* The chapter delimiter parameter of the \C tag was documented as \chapterdelimiter, not \chapter\Belimiter.



* New basic tag \C. The tag collects index entries of a chapter range (and all included subchapters) or the whole document. The stream contains all the tag options. Generator based converters can rely on additional informations, which include a list of all index entries in the chapter range specified, and a quick ranking - a unified list of indexed phrases and their occurence number.

* \C shows an example implementation for \\\INDEXCLOUD in generator based converters. As SDF cannot paint clouds, the tag is transformed into an ordered ranking list.

* \\\INDEXCLOUD implementations in traditional backend based converters are more expensive and depend on the converters internal index management and data structures, so they can vary. \pp2sdf shall show one possible implementation in future versions.

* New value for \C's \C option: \C works like \C, but embeds the file contents as a standard block (which is parsed), instead of as a verbatim block like \example. Parsing an example file is especially handy when working with (highlighting) paragraph filters.

* \ generator had not been adapted to several of the last stream format updates.

* Code passed to \C was not accepted if it evaluated to "false" in Perl. Now the valid "0" is handled correctly.





===Import filter API for the INCLUDE and EMBED tags

PerlPoint supports import filters for other formats since package version 0.38, via tag option \ifilter:

  \\INCLUDE{type=pp ifilter='pod2pp' file="file.pod"}

Nevertheless, this could be simplified. With the new option import this can be written as

  \\INCLUDE{type=pp import=pod file="file.pod"}

, given a module \PerlPoint::Import::POD is installed (note: see the \Pod::PerlPoint distribution for this module). The import module API is new and requires the module to define a function \importFilter(), with the same interface as functions for \ifilter. In fact, \import=pod is a shortcut for \ifilter="PerlPoint::Import::POD".

With \\\INCLUDE, the import option can use the \source file extension as the source format. This is enforced by using a true numeric value for this option, instead of a string:

  \\INCLUDE{type=pp import=1 file="file.pod"}

When both \import and \ifilter are used together \import is ignored.

===Automatic import of entire files

The new standard import filter API is used to allow automatic import of source files in other formats. Import is requested by the special prefix \IMPORT:, like in

  > perlpoint ... \B<IMPORT:>source.pod

With the prefix, PerlPoint treats the file extension (\pod) as the name of the source format, searches for the related import module (\PerlPoint::Import::\POD, again the format is all uppercased in the module name by convention) and invokes it automatically.

This feature applies to all existing converters automatically.


* new option \B-version for \perlpoint and \PerlPoint::Generator, thanks to Achim Grolms for the suggestion;

* the default file type for \\INCLUDE and the default language for \\\EMBED are \pp now;

* all generator based converters now accept the new option \-acceptedFormat, which specifies languages to be accepted in included files or embedded snippets (with \\\INCLUDE or \\\EMBED);

* parser: slight code optimizations;


The main intention of this update is to allow the production of better structured results (like correct XHTML). On the way to there, fixes and improvements were made.

There should be no incompatibility with 0.41.



* Tag authors can declare their tags as standalone now. If the parser finds a text paragraph is occupied by such a tag entirely, the paragraph "wrapper" is removed from the stream. This allows to produce better output for tags that are transformed into something that does not fit into a paragraph. (We had this before for images and a few other special tags, but now this is a general mechanism.) Standalone tags not used standalone have no special effect.

* Likewise, if all that is found in a paragraph is an embedded area (\\\EMBED{...} ... \\END_EMBED, the wrapping paragraph directives are removed from the stream.

* \IMAGE tags now get an \alt option by default unless they define their own (the default is "Image").

* \perlpoint now reads option files from the start directory, allowing to have a \.perlpoint file in a project tree.


* statistics for list shifters did not work;

* macro default parameters were not documented!


* dummy tokens inserted by the parser now are special strings that are filtered out by the backend module, the old string solution injected empty tokens into the stream;


This is a light update performed with the introduction of \PP::Template::TT2. No adaptations should be required except for the name change from \pp2tdo to \perlpoint.

As \C is intended to be one converter for many formats, it is now renamed to \C. It does \not mean other \pp2... converters are out of support in any way which they are not. The idea just came up when thinking about a binary package built with pp (from \PAR), for which a name of "perlpoint" seems appropriate and memorable.


* \PP::Generator provides a new option \-mainstream which allows to specify an alternative name for the document stream "main" (as a usage example, this helps to produce S5 slides which require DIV areas with certain names)

* main document stream now produces entry and leave point as well, allowing to format this stream a special way in the target format (again, S5 is an example of how this can be used)


* Generator classes did ignore the parsers nested table configuration, therefore no generator based converter could provide nested table support. Fixed.


* Updated \MANIFEST and \TODO.

* Prepared for use with \PP::Template::TT2.

* Took care to let the distribution pass more Kwalitee (CPANTS) tests.


This update has two parts.

First, it is a feature update. It has a few incompatible changes (see below), but should require only slight adaptations of existing converters. All converters should continue to run or should be adapted soon (as I know there is an adapted Converters package on the way).

Second, 0.40 introduces an extended framework, which in this release is in \beta state because of lacking tests and incomplete docs. Nevertheless, it is well tested in production and should be save to use.

Read on for all the details!



There are lots of new features, most of all an extended framework which should make it much easier to write converters.

The new parts are just \extensions of the "old" framework which remains in action (and is used by the new parts), so converters using the traditional interface should continue to run.

==Extended framework

This version introduces a new converter, \Bpp2tdo ("template driven output"), and a new part of the module framework, the \PerlPoint::Generator hierarchy, which implement a very generic and general converter approach.

See the new section "The formatter approach" in "Writing converters" for details about this new model.

Here are a few highlights:

* Object oriented extension design.

* Separation of general function, language specific actions and formatting.

* A clear, simple to use, entity based interface for subclassing of only a few methods to overwrite.

* All the basic work - dealing with parser, stream, backend and backend events - is already done by \PerlPoint::Generator.

* Easy to extend. Generator authors just add formatting.

* Every subclass can add additional options.

* Every subclass can add additional help portions.

* As formatting is determined by the formatter subclass, several templating systems can be used in parallel by just switching the template engine. A general template layer is provided.

* \All generators can use styles.

* General style format with parts for configuration, templates, docs and screenshots.

* Subclasses can be defined even in a style, so a style can come with its own formatters and template systems if necessary.

* Most of the docstreaming and index work is now done a central place, so all generators can use it and do not have to deal with it themselves.

* \The "old" approach is still available and supported, and existing converters will continue to run, as the old part of the framework is the base of the new part. The new generator part just adds an additional level of abstraction, making it easier to focus on the target format and layout.

===New SDF generator

As a reference implementation, the functionality of \pp2sdf was reimplemented in the generator model. To produce SDF this way, invoke \pp2tdo with option \-target SDF.

===First pure XML generator(s)

Well, we already had the demo converter that produced XML for PPresenter, but now a \common implementation is available, implemented with the new generator approach. To produce XML, invoke \pp2tdo with option \-target XML.

Best of all, if the result does not meet your needs, it is easy to control entity translation by options. If this is not sufficient, it is easy to write a modified converter by subclassing \PerlPoint::Generator::XML. See \PerlPoint::Generator::XML::Default for an example and "Writing converters" for instructions.

Various formatters are available to produce XHTML, paged XHTML and AxPoint. These are published as standalone CPAN packages, watch the PerlPoint namespace or the project side on SourceForge.

===Further reading

To learn more about the new features and generators, please read via \perldoc

* the docs of \pp2tdo;

* the \help of \pp2tdo, activated by option \-help, which adapts itself to the \-target, \-formatter and \-template options:

  Here are a few example calls (each call on one line):

  pp2tdo -target SDF
  pp2tdo -target XML
  pp2tdo -target XML -formatter XHTML
  pp2tdo -target XML -formatter XHTML::Paged
  pp2tdo -target XML -formatter XHTML::Paged -templatetype Traditional
         -templatesAccept XML/XHTML::Paged
  pp2tdo -target XML -formatter AxPoint

... and have a look at the example template provided in the \demos subtree.

==Incompatibel changes for users

* Empty (reset) variables will no longer be replaced by empty strings. This might become a problem - such variables were occasionally used to start generated text paragraphs, as by Pod::PerlPoint's pod2pp, which wrote:


    ${__pod2pp__empty__}Converted POD text ...

  To work around this, use the new optional dot start for text paragraphs:

    .Converted POD text ...

  And please install an updated version of Pod::PerlPoint, please ;-)

* The first chapter now \requires a headline. Comments and all paragraphs not producing visible content (like macro definitions, code fragments and variable asignments) can appear \before the first headline.

