Wed, Jun 15 2011

   *  zoem-11-166 released.

   *  The default search path was changed. The obsolete PREFIX/zoem/mac
      component was dropped, and PREFIX/aephea was added.  Now, if zoem and
      aephea are both configured with the same prefix for installation (e.g.
      --prefix=/usr/local or --prefix=$HOME/local), then aephea include files
      will be found automatically by zoem.

   *  set#3 can now splice into a variable (that is, the string
      refered to by a macro). This is achieved by pairing the respective
      keys {start} and {width} with appropriate values in the first
      argument to set#3. The start position is relative to an offst
      of zero, so

      \set{foo}{kaboom}
      \set{{start}{2}{width}{3}}{foo}{za}
      '\foo'

      yields 'kazam'.

   *  The example zoem solution 8q2.azm to the N queens problem
      has been updated. It now pretty-prints asciific chess boards using the
      new splice mode to set#3.

   *  Small documentation fixes.


Wed, Sep 22 2010

   *  zoem-10-265 released.

   *  In interactive mode zoem now utilise readline editing and history
      capababilities if available. This can be disabled completely at compile
      time or optionally at run time.

   *  A new builtin macro seq#4 can be used as a simple type of
      for loop.

         \seq{i}{a}{b}{ ... }

      is equivalent to the pseudo code

         for (i=a;i<b;i++) {
            ...
         }

      Internally, its definition is

         \set{\1}{\2}
         \while{\let{\get{''}{\1}<\3}}{
            \4
            \setx{\1}{\let{\get{''}{\1}+1}}
         }

   *  Environments that are not closed at end of processing are now reported in
      a diagnostic message.


Fri, Jan 8 2010

   *  zoem-10-008 released.

   *  The HTML document framework has been wrapped up and split off under the
      name Aephea, Adaptable Exo-skeleton for Practical HTML Extension and
      Abstraction. PUD (Portable  UNIX Documentation) is now shipped with
      Aephea.  PUD provides limited support for two mini languages (manuals and
      FAQs) for two output devices, troff and HTML. Aephea provides more
      elaborate support for a single output device, HTML.

      The zoem package from now on provides only resources and documentation
      for the macro/programming language zoem itself.

   *  This release mostly provides improved and streamlined support
      for using zoem as a convenient and powerful layer on top of HTML,
      providing full extension and abstraction capabilities.

   *  New device scope syntax was introduced specifically targeted
      at HTML: \@e{ent} will expand to the character entity &ent; --
      it is equivalent to and minimally shorter than \@{&ent;} .

   *  Trailing whitespace is removed from the output captured
      by \system#3 invocations. This may be made optional in the future.

   *  Set macros can now be targeted at named dictionaries within the
      dictionary stacks (of which there are two, the stack associated with
      \begin{} - \end{} environments, and the default user stack). A new get#2
      primitive allows getting from named dictionaries.

      User stack
      Getting: \get{foo}{tim}                   (looks only in foo dictionaries)
      Setting: \set{{dict}{foo}}{tim}{zut}      (sets in topmost foo dictionary)

      Env stack
      Getting: \get{foo}{$tim}                  (looks only in foo dictionaries)
      Setting: \set{{dict}{foo}}{$tim}{zut}     (sets in topmost foo dictionary)

      The stack type to query is inferred from the key signature.  No mechanism
      is yet provided to get and use, in this manner, keys that take arguments,
      although it can be achieved using apply#2.

   *  The f#2 (mathematical functions) primitive has new modes (first argument
      to f#2) irand and drand for generating respectively integer and double
      random numbers within the range specified in the second argument to f#2.

   *  An update mode for setting keys has been added to set#3, where a key is
      required to be already present. The dictionary stack will be traversed
      downwards until a key is found. This can be used to update keys in a
      first environment from within a second environment. Aliases update#2 and
      updatex#2 are provided.

   *  When the html device is set using the command line option -d html,
      zoem automatically sets up mapping tables so that the HTML special
      characters are converted where necessary.

   *  New syntax for HTML/XML tags that do not have an associated close tag,
      such as <meta>, <link>, and <hr>. This is now specified in zoem
      as for example \<*meta>.  The previously accepted syntax \<meta!> is now
      deprecated. The new syntax is more reasonable and makes it easier
      to recognize such special tags.

