Re: [PATCH 0/2] Add script to convert BiBTeX output (was Re: [RFC PULL] Bibliography URL cleanup)

["Paul E. McKenney" <paulmck@linux.vnet.ibm.com>]

On Mon, Dec 26, 2016 at 11:52:40PM +0900, Akira Yokosawa wrote:
> On 2016/12/25 08:15:09 -0800, Paul E. McKenney wrote:
> > On Sun, Dec 25, 2016 at 08:42:18PM +0900, Akira Yokosawa wrote:
> >> On 2016/12/24 16:36:32 +0900, Akira Yokosawa wrote:
> >>> On 2016/12/24 23:05:05 -0800, Paul E. McKenney wrote:
> >>>> On Sun, Nov 06, 2016 at 03:47:19PM -0800, Paul E. McKenney wrote:
> >>>>> On Sat, Nov 05, 2016 at 05:16:19PM +0900, Akira Yokosawa wrote:
> >>>>>> Hi Paul,
> >>>>>>
> >>>>>> On 2016/10/28, 11:30:38 -0700, Paul E. McKenney wrote:
> >>>>>>> On Fri, Oct 28, 2016 at 07:45:16AM +0900, Akira Yokosawa wrote:
> >>>>>> [snip]
> >>>>>>>> So, these bib files are an library collected for nearly three decades!!!
> >>>>>>>> They are invaluable as they are, and I'd appreciate your decision to
> >>>>>>>> make them public.
> >>>>>>>
> >>>>>>> Unfortunately, many of the comments on the early entries reflect my
> >>>>>>> relative youth and impetuosity, so unless or until I get time to edit
> >>>>>>> the whole mess so as to avoid offending any number of authors (to say
> >>>>>>> nothing of their disciples!), I must keep the originals private.
> >>>>>>
> >>>>>> I see. I misunderstood the circumstances. So you made only a part of your
> >>>>>> bib files public.
> >>>>>>
> >>>>>>>
> >>>>>>>> There are two issues in urls in the bib files.
> >>>>>>>> One is the inconsistency of format discussed here.
> >>>>>>>> The other is the dead links. There are quite a few urls that end up in
> >>>>>>>> "not found" now. Maintaining urls would require a great deal of work itself...
> >>>>>>>>
> >>>>>>>> To make the format consistent, a script would work. But before beginning
> >>>>>>>> implementation, we need to clarify what the script would do.
> >>>>>>>> So I'll make some sample replacement patches to confirm your preference.
> >>>>>>>
> >>>>>>> Sounds good, and I look forward to seeing them!
> >>>>>>
> >>>>>> I said to make "some sample replacement patches", but it became quite
> >>>>>> intensive changes. So I'm sending them as a pull request. I don't expect
> >>>>>> you to actually pull them as it is, but just to pull them on a local
> >>>>>> branch and see what they look like.
> >>>>>
> >>>>> I have pulled thme into akiyks.2016.11.05a, and pushed the first nine
> >>>>> patches.  I am reflecting those changes in my bib sources as well.
> >>>>> Looked sane at first glance, but yes, I need to work out how to handle
> >>>>> the later ones with other documents...
> >>>>
> >>>> And the later .bib edits seem compatible with my current tools, even
> >>>> without the script and .bst changes.  So I have applied them in tandem
> >>>> to my .bib source and to the perfbook bibliography.  I was able to
> >>>> find valid URLs for a few of the entries, so used them instead of
> >>>> \nolinkurl{}, but several do appear to be quite dead.
> >>>>
> >>>> I also applied the alphapf.bst changes, but left inlinelinks disabled
> >>>> for the time being.  (I am concerned about leaving authors off.)
> >>>>
> >>>> Thank you very much for your work on this!!!
> >>>
> >>> You are welcome!
> >>>
> >>> Now, the branch "bib-append-doi" might have some conflicts.
> >>> I'll rebase it and submit a v2 of the PULL request.
> >>>
> >>> inlinelinks seems to have upper limits of around 13 authors.
> >>> I've not found any workaround of the issue...
> >>> Can't you live with a short author field supplemented by a note of
> >>> full authors list?
> >>
> >> Hi Paul,
> >>
> >> As you have switched to alphapf.bst, I investigated this issue.
> >> It turned out that the length of author list is *not* the root cause.
> >> By changing the length of the list, I unintentionally moved the
> >> page boundary in Bibliography, and that resolved the build error.
> >>
> >> The root cause of the error was an href-ed *long* title which crossed
> >> a page boundary.
> > 
> > Thank you for tracking this down!  This is a title as in the title
> > of a citation, or as in the title of an href?  (My guess is the latter,
> > but it has been about 20 years since I last used perl.)
> 
> Yes, the latter.
> 
> > 
> >> In perfbook, you can't control page breaks in Bibliography by the
> >> nature of the project. So, I wrote a perl script to convert output of
> >> BiBTeX on the fly.
> > 
> > Indeed, and this presumably means that the current project is
> > randomly vulnerable to this bug.  Are we the only ones to see it?
> > If not, is a fix on the way?
> 
> If we were the only ones, I could have never found the root cause :-)
> I did a search of the error message:
> "! pdfTeX error (ext4): \pdfendlink ended up in different nesting
> level than \pdfstartlink."
> 
> It brought me to a page at:
> 
> http://tex.stackexchange.com/questions/1522/pdfendlink-ended-up-in-different-nesting-level-than-pdfstartlink
> 
> This issue was opened 6 years ago. A recent comment (of a workaround) there
> was given by Heiko Oberdiek who is one of the current maintainers of hyperref
> package.
> 
> There is another page at
> 
> http://tex.stackexchange.com/questions/249579/pdfendlink-ended-up-in-different-nesting-level-than-pdfstartlink-error-with
> 
> It says that two column layout is related to the issue.
> 
> You can also find the following description in Limitations section of hyperref's README:
> 
>   Links across pages
>   In general they have problems:
>   * Some driver doesn’t support them at all (see above).
>   * The driver allows it, but the link result might include
>   the footer and/or header, or an error message can
>   occur sometimes.
> 
> So, I can say that this issue is well recognized by the maintainers.