==Incompatible changes of the converter API

* Modified \anchor interface. Anchors now store the absolute number of the page they are defined in. As this is an internal interface, this does not affect document authors. On the other hand, converters will have to be adapted \if they make use of anchors (as far as I know, mostly converters of this package do, which \were adapted as necessary).

* \LOCALTOC added to the list of standalone tags (which are stripped of of an enveloping text paragraph if they are its only contents);

* \\\A and \\\F are basic tags now, which means they are available to all converters. The original meaning of "F" was \\font, but can be generalized now (to "formatting" or something like that). In fact, it formats a selected text, and it is up to the converters how to tranform the various options. (As with \\\FORMAT, conventions will help to establish documents that are portable between converters.)

* \B\\A now checks if it is the innermost tag/macro. This \can break existing sources.


Beginning with this release, a tutorial is part of the package. Being in an early state till it already describes the PerlPoint basics. Please have a look. Comments are appreciated.


* Anchor objects now can produce "arbitrary" anchor names on request.

* The \\\REF tag now provides the absolute chapter number of the related anchor, in the internal option \__chapter__.

* Backends can register a handler for \all directives (\DIRECTIVE_EVERY). This slows down operation, but allows central prehandling of all stream parts where necessary.

* Headlines provide additional data: their full and shortcut pathes as well as level and page number pathes, plus variables that are valid when the chapter is complete (for templates).

* New warning if the maximum columns number is detected in another line than the first table line (which is the base of normalization).

* New parser option \criticalSemanticErrors treats semantic errors as critical, causing the parser to terminate immediately (as after syntactic errors).

* New parser option \skipcomments excludes comments from the stream.

* Input filters can access the source file by a variable \$main::_ifilterFile now.

* \Tag hooks take (provide) an additional parameter: the absolute number of the page the tag is used on.

* Table directives provide additional hints: \__maxColumns__ and \__titleColumns__.

* New directive DIRECTIVE_DPOINT_TEXT now encloses definition point explanations in the stream.

* Text paragraphs can be started by an (optional!) dot now:

   .Text begins.

  This is mostly useful when \I<generating> PerlPoint from other sources,
  to make sure the first characters of the converted text will not be interpreted
  as special PerlPoint characters.

* As Clinton Pierce's Perl Projector seems no longer to be supported, the demo converter \pp2cppp and the module \PerlPoint::Tags::CPPP were removed from the distribution.


* A subtile bug could cause parser crashes when an included document contained just a macro without trailing newlines, in a deeper nesting level. The parser now behaves more robust in this area. Thanks to Heike Metz who reported the crash.

* Docstream "main" was ignored like any other docstream if working in the "docstream ignore" mode.

* Undefind variables were expanded. Fixed. Note: this might turn out to be an incompatible change for you, if you unset an empty variable in order to get an empty string (like Pod::PerlPoint did):


    ${empty}Text starts here ...

  To work around this, use the new optional dot start for text paragraphs instead:

    .Text starts here ...

* There was a general error in variable handling, which caused wrong resolving of backslash preceeded variables in several cases. Thanks to Lorenz for the bug report.

* When continuing an ordered list after returning from a nested list, the continued list was treated as "new". This was not really a bug as it was treated as a new list, but continuation would have been more intuitive - now it is there.

* The \alt option of \C can handle backslashed commata now - it was not possible before to use headlines with commata as alternatives.


This update moves a tag (\\\X) from converter definitions to the basic tags. To make use of the new functionality, you need to update to adapted converters. Alternativelely and as a workaround, go to the definition lib, e.g. \PerlPoint::Tags::HTML, and delete the definition of the \\\X tag. (Please note that full support of the new index features \needs updated converters.)


* Guarded variables were \expanded if set up. Oops. Fixed, test added.

* \pp2sdf did not print tables ...

* The parser now automatically adds \all composite anchors for a headline, not only for the full headline path. This allows to use partial links like \C|D|E or \D|E for a headline with the full path \A|B|C|D|E.


* \B\\X became a base tag like it is in POD. Many thanks to Lorenz who implanted it into PerlPoint with his converters. The tag is defined in \PerlPoint::Tags::Basic now. Converters still defining it by their own should be adapted (unless they want to overwrite the basic definition).

* \B\\X now checks if it is the innermost tag/macro.

* New basic tag \BINDEX provides a full index structure. See \PerlPoint::Tags::Basic for details.

* Adapted perlpoint-hilit19.lisp to Emacs 21.x. Unfortunately it enforces a warning, but nevertheless does what expected. Might be someone with Emacs Lisp experience can fix this?

* Parser change to run under perl 5.8 without errors: \no utf8. Does it work under older perls as well?


===Index based references

Introducing new "index based references". What's this? Imagine a document collection by many authors. It would be nice to cross link related articles. This requires knowledge of all the document parts and would usually be arranged by an informed document manager, or by using technologies like Topic Maps or the like. In any case, it takes time.

But what if the collection is frequently updated, or when it is impossible to instrument them for Topic Maps or similar technologies, or if there is neither person nor time to investigate all the stuff in detail? Then you could use this (still experimental) feature of \index based references.

The idea is that index entries point to documents/chapters with substantial information about the index entry. Documents/chapters matching in their index entries are probably content related. In linking to other documents or chapters speaking about the same indexed issues, or to a relevant percentage of the base documents (indexed) issues, one can build cross references automatically.

To do so, the package now provides a new base tag \BINDEXRELATIONS which builds a list of related pages. Options specify if the keyword base should be filled with entries from the start chapter only, or shall include all subchapters. Likewise, one can specify if only the index entries at other \startup pages should be taken into account, or if \all their subchapters should be scanned, too. Finally, there's an option to configure a minimal match threshold (absolutely or by percentage).

Converters can take the provided chapter list and present them as they like.

The appropriate tests and docs were added.


* The documentation now mentions explicitly that list indentation is reset \automatically by any subsequent non list paragraph. Thanks to Marcus Holland-Moritz for suggesting this.


This is maintenance update for paragraph filters. If you are using paragraph filters, it is strongly recommended to upgrade.

But hopefully the new features make it worth to update for others, too ;-) The POD input feature, for example. Or any language you like. Read below.

There are no incompatibilities between \REF{name="0.37"}<0.37> and 0.38. All PerlPoint sources which passed 0.37 should pass 0.38 as well. All converters should continue to work.


* Paragraph filters did remove guarded backslashes (\\\\\) and guarded ">" characters (\\\>). Improved.

* Paragraph filters restored the original source with bodies for bodyless tags, fixed.

* Paragraph filters did not reinsert all backslashes (they handled the first occurence only) when restoring the original source.

* Paragraph filters inserted backslashes into verbatim paragraphs when restoring the original source (thanks to Lorenz for the report).

* Empty headlines could cause errors and infinite parser loops.

* \pp2sdf: removed occasional warnings displayed without real reason. Fixed a \splice() usage bug.


* Tags are streamed with an additional information. Both opening and closing directives now provide a body hint telling a backend programmer if the tag has a body or not.

* The lexer makes use of precompiled patterns now, but it seems not to affect the performance of parsing.

* \\\INCLUDE, \\\EMBED and \\\TABLE now support the \_cnd_ option, like tags defined externally.

* Embedded parts and included files can be preprocessed by the new "input filters" - snippets of PerlPoint author defined Perl code transforming the included or embedded parts before they are processed as usual. This is a generic implementation of sourcefile import.

  For example, it is possible to process POD sources
  directly now. Here is a simple filter that does the job:


  use Pod::PerlPoint 0.03;

  sub pod2pp
    my ($pod2pp, $result)=(new Pod::PerlPoint());


* Or try the new demo source file, \demo/perlpods.pp, with your favourite PerlPoint converter. (Make sure the cache is deactivated and you are using \-active -safe ALL. Depending on your system, this is a CPU, memory and time consuming job producing over 12.000 pages when processed with perl 5.8.0.)

* Demo converters \pp2pod, \pp2pp and \pp2tree got the usual interface (supporting traces etc.).

* New trace switch \\ avoids removal of temporary files (which are new as well - and written when on includes files via input filters).


* Brackets (\[]) seem to have a special meaning in SDF: \sdf evaluates their contents by \eval(), which causes a lot of confusion in (Perl) example paragraphs - as a workaround, all opening brackets in blocks are now written as \\\[.

* Anchor and links names (in \id= and \jump= specifications) are enclosed by Perls \q() operator now, in the hope to avoid trouble with quotes in the anchor. (But what about parantheses in an anchor name now? It seems to work.)


This is a maintenance update. It can replace version 0.36 transparently except for verbatim block paragraph filters and \pp2sdf output handling. See next section for details.