   *  Zoem recognizes special classes of HTML/XML tags that do not have an associated
      close tag. For now these are tags that start with '!' or '?'.
      Previously accepted syntax such as \<!-- comment --!> is deprecated
      and is now simply written \<!-- comment -->.

   *  The read routines now ignore nul bytes in files, improving zoem's
      robustness. It no longer yields unpredictable results on files with
      embedded nul bytes.

Fri, Sep 5 2008

   *  zoem-08-248 released.

   *  A bug manifested itself on powerpc + linux in the form of a crash,
      resulting from using vsnprintf repeatedly without sufficient
      corresponding va_start and va_end invocations. On some other platforms
      this bug would manifest itself as a mysterious zoem syntax error due to
      heap corruption. Fixed. Thanks to Philipp Benner, Tobias Quathamer, and
      Joost van Baal for bug reports.

   *  Two modes for the defined#2 primitive have been renamed.
      "prim" has become "primitive" and "alias" has become "builtin",
      in line with zoem nomenclature and documentation. This change
      is backwards incompatible, but unlikely to affect anyone.

   *  Option -ndollar was not working. fixed.

Thu, Nov 29 2007

   *  zoem-07-333 released.

   *  SYNOPSIS
      A bug in set#3 was removed, and two directives were added, namely batch
      definitions and a more legible way of appending to keys (see below).

      Keys local to environments (aka dollar keys) can now be defined using
      the same signature as when invoked, e.g. \begin{env}{{$k1}{v1}{$k2}{v2}}.

      This release also continues the focus on retiring what little special
      case syntax still exists.  The format#2 specification syntax has been
      reworked (in an incompatible way) into a standard and extensible
      key-value syntax (see below). The format#2 implementation was redone and
      several new format#2 directives were added.

      xml/html syntactic sugar now stops introducing whitespace formatting
      when so instructed with the at directive \@{\w}.

   *  Fixed bug in set#3 where usage of the {if} or {unless} directive
      corrupts state when combined with the {modes}{x} directive.

   *  xml syntactic sugar cubes now obey the device scope \w directive.
      This means that xml elements, in the scope of an at directive
      \@{\w} no longer introduce white-space. This only affects the
      formatting of the XML/HTML code itself.
      This is for example useful when creating <pre> elements.

   *  The set#3 {modes} directive now accepts a v (for vararg mode)
      where a batch of key-value pairs can be defined within a single set#3
      invocation. This may improve the readability of resource files
      containing many definitions. Example:

         \set{{modes}{vw}}{}{
            {is_last}   {0}
            {ctrprev}   {0}
            {ctrnow}    {1}
         }

   *  Introduced append mode in set#3:
         \set{foo}{bar}
         \set{{modes}{a}}{foo}{bar}
      sets foo to barbar.

      append#2 is an alias for \set{{modes}{a}}{\1}{\2}
      appendx#2 is an alias for \set{{modes}{ax}}{\1}{\2}

      Hence append#2 appends to keys without expanding and appendx#2 appends
      to keys after expanding the to be appended text.
      Keys that are not defined when appended to will first be instantiated
      with an empty string.

   *  Keys local to environments (started by \begin#2) can now
      be defined using the same signature as used when invoking them, i.e.
      \begin{env}{{$foo}{bar}} as well as \begin{env}{{foo}{bar}}.

   *  Some bugs were found and removed in the format#2 implementation.

   *  format#2 has acquired a reuse mode where a format specification
      can be be reused a number of times or exhaust all arguments (see below).
      It has also acquired the 'border' directive for adding constant strings
      to the left and right of the formatted units.

   *  The tr#2 specification language is now also described in zoem(1), the manual
      page of the interpreter. The idea is that all - as few as possible -
      special purpose sub-languages will be described in there for ease of
      reference.

