
>>
>> Firstly, I think on this page
>> http://www.graphviz.org/content/dot-language at the 1st reference to
>> HTML labels - "an HTML string (<...>)." - there should be a link to
>> this page/anchor http://www.graphviz.org/content/node-shapes#html.
>>
> This page describes the DOT language, for which an HTML string can be
> any legal XML string and used like any other array of uninterpreted
> bytes. But
> I did add a note concerning labels and a link to the more restrictive
> syntax.

Excellent.

>> As for the attached, would it make more sense that the name of
>> node_59_1 be rendered as node_59_1 and not as <....>?
> For the language and parser, it was simpler to treats strings as
> strings. And we assumed users wanting the special HTML-like labels would
> use an HTML string for the label value. So you are touching on a dark,
> unclarified aspect. By default, a node label is "\N", which is replaced

I agree I was pushing the envelope :-).

> by the node's name
> fairly late during processing. The label is an ordinary string, so it is
> interpreted as an ordinary string, with \N being replaced with the raw
> bytes of the
> node's name. So what you are seeing is what would be expected.
>
> One could suggest that we should catch the case of a raw label=\N,
> replace it with an HTML string if that is the node's name, and go on
> from there.
> With this interpretation, node node_59_1 would appear like node
> node_59_3. Another alternative is the one you suggest but what happens when
> there are multiple text fields?
>
> My feeling is that these last two interpretations aren't that helpful to
> justify the coding changes required, especially in the latter case. My
> inclination would
> be to leave things as they are with a comment stressing that, if you
> want an HTML-like label, use an HTML string for the label value.

That's fine, as long as something is documented, and of course people 
can be pointed to the mail archive.

-- 
Ron Savage - savage.net.au
_______________________________________________
[email protected]
http://lists.research.att.com/mailman/listinfo/graphviz-interest