==Incompatible changes

* Paragraph filters intended to be applied to verbatim paragraphs need to be adapted: they now receive the \full paragraph text \including the enclosing heredoc parts (e.g. \<<EOE and matching \EOE).

* Extended headline stream interface. This only effects converter authors, but not really, because there are just two \additional parameters now. See below.

* \Bpp2sdf now needs an explicit result file setup by the new option \-sdffile. Previous versions wrote to \STDOUT.


* \pp2sdf: \B\\SEQ did not set an anchor when the \name option was specified.

* A combination of \\\REF, \\\SEQ and activated cache caused trouble illustrating open tasks in the implementation both of finish hooks and anchor objects. This is fixed, see \REF{type=linked name="0.37 | Features | Cache"} below for details.

* For grammatical reasons, paragraph filters applied to blocks or lists could cause trouble in detection of the subsequent paragraph. This is fixed.

* Filtered paragraphs were cached but are made potentially dynamic in an unpredictable way by the filter(s). So they are cached no longer.

* Filtering of headlines and verbatim blocks was not really fully supported yet. Improved.



===Document streams

Robert Inder asked how to achieve an HTML layout where notes are placed below the slide context. When we discussed several ideas, I thought it might be useful to have a general solution which I called \"document streams". Thanks, Robert!

So what is a document stream? Consider the following layout:

  |         header navigation         |
  |                                   |
  |              Title                |
  |                                   |
  | Introduction text, points etc.    |
  |                                   |
  |                 |                 |
  | Here we are     | Here we are     |
  | talking about   | talking about   |
  | subtheme 1.     | subtheme 2.     |
  |                 |                 |
  |               notes               |
  |         trailer navigation        |

The kernel parts can be easily generalized into a more abstract scheme:

  |                                   |
  |            main stream            |
  |                                   |
  |                 |                 |
  |  item 1 stream  |  item 2 stream  |
  |                 |                 |
  |              stream 3             |

This may be used in many ways - to compare software or hardware items, to place reader hints besides a main text, to write about the lifes of two or more individuals and present certain life parts in parallel, and so on.

But how to distinguish document streams in a PerlPoint document? Well, here are the simple rules:

* Every headline (re)enters the main stream.

* \Stream entry points enter certain streams. They are a new one line paragraph type, beginning with a \B~ character and ending with the line (like comments do):

  =Main stream

  Bla bla

  \B<~The first docstream>

  Blu blu

  \B<~The 2nd doc stream>

  Bli bli

  ==Back to main stream

Converters need to add docstream support, so currently only \pp2sdf provides a first (reference) implementation. \It is completely up to a converter how it supports and "translates" document streams. But to implement generally useful methods in the framework, the parser can be set up to transform all docstream entry points into headlines (at a sublevel of the current headline level) or to ignore all document streams at all (except of the main stream).

More, docstreams can be skipped more selectively as well: the parsers \run() method provides a new parameter \Bdocstreams2skip to pass the parts to be ignored. When an ignored docstream is entered, all subsequent document parts till the next entry point (or headline) are completely ignored (including conditions - take care).

Finally, \headline start directives now provide a list of docstreams used in this certain chapter - allowing converter authors to adapt the chapters layout dynamically. See "Writing converters" for details.


Now it's easy to use a central repository of (macro / variable / code / filter / document) libraries:

* \B\\INCLUDE now searches pathes specified by an environment variable \BPERLPOINTLIB (similar to \perl, shells, linkers etc.) if the specified file cannot be found directly. In the tradition of \PATH and \PERL5LIB, pathes are delimited by (optional whitespaces and) semicolons.

  path1\B<;>path2\B<;> path3   \B<;>   path4\B<;>path5

* The new PERLPOINTLIB path feature is available immediately with this package version, regardless of converter adaptations.

* Converters can pass a new parameter \libpath to the parsers method \run. The argument of \libpath should be a reference to an array filled by converter options. I suggest a new conventional option \\-includelib to be invoked multiple times (in the tradition of \-I options available for \perl and various compilers).

* \-includelib needs converter adaptations, so please use \PERLPOINTLIB till updated converters appear.

Pathes specified via \-includelib (or the internal representation \libpath, respectively) are searched \before pathes specified via \PERLPOINTLIB.

* Pathes of included files are resolved now - circular inclusions caused by links pointing to the same file can be detected earlier this way. As a side effect, the resulting absolute file names improve error messages.


* Paragraphs containing tags using finish hooks are no longer cached. A tag finished after parsing makes the whole paragraph depending on something that cannot be cached on base of the parsed text.

* Anchors defined by a cached paragraph are cached now as well - and restored after a cache hit.

* The implementation of the mentioned last point required an extension of the internal cache format. Cache files made by versions prior 0.37 will be rebuilt automatically next time they are used.


* Avoiding more backslashes in paragraphs starting with an SDF command.

* New option \B-result to specify an output file.

* Supports document streams by three modes: docstreams can be ignored, entry points can be made headlines, or \pp2sdf can produce one document per docstream.

* Provides new conventional converter options \B-includelib, \B-docstreaming and \B-skipstream.


* Slightly improved the \README file.

* \BflagSet() now takes a \list of flag names and checks if at least one of the flags is set (implicit \or condition). This makes it more handy to write combined conditions. Thanks to Helmut Steinbach to make me thinking about this.


  ? flagSet('a') or flagSet('b') or flagSet('all')


  ? flagSet(qw(a b all))

* \PerlPoint::Anchors was extended by a logging mode to trace newly added anchors.

* Added tests of condition interface functions (currently \flagSet() and \varValue()).

* Improved headline tests.

* Added paragraph filter tests.

* Paragraph filter functions can access the type of the paragraph using the new variable $main::_pfilterType.

* Empty text paragraphs are no longer made part of the stream.

* Blocks were streamed with a final newline, improved.

* Demos \pp2pod and \pp2tree auto-anchor headlines now (so \\\REF can be used).

* New headline shortcuts. Thanks to Robert Inder who asked for them. A shortcut is a plain string appended to the headline title, separated by a "~". \Converters need to be adapted - currently there is no shortcut support available.

  =A very long headline which may not fit
   into the navigation bars \B<~ \I<Long headline>>

* Updated \hilit19 extensions for Emacs (in the \utilities directory).

* Updated docs.


This is a major update - in the sense that it modifies a lot of internals, especially stream handling. Nevertheless, it's almost completely compatible with earlier versions of the package (and the few \REF{name="Summary: Incompatible changes" type=linked}<incompatible changes> most likely do \not affect existing converters).

So \this version should run whereever 0.35 could be used.

While working on this version, I began to make use of the modified internals myself. As a result, there are various new \visible features as well. See section "\REF{name="Visible Changes" type=linked}" below for all details. Internals are described in chapter "\REF{name=Internals type=linked}".

\Please note that there's a new dependancy to run the installation tests because I switched to \BTest::More (included in CPAN package \Test::Simple - do \not install the still available standalone distribution of \Test::More). You may have to update \Test::Harness as well.


* Several very slight fixes in calls of \PerlPoint::Parser::run() (these bugs caused no errors).

* \pp2sdf embedded native SDF incorrectly into the produced result. Thanks to Lorenz for the bug report.

* \pp2sdf: unintended SDF recognitions caused by ">" characters are now avoided by generating ">" as "{{CHAR:gt}}" (outside verbatim blocks).

* If the only contents of a table cell was a \0, this number was removed. Fixed.

* If a macro overwrote a tag with the same name and a (parsing) hook, the tag hook was invoked when the macro had been used.

* Tag parsing hooks received tag options instead of the tag body.

* Helmut Steinbach found that the closing sequences of verbatim blocks made by Active Content were not recognized correctly. This could be fixed.

* Investigating the previous point I found that is was also impossible to dynamically make a usual example block or a paragraph that starts with a tag or macro. This is now fixed as well.

* Internal bugfix in macro body reparsing. The bug could cause infinite parsing loops.


This chapter can be safely skipped by "pure" PerlPoint users which might like to continue with chapter "\REF{name="0.36 | Visible Changes" type=linked}".

This chapter describes these news:


===PerlPoint::Backend, stream format

\\The format of the intermediate stream was changed. I do not know of any direct stream access in a converter, but \if anyone did access the stream directly, he will need to adapt his code. Code using \PerlPoint::Backend does \not need to be adapted.

The stream is still a stream basically, but with additional structural hints stored in parallel, which are currently a stream of headline index numbers. This way I hope to stay with the performance advantages of a stream while enabling things like chapter inspection and arbitrary chapter navigation.