   *  The unwieldy format#2 specification language has been redefined to
      follow standard zoem syntax using keys and vararg value lists.  It is
      more verbose, but also more robust, compatibly extensible, powerful, and
      more easily understandable. As an example one now writes

      \format{
         %{{padding}{-=-_}
           {delimit}{ }
           {align}{left}
           {length}{length}
           {double}{{{arg}{foo}}}
           {width}{40}
          }%{
            {align}{right}
            {width}{40}
          }
      }{
         {arg1}{arg2}
      }

      rather than

      \format{%~{-=-_}{ }<40*{length}{foo}.%>40.}{
         {arg1}{arg2}
      }

      Formerly a format specifier was enclosed inbetween '%' and '.' . It
      interpreted special characters '<', '>', and '=', as well as the special
      characters '@', '*', '~' each of which had to be followed by two blocks.
      The same syntax is now obtained by a more customary key-value structure
      encoded as a zoem vararg.

      format#2 accepts new directives {border}{{left}{right}},
      {delimit}{{left}{right}}, {double}{{[{key}{value}]*}}, and
      {reuse}{<num>|*}.

      {delimit} parameters decorate the formatted argument on the sides, and
      participate in the computations involving width, alignment and padding.
      There is no padding inbetween delimiters and the argument to be
      formatted.

      {border} parameters decorate the entire formatted product and do not
      participate in aforementioned computations. Padding is computed and
      produced before the borders are added.

      {double} can be used have the length computations performed on
      fake parameters. This can be useful when certain formatting
      elements need to be present in the true parameters.

      {reuse}{<num>} will reuse the current specification <num> times,
      {reuse}{*} will reuse it until all arguments are exhausted. Thus

      \format{%{{align}{right}{width}{10}{reuse}{*}{padding}{{'}{}}}}{
         {beauty}{lies}{in}{the}{eye}{of}{the}{beholder}
      }
      \textmap{{repeat}{8}}{1234567890}

      results in

''''beauty''''''lies''''''''in'''''''the'''''''eye''''''''of'''''''the''beholder
12345678901234567890123456789012345678901234567890123456789012345678901234567890


Fri, Nov 9 2007

   *  zoem-07-313 released.

   *  \set{''foo}{bar} now works. It sets a key in the bottom (global)
      user dictionary. Same semantics as \set{{modes}{g}}{foo}{bar}.

   *  \undef#1 now checks whether its argument is a valid key signature.

   *  \undef{''foo} now works.

   *  Streamlined the error-processing part of the code.

   *  Added \textmap{{repeat}{<num>}}{<string>} mode. This results in <string>
      <num> times repeated. <num> is evaluated before use.

   *  The special magic comment sequences
         \:/   delete rest of line including newline
         \:!   replace by '\' (for quoting zoem comments)

      are now deprecated (but still supported). Preferably use
         \:{/}
         \:{!}

      instead. This syntax is both more robust and more extensible.  This
      change is similar to previous changes introducing \`{}, \={} and \*{}.

Wed, Jul 25 2007

   *  zoem-07-205 released.

   *  Zoem is now licensed under the GNU General Public License version 3
      (or later).

   *  Added \set#3 that provides all the modalities of set#2, setx#2,
      def#2, defx#2 (namely optional warning, expansion), as well as
      additional modalities. It is possible to define a key conditionally (if
      not set already), and a key can be explicitly set in the global (bottom)
      user dictionary.

   *  \''foo will look up foo in the bottom user dictionary, allowing
      this dictionary to function as a global namespace.

   *  Success status, as written in the session macro \__zoemstat__ by try#1
      and catch#2 is now associated with the string "done" rather than "ok".
      The status "done" can now be explicitly thrown with \throw{done}{}.  In
      this case the unprocessed part of the current interpretation stack is
      discarded, but no error is generated.

      When not captured by try#1 or catch#2, \throw{done}{} can be captured by
      eval#1 and while#2.  If neither of these four primitives catches the
      throw, it is captured at the file processing level and results in
      end-of-processing for the current file. This is fully equivalent
      with what used to be the primitive \done. This primitive has now
      been implemented as the alias \throw{done}{}.

   *  The primitive \done is now implemented as \throw{done}{}.

   *  Introduced \`{..} syntax, equivalent with \`..` syntax. These are only
      available within the \formatted#1 primitive (where .. is a placeholder
      for valid formatted#1 directives).  This change is similar to previous
      changes introducing \={} and \*{}. The old \`..` syntax is deprecated
      and if it is used error messages are issued. The rationale for these
      changes is that the new syntax is more pleasant on the eye and will be
      extendible if needed.