And it does sound like you extensively researched this issue!  Thank
for doing so!

> >> Now, you can enable "inlinelins". Note that in appending url info to
> >> .bib files, we may need to add more conversion patters. 
> > 
> > Actually, my current thought is to revert the alpha.bst commits
> > pending a fix.  Though I don't see the need to be too quick about
> > it, given that it seems to work for both 1c and 2c at the moment.
> > 
> > Or am I missing some reason why this cannot be fixed?
> 
> It seems unlikely to be fixed soon.
> 
> I remember tweaking book.cls to fix indentation behavior in
> two column layout. My impression is that two column layout is not
> as reliable as one column one...
> 
> One idea is to use alphapf.bst only for 1c with inlinelinks enabled. 
> 
> Thoughts?

I am good with enabling alphaph.bst inlinelinks only for 1c!

							Thanx, Paul

> And I have no objection to reverting commit ca109cf61bbf for now.
> 
>                                                Thanks, Akira
> 
> > 							Thanx, Paul
> > 
> >>                                              Thanks, Akira
> >>
> >> --
> >> Akira Yokosawa (2):
> >>   Add script to tweak output of 'alphapf' bibliography stile
> >>   alphapf.bst: Enable 'inlinelinks'
> >>
> >>  Makefile                    |  1 +
> >>  alphapf.bst                 |  2 +-
> >>  utilities/tweakhrefinbbl.pl | 37 +++++++++++++++++++++++++++++++++++++
> >>  3 files changed, 39 insertions(+), 1 deletion(-)
> >>  create mode 100644 utilities/tweakhrefinbbl.pl
> >>
> >> -- 
> >> 2.7.4
> >>
> >>>
> >>>                                                    Thanks, Akira
> >>>
> >>>>
> >>>> 							Thanx, Paul
> >>>>
> >>> [snip]
> >>>
> >>
> > 
> > 
>