For example, \pp2html 0.11 in fact needs to process the stream twice because it needs the whole picture of the headline structure when building slides. Now it would be possible to inspect headlines in the first pass \only, and to process \all tokens in the 2nd pass then. This is a common task - \pp2html might be modified or not, but future converters can definitely make use of this. (And there's an even better and faster way now to grab the TOC, see below.)

To make use of the new structure, \PerlPoint::Backend provides several new methods:

* \run() now works in \modes. By default, it processes all tokens as usual. This is called the "token mode". In \headline mode, it processes \headlines only. Switch modes by \Bmode() - even from within a callback!

* Navigate within the stream: \Breset() goes back to the very beginning, while \Bmove2chapter() allows you to jump to a certain chapter.

* Get the number of all headlines by \BheadlineNr(), and the number of the current chapter by \BcurrentChapterNr().

* The stream data structure is now bound to a backend object while \run() is running. If navigation methods, \headlineNr() or \currentChapterNt() shall be called outside \run() - which means outside a callback - \Bbind() can be used to perform this binding manually.

* Instead of using \run(), converter authors may prefer to process the stream step by step having full control (e.g. to use the stream while presenting instead of in format translation). This is possible by using \bind(), sequential calls of \Bnext() and finally \Bunbind().

The new backend method \Btoc() uses the new features to provide a table of contents. If called for a certain chapter, all subchapters are listed. The reported depth can be limited.

  # get the complete \I<table of contents>

  # get a list of all \I<current subchapters>

  # get a list of the subchapters \I<one level below the
  # current one> (do not list the complete tree, if there
  # are more levels)
  $subchapters=$backend->toc($backend->currentChapterNr, \B<1>);

All backend extensions are documented both in \PerlPoint::Backend's POD and in the converter tutorial. Hopefully.

===More stream enhancements

* Headline start directives provide the full headline title (plained, which means without possibly used tags) as new data.

* Unordered and description list directives now reserve a parameter at the position where \ordered list directives traditionally pass a list startup level. This parameter is currently unused.

* All list directives provide four new additional parameters to give converter authors a hint about subsequent or preceeding \list shifts. For reasons of stream construction, hints about \preceeding level shifts are provided in list \start directives \only, while hints about \subsequent level shifts are provided in list \completion directives \only. (But these are exactly the places where these informations can help, so symmetry is the only thing lost here.) This feature is intended to help a converter to build really \nested lists. (In PerlPoint, a list level shift implicitly closes a previous list, but this is different in various target languages.)

* Parser and backend now interact via directive hints. This is an internal interface. For example, it is used to flag that a tag and / or its contents should be skipped (see below).

===New finish hook to complete tags

With the new tag declarations by module introduced by version \REF{name="0.34" type=linked}, it became possible to hook into tag parsing. Now there's another hook to \complete a tag \after parsing - this means, when all input informations are available.

 The new \C<\\REF> macro makes use of this feature
 to verify links and to complete itself by linked content.

The new hook interface is quite similar to parsing hooks except that it does not provide informations only available at parsing time (like the tag body). See \PerlPoint::Tags for a detailed doc.

===New hook return codes PARSING_IGNORE and PARSING_ERASE

Tag \parsing hooks can return two new codes:

* \BPARSING_IGNORE causes the parser to ignore the tag. The result is just the body.

* \BPARSING_ERASE removes the tag \completely which means together with all its content. See the new basic tag \\HIDE for an example.

Finish hooks can return these codes as well, with exactly the same effect. (The tag remains streamed (as well as its body), but will be bypassed by the backend.) See \PerlPoint::Tags for a detailed doc.

===Hooks can now be interchanged

It is now possible to use hooks of a "foreign" tag (declared by another tag module) in your own tag hooks. This helps to emulate other tags, for example if an \old interface (tag and option names) shall be preserved but the \new functionality shall be used. To invoke a foreign hook, call \PerlPoint::Tags::call() anywhere in your own hook, passing tag name, hook type and parameters, like so:

 $rc=PerlPoint::Tags::call('TAG', 'hook', @_);

If the tag is not registered, or has no hook of the specified type, an undefined value is supplied, otherwise the return code of the invoked function.

===New anchor management

The parser can relieve converters from anchor management partially now. This is an \offer.

* A new minimal class \BPerlPoint::Anchors provides basic anchor collection services. Addresses and related values can be added and queried.

* The parser maintains a \PerlPoint::Anchors object.

* On request, headline titles (stripped off of tags) are automatically added to the parsers anchor collection. (This is inspired by the automatic headline anchors idea by Lorenz, built into \pp2html for versions and already imitated by \pp2sdf. Why not make it a general feature?) To activate this feature call \run() with \headlineLinks set to a true value.

* The parsers anchor collection object is passed to all tag hooks (both parsing and finishing ones), where the collection can be extended (most probably by parsing hooks) or queried (supposedly by finish hooks).

This is the complete management for now. The parsers collection is \not made part of the stream. It's just a help to verify links at parsing time, and to insert named values into the stream.

  The new \C<\\REF> macro uses the parsers
  anchor collection to resolve values not
  available at parsing time, like sequence
  numbers generated by the new \C<\\SEQ> tag.

Again: with this feature (and tag finish hooks), it becomes easy to reference strings defined \later than the reference.

===Summary: Incompatible changes

Various incompatible modifications might have been mentioned before, but it might be useful to have a check list:

* The stream format was \significantly extended, both in structur and semantics. \PerlPoint::Backend manages this for you. (And you already used it, didn't you? ;-)

* List shift paragraphs are no longer streamed by opening \and closing directives. There's only an \opening directive from now on.

* Tag (parsing) hooks have a modified interface: the tag body is now passed by reference, no longer as a list. This was necessary to add new parameters to the interface.

* Tag completion directives in the stream no longer provide exactly the same informations as corresponding start directives, because the new finish hooks are only invoked for startup directives. So, if a tag finish hook modifies tag options, this will affect the startup directive only. (Let me know if this causes problems.)

* Likewise, headline start directives in the stream now provide more informations than related completion directives: the full (plained) headline title is only provided with the startup hint.

* \\\TABLE tags are streamed with additional informations about the separators used (options \separator and options \rowseparator). If they were made from a table \paragraph, there's an additional option \_paragraph_ set to 1. These informations are intended to be used when a paragraph is retranslated for a \REF{name="New type of active contents: paragraph filters" type=linked}<paragraph filter>.

==Visible Changes

While the modified internals are especially of converter authors interest and do not affect existing documents, there are already \new features in this version implemented on base of the new internals which \might find your interest and help you in writing documents. These new features include:


Please see the \REF{name="0.36 | Misc" type=linked}<"Misc"> section below for more visible changes which are not based on the modified internals.

===More active content: tags and macros can be made optional

If a tag or macro has an option \B_cnd_, the parser treats this option as a \condition to activate the tag/macro or not. The condition is evaluated as Active Content. If Active Content is disabled, the condition defaults to "false".

  \\IMAGE{_cnd_="$illustrate" src="piegraph.gif"}

Unfortunately, because it is done via option, only tags and macros \providing options can be made conditional. So,

  \\I{_cnd_="$italics"}<conditionally italic text>

cannot be written because of a parsing error. That's bad. If anyone has an idea for a well fitting general solution, please let me know.

(Hint for converter authors: the condition option is stripped off of the options hash if the condition (and therefore the tag) turns out to be valid. You cannot evaluate it yourself again.)

===New basic tag \\LOCALTOC

\B\\LOCALTOC inserts all subsections of the current chapter (in requested depth). Simply say:

  =Parent chapter

  In this chapter:


  ==Subchapter 1

  ===Subchapter 1.1

  ==Subchapter 2

, and a converter will produce results according to \this source:

  =Parent chapter

  In this chapter:

  \B<* Subchapter 1>

  \B<* Subchapter 1.1>

  \B<* Subchapter 2>

  ==Subchapter 1

  ==Subchapter 2

I missed something like this tag for a long time, indeed. Think of all the empty pages produced by headlines which just collect various subchapters, but without own text.

It is possible to limit the subhierarchy depth taken into account (suppress the display of subchapter 1.1 in the example above by setting \depth to 1), to format the produced list in various ways (ordered, unordered, with chapter numbers) and even to make the chapter titles hyperlinks to their related chapters (if supported by the target format).

The transformation has to be done \by the converters. As I know, the tag shall be supported by \\PP::Converters 0.12. For now, you may get an impression by using \pp2sdf which as a reference implementation is already supporting \\\LOCALTOC.