   *  Added \ifdef#3 and \ifdef#4 aliases. This is the ifdef#3
      definition:

      \set{ifdef#3}{\if{\defined{\1}{\2}{\3}}}

   *  \defined{mode}{} now accepts modes 'zoem', 'prim', 'alias'.
      mode zoem accepts the combined classes of prim(itives) and
      alias.

   *  The command line option -s accepts 'key' as well as 'key=val'
      syntax. In the former case the value of key will be set to '1'.

   *  Added two solutions to the N-queen problem in the examples directory.

Wed, Dec 13 2006

   *  zoem-06-347 released.

   *  Fixed display bug in PUD -- the { itemize, long-item, html, safari }
      combination where long-item would be split over lines.

   *  Clean-ups and modifications in underlying util library (mostly
      hash-related).

Wed, Aug 23 2006

   *  zoem-06-234 released.

   *  Added beta key \__env__#1 to retrieve environment variable settings.
      \defined#2 tests for existence of a name in the environment with
      the ENV type.

   *  Added caveat in zoem(1) and Zoem User Manual documentation;
      \done will stop processing the current file only if it was read in its
      entirety, i.e. if the chunk size is sufficiently large. The default
      chunk size (1Mb) should be more than sufficient for authoring
      environments.

   *  The session key \__fnup__ is set to the file that includes
      the current file to aid in diagnostics.

   *  The underlying utility library was overhauled - it's integer
      type system was redone and a more rigorous design was put in place,
      following common practices (size_t and ssize_t wise).

Tue, Mar 22 2006

   *  zoem-06-080 released.

   *  A number of developments have settled with release. To sum up the most
      noteworthy items (C, F and H stem from this release):

         A) primitive namespace identification with \'
         B) user keys can shadow primitives
      +  C) user keys can inspect (vararg) argument structure with \nargs#1
  ABC => D) improved facilities for backward compatability
         E) PUD or portable UNIX documentation is now a separate entity
      +  F) \__line__ now yields what it promises
         G) many primitives recoded as user macros (counters and references)
      +  H) \switch#2 can group branches
         I) \register adds useful processing hook registration
         J) \whilst#2 flushes output immediately

   *  A long standing minor bug has been fixed. Comments up to and including
      the newline can be stripped with the sequence \:/.
      This used to corrupt the file line counter as seen in the session macro
      \__line__ and in error messages. No more.

   *  \switch#2 can collect several cases together using a vararg.
      In the \nargs#1 example below one might use e.g.

      \def{test#1}{
         \switch{\nargs{\1}}{
            {{-1}{-2}} { .. regular stuff .. }
            {   .. apply stuff .. }
         }
      }

      to do the same thing for both of the -1 and -2 cases.
      Different devices can easily be lumped together using the idiom

      \${ {{txt}{roff}}  { .. stuff A .. }
          {html}          { .. stuff B .. }
       }

       In this case the txt and roff devices pass through the same
       branch. \$#1 is a new alias introduced further below.

   *  Added \nargs#1 primitive which counts the number of scopes in its
      argument - WHICH IS NOT EVALUATED. This allows construction of user
      macros that can overload one or more particular arguments, i.e.  behave
      differently depending on the number of scopes in the argument.

      A likely application is an argument that either specifies a simple
      string (assuming default modes) or a vararg with strings and 
      modes specified using a key-value vararg.
      One usage is to extend a macro's functionality while retaining
      backward compatibility.

      \nargs{{abc}def}     : -2      \: neither fish nor fowl
      \nargs{abc}          : -1      \: regular argument
      \nargs{abc{def}}     : -1      \: regular argument too
      \nargs{}             :  0      \: empty vararg (not empty regular arg)
      \nargs{   }          :  0      \: empty vararg again
      \nargs{{abc}}        :  1      \: vararg with 1 argument
      \nargs{{abc}{def}}   :  2      \: vararg with 2 arguments
      \nargs{   {abc}
                {def}
      }                    :  2      \: vararg with 2 arguments
      etc.

