On Wed, 2003-06-25 at 21:32, Jewett, Jim J wrote: > Patrick Ohly > > > I have already verified manually that it is possible to replace > > the <a href="javascript"></a> with a <form>...</form>, but this > > doesn't really help I guess. Is there any way how I can cause > > any of the plucker frontends to trigger a POST operation and then > > include the result in the pdb file? > > Python can handle posts. (url-fetching functions take an optional > third argument of "data".) PyPlucker has some support for it, but > not quite. (It may be urltext_key). To my knowledge, you would > probably have to make code changes. > > Part of the catch is that post actions are supposed to involve > the user saying "yes, I really want to do this", but get actions > can be automatic. This means that at a properly designed site, > GET is safe, but POST might order thousands of dollars worth > of junk. You don't want that happening by accident.
Correct, so the default really should be not to submit forms. Suppose I wanted to extend either PyPlucker or JPluck to do that, what would be a good rule to decide when a form is to be treated like a link and when not? I can think of one criteria that should prevent submitting a form: if the form has visible input fields, then an automated posting probably makes no sense. Unfortunately the opposite is not true: even if all fields are hidden, posting the form might still have undesired side effects. I guess I'll just settle for a Python front-end script that downloads the initial page, downloads the other pages using POST and then calls JPluck to convert the local files. > > Ideally, I'd let JPluck transform the starting page with XSLT > > automatically. As I have never used XSLT before, I am looking for > > an easier way to debug a stylesheet than running JPluck on > > a page and looking at the resulting pdb: in particular, it would > > be nice to get a serial dump of the parse tree that the stylesheet > > is applied to and the result of the transformation. > > I agree, but can't offer much help just yet. I have heard that IE > can use the same XSLT, which might speed your testing. I get the pages transformed as desired with xsltproc, but not if I use the same stylesheet in JPluck directly. Hmm, looking at the source of JPluck I found Transform.java and transform.jar, which I should probably have used instead of xstlproc. [a little while latter] java -jar transform.jar also works as expected, but I still cannot use stylesheets in JPluck 2.0 pre 5 directly. Here's what I did: - clean the start page with JPluck 0.9 - transform clean start page with transform.jar from JPluck 2.0 - convert transformed page with convert.jar from JPluck 2.0 -> pdb looks as expected - convert the original, uncleaned (but downloaded) page with JPluck 2.0, applying my stylesheet on-the-fly -> pdb shows that stylesheet was applied, but most of the text is missing - convert the page that was cleaned by JPluck 0.9 with JPluck 2.0, applying the my stylesheet on-the-fly -> same problem It seems to me that doing the xsl transformation on-the-fly somehow leads to different results than doing it seperately, even if the same Java class is used. How can I debug this problem? I could provide the page that I try to plug and my xsl/jxl files, if that would help. -- Freundliche Gruesse / Best Regards Patrick Ohly Senior Software Engineer -------------------------------------------------------------------- //// pallas Pallas GmbH / Hermuelheimer Str. 10 / 50321 Bruehl / Germany [EMAIL PROTECTED] / www.pallas.com Tel +49-2232-1896-30 / Fax +49-2232-1896-29 -------------------------------------------------------------------- _______________________________________________ plucker-list mailing list [EMAIL PROTECTED] http://lists.rubberchicken.org/mailman/listinfo/plucker-list