===New basic tag \\SEQ

Generates the next value of a certain sequence "type" which is declared by option:

  \\SEQ{type=example}, \\SEQ{type=example}, \\SEQ{type=example}

will produce \1, 2, 3, while

  \\SEQ{type=example}, \\SEQ{type=image}, \\SEQ{type=table}

results in \1, 1, 1 (provided no sequence was started before).

It is possible to name the generated number, which makes it easy to reference it. Naming is done by option \Bname.

  \SEQ{type=image name="Blue water"}

(Note: Converter authors have to deal with this tag a special way because the number is passed by a generated tag \option, not by tag body. Nevertheless, it's quite simple, see \pp2html for an example and "Writing converters" for a description.)

===New basic tag \\REF

This is intended to be a very general reference tag. We already have several reference tags, mostly introduced by \pp2html - this one is intended to continue their tradition a generalized (and generally available) way and is implemented on base of new parser features (see \REF{name="0.36 | Internals | New anchor management" type=linked}<"anchor management"> above).

There's a complete doc. Here's an overview:

* At least one option is required, which is \name. It specifies what shall be referenced.

* Option \type to say if the result should be made a hyperlink or not. Defaults to \"plain" (text) which means that the value of the referenced object is replied (think of a named sequence number).

* The body is optional in general, but forbidden in conjunction with a \"plain" result formatting (if a body is specified, there's no room for the referenced text).

* Alternative references can be specified via an optional \alt option, delimited by commata. (This delimiter might not be optimal because commata can be used in headlines, I'm open to better ideas.)

  \\REF{name=A alt="B, C, D"}

* The final link is checked by the parser.

* Optional option \occasion (set to a true value) specifies that a missing link is no error. In this case, the tag will be ignored.

===New basic tag \\FORMAT

This is another generalized tag, this time in the tradition of \pp2html's tags which configure the rendering of examples. \\\FORMAT is just a container which can be filled by various options which represent the settings to made.


This should make all following text paragraphs and table cells left justified, until the end of the document or another alignment setting is made. So the meaning of \\\FORMAT is to configure \default formattings.

As formatting is very target language specific, each converter can define valid settings it accepts and evaluates. \\FORMAT may then be used to make settings interpreted by \various converters \at once:

  // this example uses imaginary settings
  \\FORMAT{\B<html_xxx=2 sdf_xxx=5>}

Nevertheless, certain configurations may have a common meaning. I encourage converter authors to establish conventions. The first conventional setting established is \align which shall accept the values \left, \center, \justify and \right.

This tag was invented after Carlos Malvar suggested we could offer justification. Thanks, Carlos.

===New basic tag \\HIDE

This tag removes all its contents. This makes sense if used together with a tag condition.


===New basic tag \\STOP

Another tag that makes most sense if used with tag conditions. It raises a syntactical error and thereby stops document processing immediately. So, to check if the used parser version meets the needs of your up to date document, write


===New type of active contents: paragraph filters

Carlos Malvar asked me for automatic example highlightning. Thinking about how users could afford automatic highlightning without having to support various example languages by PerlPoint directly, I thought a general solution might meet such needs. So now users can declare \paragraph filters which means they declare that certain paragraphs should be passed to a specified piece of code when parsed. This is, obviously, Active Content. The filters results must be valid PerlPoint which will then be reparsed.

So finally those filters are there. Here is a complete example to demonstrate how this feature can be used. The example document contains a verbatim block of Perl code, which is made a block by a user filter. The simple filter marks all Perl comments bold (using a straight forward comment detection).

  // define the paragraph filter

  sub formatComments
    # $main::_pfilterText is used to
    # pass the paragraph content


  Now, an example in a verbatim block paragraph:


    # 1st comment
    This is my example. # 2nd comment


    # another comment


  Example completed.

Please note the preamble which installs the filter to use.

Here is another filter that automatically numbers example lines:

  sub filter2
    # init line counter
    my $c;

    # transform example text
    my $t=$main::_pfilterText;
    $t=~s/^([ \t]*)/join('', ' ', ++$c, ": $1")/mge;

    # provide result

Filters can be chained by using a single pipe character:




To try this successfully, Active Content has to be enabled (converter option \-active). If Active Content is disabled, filters cannot be installed. (Copy the example from a converters result, not from this Changelog source to avoid backslash trouble.)

Paragraph filters can be assigned to almost all types of paragraphs, not only verbatim blocks. But because they work after a paragraph is processed, they make no sense for paragraphs which do not produce output directly. So filters are not allowed for conditions, alias definitions and variable assignments (for grammatical reasons filters \can be assigned to the last mentioned paragraph type, but I hope to fix this in future releases).

There is no loop check currently. If a filter produces a filtered paragraph, things might end in an endless loop.

As always, feedback is highly appreciated. Documentation and tests are planned to be added.


* Supports the new features.

* "{" characters outside verbatim blocks and embedded HTML or SDF are now replaced by SDF escape \{{CHAR:lbrace}} to avoid SDF confusion by unbalanced braces;

==New demo converter pp2ppresenter

The first XML converter is here, producing input for PPresenter's XML interface. PPresenter is a Perl/Tk presentation software by Mark Overmeer, with an XML interface by Johan Vromans. See ppresenter.org for details about this software.

This demo is not complete yet, but I hope Mark will take it over ;-)

The XML generation is by no means perfect, but a first proof of XML generation from PerlPoint. Suggestions are very welcome, as there are various more presenters using XML.

As with all demos feel free to make this a starting point of an own converter. Just let us know.


* "Writing converters" is up to date now, including a reference of the backend callback interface.

* Added \\\LOCALTOC, \\\SEQ, \\\REF and \\\HIDE tag docs. Updated other docs where appropriate.


* Improved error messages visualize where an error was detected.

  [Error] test.pp, line 6: found " ", expected:

          \B<org.: \\IM{src="blibblub.gif"}
          exp.: \\B<\\C<\\I<\\IMAGE{src "blibblub.gif"}\>\>\>

* Removed \Safe invokation in installation tests where it was unnecessary.

* There's a new dependancy to run the installation tests because I switched to \BTest::More (included in CPAN package \Test::Simple).

* Helmut Steinbach suggested that macro definitions could declare parameters. That's implemented now. To declare defaults, simply add an option list to the macros name:

  +MACRO\B<{value="default value"}>:value is __value__

  Now I'm using the macro the usual way, \MACRO{value=passed}.
  And the new way: \MACRO.

* Several modules were stored in DOS format (with \r\n line endings). This confused perldoc on Solaris and is fixed. Thanks to Lorenz who detected it.


This is a minor update, but it is strongly recommended. You will especially enjoy it in case you are running perl 5.00503 or Windows (or both ;-).


* The lexer did not combine plain strings as completely as possible because of a pattern bug.

* \pp2pod now completes examples ((verbatim) blocks) by additional empty lines.


* Adapted to perl 5.005 (once more ;-). Please note that under perl 5.005, parent classes need to be loaded explicitly, so \to make your converter run under this perl version please include a \use PerlPoint::Tags; statement. See the demos for examples.

* The package now passes all tests under Windows as well. They failed because files need to be prepared by \binmode() under DOS to make \seek() work correctly.

* New demo application \pp2cppp converts PerlPoint into Clinton Pierce's PerlProjector format.


* New intro \"Getting started (with PerlPoint)".

* Now doc \"How to write a PerlPoint converter".

* Providing a POD version of "Getting started". I'd like to add a POD version of the converter tutorial as well, but unfortunatley there are limitations because POD (as I know it) does support neither tables nor headlines of a level greater than 2.

* Slightly rearranged existing docs.


\This is a major release. It needs adapted translators. Old fashioned converters will stop working after installation of package version 0.34.


* Lorenz detected that \B0's were not allowed in macro names: fixed (this integrates patch \0.3301);

* Verbatim blocks were documented as something left completely untouched by PerlPoint but variables \were replaced, fixed implicitly with new line reading behaviour (see below).

* Multiline tags in examples are correct PerlPoint but were mistranslated into equivalently multilined SDF tags. Because SDF tags cannot enclose line breaks pp2sdf needs to automatically \close open tags at the end of an example line and to \reopen them in the subsequent line, which is arranged now.

  \I<The example code>

      My tag \\I<encloses
      line breaks>.

  \I<formerly transformed into>

  E:  My tag {{I:encloses
  E:  line breaks}}.

  \I<is now made the valid SDF code>

  E:  My tag {{I:encloses\B<}}>
  E:  \B<{{I:>line breaks}}.

* A bug in an internal stack handling caused several operations to be performed when a macro was replaced, including line number updates and removal of whitespaces following the macro (context dependend). The fix "saves" such whitespaces and accelerates processing, especially if macros are used extensively in a document processed with activated cache.