      User macros can now use a framework such as

      \def{test#1}{
         \switch{\nargs{\1}}{
            {-2}{\inform{mixed feelings}\exit}
            {-1}{ .. regular stuff .. }
            {  \switch{\let{\nargs{\1}%2}}{
                  {0}{\apply{_#2\!{{ .. stuff with \1 and \2 .. }}}{\1}}
                  {\inform{vararg invocation requires paired arguments\exit}}
               }
            }
         }
      }

      to branch on scope presence/absence. The purpose of the -2 return value
      is to promote frameworks that spot errors early. Group -2 and -1
      as {{-2}{-1}} within switch to make them pass through the same branch.

   *  Added $#1 and $#3 aliases.
      $#1 maps to \switch{\__device__}{\1}            \: switch
      $#3 maps to \switch{\__device__}{{\1}{\2}{\3}}   \: ifelse

   *  within write#3 invocations the key \__fnwrite__ is set to the
      appropriate output file name. This allows easy branching on the file
      being written to from macros accessed within the third argument. The key
      is deleted after evaluation of the third argument has finished.  Nested
      invocations will currently overwrite this key.

   *  Spreading \foo over two lines as in
      \fo\:/
      o

      (with the 'o' at the beginning of line) will be correctly parsed
      regardless of chunk size.

Wed, 22 Feb 2006

   *  zoem-06-053 released.

   *  inspect#4 was broken since the previous release; the parse
      code responsible for firing of the callback macro had not
      been updated properly. Fixed.

   *  minor documentation improvements.

   *  The TODO was shrunk to a more reasonable size and PUD
      now has its own DOPUD.

Wed, 15 Feb 2006

   *  zoem-06-046 released.

   *  Major clean-ups in both code and documentation. Facilitated
      syntactic separation of primitives and user macros.

   *  (minor) Backward incompatible change in the PUD macro package.

      As a further clean-up of the reference framework (in which refload#2
      and friends moved to ref.zmm), publanch#2 and publref#1 have now
      moved to ref.zmm and have been renamed to reference#2 and refer#1
      for better consistency.

      ref.zmm is now documented in pud-ref(7).

   *  pud-ref(7)
      pud(7)

      were added. The first describes the reference framework, which was
      recently removed from the Zoem User Manual as it is no longer
      implemented by primitives. The second is an introduction to and an
      overview of Portable Unix Documentation.

   *  The counter primitives \ctradd#2, \ctrset#2, \ctrput#1 and the builtins
      \ctrinc#2 and \ctrdec#2 were recoded as user macros in the new macro
      package ctr.zmm .

   *  For primitives or builtins with standard alphabetic names
      and the special primitive \$#2 and the special builtins
      \"" and \""#1 the syntax

         \'standard{arg1}..
         \'${arg1}{arg2}
         \'""
         \'""{arg1}

      is now accepted. Moreover, primitive definitions can be shadowed in
      the user dictionary. Currently a warning is issued when this happens.

      The prefered way of accessing primitives is now with the new syntax.
      This makes existing zoem files compatible in case future zoem
      primitives clash with user-defined keys.

      It also makes works based on such files still extendible since the
      new primitive can be unambiguously accessed by the new syntax.

      Finally, with the new syntax zoem needs to do less hash table
      lookups, making it again slightly faster. Said tiny speed increment
      has been fully negated by the redefinition of counter primitives
      as user macros however.

   *  The number of builtin aliases has been cut back.

   *  The idiosyncratic \> (close XML-type tag) syntax has been removed.
      Use \<> or explicit \</foo> style syntax.

   *  Cleanups and deduplication in the parsing code. Examples:
         \apply{<>#2}{{foo}{bar}}
         \apply{$#2}{{html}{yes}}
      now do something. You probably will not need it.
         \def{<>}{foo}
      now chokes.

   *  Documentation fixes and additions.

Tue, 10 Jan 2005

   *  zoem-06-010 released.

   *  Assembled the man and FAQ macros under the Portable Unix
      Documentation label.

      -  the man_zmm(7) manual page was renamed to pud-man.
      -  the faq_zmm(7) manual page was renamed to pud-faq.
      -  the generic_zmm(7) manual page was renamed to pud-base.

