[kaffe] japitools, subMap, and html output

[Stuart Ballard]

Dalibor Topic wrote:
> Hi Stuart,
> 
> I'm really impressed how cool japitools has become.

Thanks :)

> It looks acceptable in w3m (text mode browser), but
> the bars don't really look good. The tables are o.k.,
> though. You might want to add some space between the
> columns.

Yeah, I can imagine that rendering the bars in ascii mode would look a 
bit funny. Do you know if there's any way to specify something in a 
stylesheet (eg display:none on the whole "bar" table) that would apply 
to w3m and other text-mode browsers only? Maybe an @media type...

> It doesn't look too good on lynx (another text mode
> browser), but lynx has problems with tables anyway.

Yeah, lynx wasn't really in my target market :)

> An alternative to using the spacer gif could be to use
> colored empty table cells with ascii symbols in the
> background color. That should work nicely on w3m and
> the ascii symbols (a different one for each color, of
> course) should work nicely on lynx, too. It will make
> the web pages somewhat bigger, though. Check out
> http://www.illuminated.co.uk/gif2html/ to see what I
> mean. ;)

Hmm... that's a really cool program and a nice idea in principle, but it 
doesn't seem to be able to get down to single-pixel resolution. If we 
could find a way to degrade to that kind of output for text-mode 
browsers (and perhaps browsers with poor or missing CSS support) but 
still give pixel-accurate results like the current ones are for 
modern/graphical browsers, I'd be happy :) (my first thought, in fact, 
was to use a "." character and just skip it if the cell was less than 
5px wide. Unfortunately, in japitools output it seems that cells less 
than 5px wide are very common, so the results looked really funky in NS4.x).

> Also, the comparison document for jdk1.3 doesn't
> appear to validate as HTML 3.2, according to
> http://validator.w3.org/check?uri=http%3A%2F%2Frainbow.netreach.net%2F%7Esballard%2Fjapi%2Fhtmlout%2Fh-jdk13-classpath.html&charset=%28detect+automatically%29&doctype=HTML+3.2
> or any other HTML version I tried. You could try
> running jtidy over your generated output and patching
> the output method accordingy. That might explain some
> of the browser rendering results.

According to the validator, "class" isn't a valid attribute on tables. 
That seems to imply that the wrong doctype is being used... of course, 
that's my fault for not including a doctype in the output.

Specifying doctype 4.01 transitional gives more realistic errors. It 
seems I closed a <th> when I meant to close a <tr>, and I need to put 
'alt=""' on each of my 1px images (hmm, I suppose I could put 
alt="XXXXXXX" where the number of X's is dynamic...). I'll fix both of 
these, and add a 4.01 transitional doctype, in the next day or two.

> As running a kaffe nightly job to generate a
> kaffe.japi.gz file depends on getting TreeMap &
> TreeSet from Classpath integrated into kaffe, and
> Archie is o.k. with that approach, I'll give it a try
> in the next couple of days and see how far I get.
> Right now I'm trying to fix some build issues on
> Mandrake 9 & Cygwin, and it looks promising. I'll get
> to TreeMap/TreeSet afterwards.

Well, the "usual" way of generating a japi seems to currently be to run 
japize with the JDK. But if you can get Kaffe to do it, that would make 
me extremely happy :)

Stuart.

-- 
Stuart Ballard, Programmer
NetReach - Internet Solutions
(215) 283-2300, ext. 126
http://www.netreach.com/