* Internal table data were not completely reset at the end of a table, thus causing addition of incorrect elements to the stream representation of subsequent tables. Fixed.

* Backslashes in conditions were removed, fixed.

* One column tables were handled incorrectly.

* Headlines couldn't begin with a character that opens a paragraph.


===Tags and macros

* \New tag declaration by modules: options and body can be declared optional, mandatory or disabled, and this influences the parser in searching for parts or not, checking mandatory parts, and invoking user (translator/tag author!) declared functions to check options or body more detailed. Please have a look at the documentation of the new module \PerlPoint::Tags for all the details. This modification improves tag parsing significantly, making it much more flexible. Additionally, there's now a convenient way to share tag definitions between converters and to allow users to use tags unsupported by the current converter but accepted by another one. See next point.

* Translators not declaring tags as mentioned above can be forced by the user to \accept foreign tags syntactically. As a side effect, the experimental \\\ACCEPT_ALL became obsolete and is no longer supported.

* Macros are adapted to the new tag evaluation: they as well are now evaluated depending on their definitions: the parser will no longer recognize an option part unless the definition makes use of parameters, likewise a body is only detected if it was mentioned in the macro definition by \__body__. This rule makes macro usage more intuative and allows simpler embedding into the documents text.

* New basic tag \\\READY uses the new definition features to complete parsing immediately. This means that remaining parts of a document will be \ignored. This can be useful if only parts at the beginning of a document are of interest in certain circumstances:

   ? flagSet(overviewMode)

   \\B<Short Overview>

   This document will give you an impression.

   // ready in overview mode

   ? 1

   // usual document


* Modified lexing to recognize the longest "words" possible. Previous versions supplied real words or even single characters. The "longest possible word" is the complete string before the next character (or string) with a special meaning in the current context. This change accelerates parsing incredibly: there are less lexer calls and less parser states to walk through (and this state machine is by far the slowest part in source processing). Parsing is speed up by about five or seven times, depending on a documents structure. This acceleration is even more impressive if only a few parts were changed and the cache is in use. More, because there are less parts in the stream now, backend processing is significantly accelerated as well!

* Accelerated comment and verbatim block reading: these paragraphs were lexed and parsed for no purpose, so now they are read and streamed line by line.

* Accelerated the parser by replacing a clean object method calls of the used Parse::Yapp parser object by accessing the objects internal USER hash entry directly. This of course breaks encapsulation but is suggested by the Parse::Yapp manual for reasons of efficiency.


* Improved cache: the usage of variables and macros does no longer prevent a paragraph from being cached, so the cache becomes more effective. In our real life documents, the cache hit rate is increased from about 50 or 60% to about 75%. The cache detects a paragraph hit if the paragraph checksum matches \and the new variables / macro definition checksum match as well. Thanks to \Stefan Sautter who suggested this improvement.

* More cache improvements: until now, it was a little bit dangerous to use something looking like a tag or macro but being none of them, at least if the cache was activated. The cache detected that there was no macro and stored a checksum for the paragraph. Everything worked well until this macro really was defined. Now it would have to be replaced where it was used! But because there was no change in the checksum, and there was no macro in the paragraph before, the cache detected a hit, reusing the \old paragraph stream. So in fact, the new macro definition seemed to be ignored in the paragraph until the cache was disabled or rebuilt. This is all history now - if something looks like a macro in a paragraph, the cache will activate the new macro checksum feature, so whenever a used nonmacro becomes a macro there will be no cache hit after that change (but of course, the \new stream is cached, so the next parsing will profit from the cache again).

* Even more cache improvements: the cache now stores the parser version which built it and the constant declarations version which was used. Likewise, the version of the used Storable module is saved. These data are used to automatically rebuild a cache after cache structure changes. (To demonstrate this feature, version 0.34 introduces an incompatible, modified cache structure and all existing cache files will be rebuilt automatically. Well, or because of the cache improvements described above. ;-)


* Extended the documentation building set by adding detailed tag documentation. As usual, this is intended to make it easier for translator authors to document their tool and can be integrated into and adapted to their documents. Second, these short documents are hopefully used as templates to document all PerlPoint parts a consistent way.

* Added a portability discussion and more to the FAQ.

* Macros are no longer called experimental.

* It is now documented that macros overwrite tags of the same name.

* It is now documented that "empty lines" containing whitespaces are not recognized as empty by Perls paragraph mode which makes the cache fail especially in examples (subsequent example parts "disappear"). Maybe future versions can improve this by either implementing an own "paragraph mode" or handling example caching differently.


* Table paragraphs are now "normalized": if a table row contains less columns than the headline row, the missed columns are automatically added (this helps converters to detect empty columns).

  A | B | C
  1 |
  1 | 2
  1 | 2 |
  1 | 2 | 3

  is streamed now exactly as

  A | B | C
  1 |   |
  1 |   |
  1 | 2 |
  1 | 2 |
  1 | 2 | 3

* Tables built by tag (\\\TABLE, \\END_TABLE) are normalized analogously to table paragraphs (see above).

* Table fields are trimmed now: leading and trailing whitespaces are deleted by the parser (previous versions did pass them to the stream).

* Tables made by tag were streamed different to tables made by paragraph, because as tags they are usually embedded into a text paragraph. Although "correct", it makes no sense to provide the extra text paragraph information in the stream. That's why the extra text information is stripped away now. In other words, the following tables are streamed the same way (except of headline formatting):

  // table by paragraph
  A | B | C
  1 | 2 | 3

  // table by tag
  A | B | C
  1 | 2 | 3

* It is now possible to close a tag made table an unusual but correct way by placing the closing tag at the end of the final table row. It's a tag, so this must be possible! And now it is.

  A | B | C
  1 | 2 | 3\B<\\END_TABLE>

* Tag declared tables can be \inlined now, declaring a string as row separator by the new option \Browseparator. This means that you can write

  Look at this table! \B<\\TABLE{rowseparator="+++"} A | B | C +++ 1 | 2
  | 3 \\END_TABLE> Incredible data!

* Inlined tables make it easy to write a macro that is replaced by a table:

            1st      | 2nd &
            __val1__ | __val2__

* Without an explicit setting, the row separator defaults to a newline (as before).

* Inlining tables enables us to \nest tables. In fact, this is still blocked by the parser \by default but \can easily be permitted by setting the new parser option \BnestedTables to a true value. The only question is how to handle nested tables converting PerlPoint to languages which do \not support them, like SDF, and that's why it is made optional and is recommended to be implemented as a \user choice.

  \\TABLE{rowseparator="+++"} column 1 | column 2 |
  \B<\\TABLE{rowseparator="%%%"} n1 | n2 %%% n3 | n4 \\END_TABLE>
  +++ xxxx | yyyy | zzzzz +++ uuuu | vvvv | wwwww \\END_TABLE


* Converters now can \predeclare variables to pass converter (call) specific informations to the document. Predeclared variables can be used like user assigned PerlPoint variables. To help users recognize them, they are capitalized by convention.

* The parser as well makes use of variable predeclaration. His settings can be overwritten by converters if necessary. For converter authors convenience, predeclared parser variables begin with an underscore. The first implemented settings are \B\$_STARTDIR which contains the absolute path of the startup directory where parsing was initiated, as provided by Cwd::cwd(), and \B\$_PARSER_VERSION providing the version of PerlPoint::Parser being used.

* \B\\INCLUDE got another special option called \"localize", enabling document authors to declare variables to be restored when the included source is processed completely. This idea was born while building documents from numerous \partial documents by different authors, each of them setting something like \$documentAuthor or \$authorMailAddress by convention. These settings are really helpful but should obviously be restored to their original values after leaving the included source! And that's exactly what this new option does. Just pass a comma separated list of the variables to be "localized" (the term is borrowed from Perl, of course) or pass the keyword \"__ALL__" to restore the current variables \completely.

  \\INCLUDE{type=pp file="nested.pp" \B<localize="docAuthor, docMail">}

  \\INCLUDE{type=pp file="nested.pp" \B<localize=__ALL__>}

* To make variable restoration possible in backends as well, all changes are propagated if \var2stream is set. A new directive is introduced to flag that all variables are reset (or deleted): \BDIRECTIVE_VARRESET.

* The new variable \B\$_SOURCE_LEVEL reflects the sourcefile nesting level. Files passed to a translator reside on level 1, this number is increased with every source nesting. Let's say certain document data should only be processed if the source is parsed as a top level document, this can be done by checking this new variable:

   ? varValue('\B<_SOURCE_LEVEL>')==1

* Umlauts became valid parts of variable names. (If you are using the Emacs \hilit19 extension provided in the \utilities subdirectory, please update to the adapted version.)


   He comes from $München.