   *  Redid a lot of documentation. Further separated package descriptions
      from core zoem documentation.

Thu 5 Jan 2006

   *  zoem-06-005 released.

   *  Pruned the Zoem User Manual to only describe the core zoem language.
      Diatribes and mini-language excursions were removed.

   *  A missing clearerr in util/io.c caused
         ./zoem --unsafe -E '\register{END}{\system{date}}'
      to loop on EOF on at least one architecture (mac OSX). Fixed.

Thu, 24 Nov 2005

   *  zoem-05-328 released.

   *  This release contains substantial behind-the-scenes changes.
      Oddbal zoem primitives were reimplemented as macros of exisisting
      primitives or as special cases of a new more generic primitive.  Users of
      packages (man, faq, doc) should not be affected, with this exception:
       ____________________________________________________________________ 
      {                                                                    }
      {  The new zoem will choke on existing *.zmr and *.zmt auxiliary     }
      {  files. Remove these and you should be fine.                       }
      {____________________________________________________________________}

      The reason being that the primitive refload#6 is now replaced by
      the macro refload#2 provided by the new ref.zmm package.

   *  Added \register#2, use e.g. as

      \register{END}{
         \if{\ctrput{zoem::ref::misses}}{
          \write{stderr}{txt}{>>> [REF] \ctrput{zoem::ref::misses} missing references\|}
         }{}
      }

      This example is from the file ref.zmm that now implements zoem
      references.  This particular registered macro outputs a warning after all
      input has been processed if missing references were found during a run.

      Currently only END is supported. registered macros are processed in
      the order of registration.

   *  Reference handling was removed from the core zoem language.
      \refload#6 and \ref#2 were reimplemented as simple macros (\refload#2 and
      \ref#2) using the zoem data facilities and the new register#2 primitive.
      They are found in the new macro file ref.zmm.

      \refloadx#2 was added, it can be used to attach custom information to
      a reference.

   *  \ucase#1 \roman#1 and \alpha#1 are reimplemented as simple
      macros - and deprecated. They are now specific invocations
      of the new \textmap#2 primitive.

   *  \textmap#2 implements a variety of text transformations. Among them:

         {word}{ucase}
         {word}{lcase}
         {number}{roman}
         {number}{alpha}
         {caesar}{<num>}
         {vigenere}{<key>}
         {vigenerex}{<key>}

      These transformations are applied sequentially, implying that

         \textmap{{number}{roman}{word}{ucase}}{\your_nice_counter}

      does what you expect it to do. Use caesar and vigerene[x] to
      encrypt your sensitive data.

   *  Added \whilst2, which sends output to the current output file
      without building up intermediate results.

   *  The tr#4 was sanitized. See further below.

   *  The zoem output implementation was thoroughly cleansed
      and is in a maintainable state now. This bears on the handling of
      multiple output sinks (altiplexing?).

      Same for environment well-formedness checks.  As part of the clean-up an
      experimental (but backwards compatible) feature was put in place; see
      immediately below.

   *  This item is important only for people that write their own
      environment definitions.  Dictionary stacks are now tied to (output)
      sinks. This implies that

      \begin{foo}
         \write{zutfut}{device}{\begin{zut}}
      \end{foo}
         \write{zutfut}{device}{\end{zut}}

      is now legal zoem, and any dollar key within a write scope is tied to the
      output stream defined by the write invocation.  Different instances of
      the same environment spread over different output sinks will thus not
      clobber one another.

      If a dollar key is not found within some write scope, it is searched
      in the default output scope.

   *  added \genoptref#1 to the generic package, for generic (html) linking to
      an option only using the option identifier string.

   *  The first argument of inspect now takes a vararg rather than
      a single string. The modifiers previously accepted as the first
      argument are now accepted as argument to the 'mods' key.
      So use e.g. \inspect{{mods}{count-matches}}{..}{..}{..}
      if counting matches is what you are after.

   *  Incompatible change in what presumably is one of the
      lesser-used primitives.

      Removed \tr#4 because its interface was broken beyond repair.
      Added \tr#2.

         \tr{  {from}{from-spec}
               {to}{to-spec}
               {delete}{delete-spec}
               {squash}{squash-spec}
            }{
               data
            }

      Data is unprotected, all specs are subject to tilde expansion.