===Active contents

* Although \Safe is a great idea and gives detailed control about code embedded into a PerlPoint source, it unfortunately limits code execution. The greatest disadvantage is that you cannot use modules in executed code, especially if they include C libraries. At least I could not find out how. So, to provide both security \and full Perl access, I decided to implement "unlimited code execution" additionally, which in fact means using \eval() and \do() instead of \Safe \on request. Just pass a true non reference value to option \safe of method \run(). It is suggested to make converters that flexible that a \user can decide if he wishes to execute active contents on his system, and on which security level. Using \eval() and \do() would be the lowest level from a security point of view. So, we already have the convention to provide options \-activeContents and \-safeOpcode by all converters. Let's say that the opcode \BALL switches to code execution by \eval() and \do().

* Another related convention: converters should avoid using namespace \main:: which might possibly be polluted by active contents in case a user chooses "unlimited code execution". As a refuge, the namespace hierarchy \PerlPoint::Converter:: is reserved by convention from now on. A converter \pp2xyz should use \PerlPoint::Converter::pp2xyz.


* installation message now contains package version;

* slight internal optimizations;

* Dealing with tags, tag parsing was optimized in general: it is no longer necessary to guard "<" characters in a tag body. (Unfortunately, this is not true yet for ">". ;-)


  # up to 0.34

  # with 0.34 and above


* Added support for \hierarchical internal links: if several headline titles are identical, it is complicated to point to one of them. An hierarchical link includes the "path" of the headline, which means all its superlevels, separated by a pipe character.

  \\SECTIONREF{name="\B<Level 1 | Level 2 | Target>"}
  \\SECTIONREF{name="\B<Level 1 | Level 2 | Level 3 | Target>"}

* Implemented first functions of a simplified condition interface which is intended to allow non (Perl) programming users to learn and perform common checks easily. The first function \BflagSet() checks if a certain setting was made. Dealing with Safe was tricky again (and therefore this function has no parameter check), but finally there's a working version!


   ? exists $PerlPoint->{userSettings}{myFlag}


   ? \B<flagSet(>myFlag\B<)>

* There's another function \BvarValue() providing the value of a passed variable, so now you can write conditions like

   ? \B<varValue(>'privateVar'\B<)> eq 'confidental'

* Text (and derived) paragraph stream representations no longer contain a final whitespace (which was made from the final carriage return character before the paragraph closing empty line). (This might be called a bugfix as well, but the trailing whitespace was never noticed as a bug and caused no trouble.)

* new headline level offset keyword "CURRENT_LEVEL" in PP file imports;


* Adapted to all modifications mentioned above where appropriate. Option \-allTags is deprecated.

* Predeclares variables \\$CONVERTER_NAME and \\$CONVERTER_VERSION.

* Unnecessary paragraph prefix backslashes are avoided in examples and inlined HTML paragraph startups now.

* Now emulates the behaviour of \pp2html to implicitly make a headline an anchor which can be used as a hyperlink target.

* Now supports \pp2html tags \\\A, \\\L, \\\PAGEREF, \\\SECTIONREF, \\\U and \\\XREF, . This support is not perfect yet for \\\L and \\\PAGEREF, but even there a quite useful approximation which should work better than the old behaviour which did in fact replace all these tags by \nothing. This crossconverter support is as well an example how the new tag definitions can be used to interchange tag syntax declarations.

* pp2sdf is no longer called alpha code.



* the value of an variable can now begin with any character:


* There was still a bug in continued list handling: if continuing points were restored from cache they contained the \old level hint, so if one inserted points before this one after the cache entry was built, the continuing point was restored at a wrong level. This is fixed.

* POD fix: \vispro option of the \run() method works on base of \chapters, not paragraphs.

* If embedded Perl code replied an undefined value this thing was tried to be parsed which caused a perl warning. Now it's completely legal to embed such code, which is especially useful to prepare variables or functions.

* Starting an ordered list by a continuing point (using "##") caused a perl warning. Code fixed.


* Installation now explicitly requires \fields.pm 1.00. I'm not sure if this was the first version of \fields that contained \phash() but it should certainly be a version providing it. An additional requirement hint was added to \README.

* The backend now provides optional process visualization as well as the parser.

* There's a new mailing list. Send an empty message to \perlpoint-subscribe@perl.org to subscribe.


* adapted to the new backend feature of process visualization;



* README adapted to new pp2sdf installation performed since 0.31;

* Lorenz recognized that macro results sometimes differed from plain usage of the replacements, investigation showed that this was the case if a macro was used without body, for reasons of parsing. The result was that the token after the bodyless macro was parsed before the macro replacement, resulting in sometimes confusing output. (As a workaround, one could use a meaningless body like in \\\MACRO\<_\>.) To make a long story short, the error is fixed with 0.32, so the stream should be correct now even when bodyless macros were used as in \\\MACRO\\MACRO\\MACRO\\MACRO\\MACRO\\MACRO\\MACRO.

* made Changelog pasing the parser again;


* Jeffrey S. Haemer suggested to extend \\\INCLUDE so that external scripts can be placed into PerlPoint without the need to copy their contents and paste it into the presentation source. To relieve presentation authors this way, \\\INCLUDE now supports a special file type of \example.

  \\INCLUDE{\B<type=example> file=script}

* The example is included as a verbatim block. Its lines can be indented by the new special option \indent:

  \\INCLUDE{type=example file=script \B<indent=2>}

* slightly rearranged \doc/parser-tags.pp when adding the example extension;



* fixed POD to be more readable by pod2man (detected by Lorenz);

* bugfix: in special constallations it could happen that headlines after file inclusion with \headlinebase=CURRENT_LEVEL were detected at wrong levels;

* the internal list counter of ordered lists used to find out where a continued list restarts is now reset when a new lists begins (as Lorenz detected, it was not reset correctly before);

* Circular inclusion of PerlPoint sources was blocked too strictly: every source was accepted only once. Now there's a better algorithm checking real nesting: you can use a source file as often as necessary, but not nested.


* added a \vim syntax highlighting file to the \utilities directory, provided by \Lorenz Domke (lorenz.domke@gmx.de);

* Ordered lists now provide the startup level, additionally to the first point which already did this if the list was continued by \##. This makes it easier for translator authors to begin the list by a certain number if necessary.

* added a tiny editing hint to the docs (FAQ);

* added \Emacs lisp code in the \utilities directory as an example how the \hilit19 module could be extended for PerlPoint;

* improved \Changelog structure;

* adapted \parser-paragraphs.pp documentation slightly;

* New option \Bsmart makes the \\\INCLUDE tag working like Perl's \require or \use for nested PerlPoint sources: the file is only read unless it was already seen before.

   \\INCLUDE{type=pp file="macros.pp" \B<smart=1>}

* This feature is \not intended for translated paragraphs but especially useful to make macros available by separate source files included whereever necessary, and to nest source parts which \both load such definitions.


* pp2sdf: all generated text paragraphs now begin with a backslash to avoid unintended paragraph style interpretation by sdf in case a colon follows the first word like in

     Remember: think twice.

* The only exceptions of this rule are made if the first word is "Note" or "Sign" because the special SDF formattings seem to make sense then.

* pp2sdf is now installed automatically by "make install";



* fixed a bug detected by \Lorenz Domke (lorenz.domke@gmx.de): embedded code can now be included into tags;

* \Lorenz Domke (lorenz.domke@gmx.de) detected unhandled macro modifications when caching was active. I figured out that this really can happen under the following rare circumstances:


* You use a string looking like a tag name not currently known as tag or alias and precede it by a backslash, like \\\NOALIAS. The paragraph does not contain any dynamically changing contents.

* Now you proceed the text with active cache. Because the paragraph looks perfectly static it is cached.

* You declare a new alias \\\NOALIAS. The paragraph remains unchanged.

* Now the next time this text is processed with active cache its checksum matches the paragraph stored before. The stream is restored and the new alias is ignored until eiher the paragraph changes or the cache is cleaned up.


* Well, this is not perfect. Users should know this at least, so I added a suggestion to clean up the cache after introducing new alias definitions.


* variable settings can be propagated into the stream now;

* improved syntactical error messages;

* According to the PerlPoint language definition tag recognition is now strictly limited to capitals and underscores (previous versions accepted lowercase letters in tag names as well). The restriction makes it easier to use backslashed strings when \all tags are accepted. Formerly, one had to write

    This is \\\\no_tag, be sure!

* Now this can be written

    This is \\no_tag, be sure!

* and only capitalized backslashed strings are still to be guarded:

    This is \\\\NO_TAG, be sure!