      Specs are largely POSIX compliant except that repeats are
      denoted

         [*c*20]     \: repeat c 20 times
         [*\012*20]  \: repeat newline 20 times
         [*X*]       \: fill with X
         [*X#]       \: repeat X until class/range boundary

      The magic repeat operator # stops on boundaries. A boundary happens at
      start or end of a class or range.  Complements are denoted by a leading
      caret.

   *  \push#1 and \pop#1 semantics have changed.  \push#1 now always pushes the
      user dictionary stack. Its argument now denotes the name of dictionary
      stack that is pushed.  The matching \pop#1 should specify the same name.

   *  made -o work with -e and -E.

   *  rewrote parts of zmm_generic for the html device. It now makes
      more use of style sheets and is reworked towards better
      customizability.

   *  Included a description of the \inspect#4 and \format#2
      (mini-mini) sublanguages in the zoem(1) manual. This covers
      respectively tilde expansion for \inspect#4 and the format
      specification strings for \format#2.

   *  itemize accepts {smallertext}{1}

   *  Added fase support for centering with \format#2. With centering,
      the default now is that strings of length 1 and 2 will be layed-out
      thusly:           (even width)
         [....8...]
         [....9...]
         [...10...]
         [...11...]
      or                (odd width)
         [....8....]
         [....9....]
         [...10....]
         [...11....]
      This can be changed by using '==' as the centering specification:
         [...8....]
         [...9....]
         [...10...]
         [...11...]
      and
         [....8....]
         [....9....]
         [....10...]
         [....11...]

      This pattern applies generally to strings of length (2*k+1, 2*k+2).
      In case you need it, the fase will alternate with each equal sign
      appended to it.

   *  Zoem's \alpha#1 macro (now implemented as
      \textmap{{number}{roman}}{\1}) transforms numbers to strings by writing
      them in base 27, using _ as zero.  It's use is primarily to enable a) b)
      c) style itemization.  While experimenting with alpha#1, I found that

         \alpha{1444} = azm

     The most wonderful coincidence as azm is the zoem suffix
     and 1444 is 38 squared - 38 is one of my lucky numbers no less.

     This is a universal truth rather than a change. However, the simple act of
     perceiving a universal truth instantly creates truth ripples through the
     fabric of everything, and in doing so changes the course of the universe
     in profound ways.  I thought you might like to know this.

Fri, 3 Jun 2005

   *  zoem-05-154 released.

   *  Fixed segmentation fault occurring with -e option, introduced
      in previous release.

   *  Added a regression test suite, stress2.azm.
      stress.azm is now also called a regression test suite.

   *  Tested a large number of error paths, fixed a few
      to yield truncated output as they should.

   *  Fixed doc.zmm to produce better W3C compliant output -
      specifically compliant <a> anchors.

   *  Added a tilde escape mechanism to inspect#4 to prevent
      cumbersome regexes when slashes are needed (either as a
      literal or as a regex metacharacter).

   *  Sanitized the \inspect#4 implementation and interface.
         supported options:
            iter-args            (was vararg)
            iter-lines           (was lines)
            discard-nmp          (was skip)  [non-matching-part]
            discard-nil-out      (was snip)
            count-matches        (was count)
            discard-miss
            dotall
            icase

Tue, 24 May 2005

   *  zoem-05-144 released

   *  Writeto obeys the --unsafe and --unsafe-silent options (or their
      absence) when its filename contains a path separator, the forward
      slash (entirely UNIX-centric). Such a path separator is considered
      a risk in overwriting sensitive files. Zoem will prompt
      the user according to the safety settings. It will never prompt the
      user if the filename does not contain a path separator.

   *  The primitive env#3 has become env#4 and env#3 is now a macro
      expanding to \env{\1}{}{\2}{\3}.
      The second argument of env#4 is a list of key-value pairs
      in a vararg that are set as default dollar keys. This simplifies
      the definition of environments.