* As before, this has only to be taken into account if users prefer to accept everything looking like a tag as a tag (see \\\ACCEPT_ALL for details).

* Consequently, lowercased letters in alias/macro definitions are no longer valid. They are accepted but automatically converted into capitals, while the parser displays a warning.


* internal code cosmetics;


* "}" characters outside verbatim blocks and embedded HTML are now replaced by SDF escape \{{CHAR:rbrace}} to guard translations of things like \\\B\<{key=\value}\>>



* modified the parser module a way that its version can be displayed by CPAN;

* New common way to pass basic data to active contents: translator authors can pass a hash of settings which is mirrored in the active contents namespace as \\%$PerlPoint. This makes it easy to pass things like the current target language or call specific user settings.

* Two conventions are already introduced: \targetLanguage and \userSettings (allowing users to control the translation process more detailled by passing simple translator options).

* See \pp2sdf for an example of usage.

* Simplified backslash handling in macros. Before 0.29, each macro level (and each tag level in the macro replacement, of course) added a need of yet another backslash before closing angle brackets (and literal backslashes) in the targetted text, making it difficult to nest a number of macros (especially if they were declared by others).

* So, until now, one had to write

<<EOE \$var-\{key}> \\$var-\{key}> EOE

* but

<<EOE \$var-\{key}> \MACR\$var-\\\{key}> \MACR\MACR\$var-\\\\\\\{key}> EOE

* - not really handable. And inconsistent (nested \tags never required aditional backslashes, but macros did).

* From now on, only \one backslash is necessary \ever:

<<EOE \$var-\{key}> \\$var-\{key}>


* \Note: existing PerlPoint sources might have to be updated.

* Accelerated macro handling - earlier versions used to \reparse a macro body which delayed parsing in case of deep macro nesting. From now on, every token is only parsed \once. As an additional benefit, a significant amount of code could be removed and the backend module needs no longer to be used by the parser which accelerates startup time of translators in general.


* pp2sdf target code embedding is now restricted to SDF and HTML, but these both \are officially supported now (as well as Perl, of course ;-);


* removed an old POD headline prefix in README which caused CPAN to run pod2html (and to display the related section only)

* made it backwards compatible to perl 5.00503 again;


* module namespace changed from "PP" into "PerlPoint";

* first CPAN release;



* bugfix in active contents handling: it might happen that lines were ignored;

* bugfix in dynamic contents handling: nested dynamic contents (resolved macros, embedded parts, included parts) might not be handled in the correct order;

* The cache now considers headline level offsets of nested sources. Before 0.26 it might happen that the headline level offset was modified but an \old offset was restored from cache.

* \The cache format was modified. You \can still use existing cache files but old data therein will never be referenced. To clean up old caches just remove them or use the parsers cleanup facility.


* new utilities directory, providing a NoteTab clipbook file initially;

* new dynamic FAQ document;

* new headline level offset keyword "CURRENT_LEVEL" in PP file imports;

* The parser now changes into the directory of a sourcefile when it is processed. This way nested sources can always use relative pathes, so assembling a patchwork document becomes easier. \Existing sources including files from other directories may have to be adapted!

* Example: if a source located in \subdir is processed and includes

      \\INCLUDE{type=pp file="\B<subdir/>subdoc.pp"}

* to nest a file in its own directory, \subdir had to be specified because all pathes were relative to the parsers startup directory. This was not very logical and made it hard to combine nested sources.

* Now the parser changes into the source directory, which means that nested source pathes are relative to the path \of the including PerlPoint file, in our example:

      \\INCLUDE{type=pp file="subdoc.pp"}

* the package now explicitly requires perl 5.6;


* pp2sdf: avoid unintended SDF phrases (\P\<something\>) now by generating "<" as "{{CHAR:lt}}" (outside verbatim blocks);


* completed prerequisites list in Makefile.PL;

* added notes about Storable updates to parser documentation;

* cache bugfix;

* trace code bugfix;

* added a new demo application pp2sdf which is a complete translator;

* process visualization: bugfix and improvements;

* new experimental tag setting "\ACCEPT_ALL" switches to a "anything that looks like a tag is processed as a tag" mode to simplify source translation by tools implementing different tag sets;


* new checksum based incremental parsing accelerates updates (depending on the document structure);

* ordered list can be continued now: simply use \## instead of \# to continue a previously opened list \in the same chapter (needs translator updates to take visible effect);

* new \run() option \"vispro" to activate process visualization;


* several fixes in embedded code handling, tables can now be inserted dynamically;

* providing a new basic PerlPoint documentation about active contents;

* new trace mode TRACE_ACTIVE (active contents evaluation);

* improved README file;


* the include directive has a new optional headline level offset parameter;

* several slight bugfixes in trace modes;

* when reading included files, the main file handle is now closed instead of duplicated and reused - perl5.6 run into difficulties in some cases (if several files were included, the file pointer seemed not to reset correctly while switching back to the original handle);

* now providing basic PerlPoint documentation in new "doc" directory, may be included into translator docs;

* added a first README version;


* completed table paragraph implementation according to the design paper: first row is now automatically streamed as "table headline" (it is up to the backend to format such table cells as it wishes);

* added macro (or alias) feature, call it still experimental because there might still be untested cases;


* table \paragraphs are implemented now;


* new condition paragraphs which allow to maintain presentation versions in several languages by one source file, for example;

* bugfix in table handling: table stream structure was wrong (the bug was discovered by \Lorenz Domke (lorenz.domke@gmx.de));


* fixed an error in Makefile.PL comment;

* embedded Perl code is evaluated now (in a safe, user configurable environment);

* Perl code can now be included as well as embedded;

* PerlPoint can now be embedded as well as included;

* variables are made accessible by included or embedded Perl code;

* added a simple visualizer to the demo directory (pp2tree);

* new optional source line hints in the stream;

* bugfix: backslashes in embedded parts shall not be special (as they were);

* new multi level list shifts (e.g. "<2"), level is optional and defaults to 1;


* Bugfix release: if a paragraph started with a tag, paragraph openers (like "*", "#", "-" and so on), they were handled as special characters until the tag was closed, which made them invalid inside the tag. Thanks to \Lorenz Domke (lorenz.domke@gmx.de) for the bug report.

* There was also a bug in tag handling which could be fixed by the way.


* added a tag test which demonstrates the usage (and function) of string parameters ;

* the grammar file became part of the distribution;

* updated parser documentation;

* definition list items were plain strings before, now they are streamed input which means that they can contain tags, for example (thanks to \A. Sigel for the suggestion);

* Added a filter feature to run(). The parser can now suppress included or embedded code of a language different from the target one. Embedded HTML code, for example, makes no sense in a stream passed to a backend which generates LaTeX or PostScript.


* this is just another attempt to make the package 5.005 clean ... thanks to \Lorenz Domke (lorenz.domke@gmx.de) for his tests;

* added the Changelog part of 0.14 which was prepared but not delivered with 0.14;


* improved handling of special characters, as a consequence, only backslashes should have to be guarded by a backslash now if they should stay for itself (except of in the beginning of a new paragraph where a number of special characters have to be recognized);

* added new demo script pp2pp in a first version which successfully processes Changelog and TODO file;

* bugfixes: stream contained trailing whitespaces for list points and headlines;

* bugfix: empty lines in verbatim blocks were not streamed;

* bugfix: stream contained leading newline for verbatim blocks;


* made it backwards compatible with perl 5.00503 again (until more people will have 5.6.1);


* added demo translator pp2pod;

* leading spaces in list points are ignored now;

* various bugfixes in parser behaviour (success flags);

* made the Changelog file passing the parser ;-)


* added embedding (enclosed blocks of foreign code);

* added including (embedded files);

* added a first version of tables;


* fixed various block bugs detected by \Lorenz Domke (lorenz.domke@gmx.de): verbatim block opener was supplied in stream in 0.09, first indentation in blocks was not supplied;


* started Changelog;

* verbatim blocks now start with a "here document" hint immediately;

* new "definition list point" paragraphs;

* safer tests;

* streamed lists are embedded into list directives now;

* modified syntax of verbatim blocks;

* added variables;

* modified tag syntax: TAbody => \\TAG[{parlist}][<body>];

6 POD Errors

The following errors were encountered while parsing the POD:

Around line 33:

Unknown directive: =Fixes

Around line 42:

Unknown directive: =Features

Around line 67:

Unknown directive: =Fixes

Around line 1911:

Non-ASCII character seen before =encoding in '$München=Munich'. Assuming ISO8859-1

Around line 2292:

Deleting unknown formatting code O<>

Deleting unknown formatting code O<>

Deleting unknown formatting code O<>

Around line 2628:

Deleting unknown formatting code G<>