   *  keys set in the second (vararg) argument begin#2 can now
      be keys taking arguments. e.g.

         \begin{foo}{{bar#2}{\!2\!1}}

      defines a key that is used as

         \$bar{a}{b}

      which would result in

         ba

   *  Environments set \$__args__ and \$__xargs__ so that they can pass
      the full argument list onwards if needed.

   *  Environments will now silently overwrite local keys previously defined.
      This enables a convenient override system when one environment
      encapsulates another environment - it can append user arguments
      (using \$__args__) to default settings.

   *  Changed --allow=<name>[:<name>]* syntax to
      -allow <name>[:<name>]* syntax.

   *  Moved option parsing to structured mcxOption framework.

Mon, 21 Jun 2004

   *  zoem-04-173 released.

   *  Repeated use of \special#1 would decrease available stack space.
      Quite rare bug revealed by Tim Hughes as one of the first to use
      clmformat/zoem on a larger scale.
      This will not affect normal usage of zoem.
      Fixed.

   *  Made the code better ISO-C compliant (split long constant strings).

Wed, 16 Jun 2004

   *  zoem-04-168 released.

   *  fixed embarassing special-related bug. It *did* affect exisiting macro
      packages (see below).

Tue, 15 Jun 2004

   *  zoem-04-167 released.

   *  \special#1 now evaluates its argument. This does not affect existing
      macro packages.

   *  Updated generic_zmm documentation.

   *  Treat spaces in manual name correctly (fixed bug in man.zmm results
      spotted by Joost van Baal).

Sun, 13 Jun 2004

   *  zoem-04-165 released.

   *  fixed bug where inline files would cause zoem to crash.

   *  introduced enclosing variants for all paragraph incarnations.
      \par     ->    \par#1         (content)
      \cpar#1  ->    \cpar#2        (caption, content)
      \car     ->    \car#1         (content)
      \ccar#1  ->    \ccar#2        (caption, content)
      Separation tags (such as \par) will be phased out from the
      zoem macro packages. \item and \sec are to follow.

   *  \formatted#1 now skips \<> scope as well.

   *  Embedded newlines in *ml syntactic sugar \<..> would screw up the
      line count. Now fixed.

   *  \<> stacks are now tied to files, so that alternating output
      can never erroneously corrupt the well-formedness stack. (Do note that
      funny user-side inverse processing could and still can do that).

   *  Environment stacks (\begin and \end stuff) are now also tied to files,
      same as above.

   *  The hash module in the underlying library code was overhauled.
      hashes now maintain private storage for links and key-value pointers.

Tue, 20 Apr 2004

   *  zoem-04-111 released.

   *  small documentation fixes.


Mon, 19 Apr 2004

   *  zoem-04-110 released.

   *  the builtin alias \error#1 is gone, replaced by \inform#1.

   *  Streamlined the examples, made them more suitable for web-publishing.

   *  Fixed a bug recently introduced into the util library;
      file read from STDIN would only read a single line.

   *  (updated web index, otherwise irrelevant).

Sat, 17 Apr 2004

   *  zoem-04-108 released.

   *  Some activity at meta-zoem primitive level.
      A new exception mechanism was fully integrated with the error
      framework.

      -  Introduced \throw#2, which generalizes \quit (now removed).
      -  Removed \zoem#1. It is generalized by \catch{label}{expression}
         and \try{expression}.
      -  Reimplemented \quit as \done, specialized for clean premature
         file exit.

      This enables

         \catch{towel}{
            \while{1}{
               do stuff
               \if{ifstuff}{\throw{towel}}{}
               do stuff
            }
         }

      and
      
         \catch{error}{
            do possibly erroneous *or* towel stuff
         }

      or

         \try{ stuff }

         inspect
            \__zoemstat__ (towel|error|ok) and
            \__zoemput__  (the possibly truncated result of \try{stuff})

      \catch#2 and \try#2 can be arbitrarily nested.
         
   *  The primitive \throw#2 is used as
         \throw{towel}{msg}
         \throw{error}{msg}

      The macros
         \throw{towel}
         \throw{error}

      expand to \throw{towel}{} and \throw{error}{} and do not issue a message.

   *  The new alias \error#1 generates an error message, it expands to
      to \write{stderr}{device}{\1\@{\N}}

   *  Removed alias \ignore#1, use \""#1 instead.
      (e.g. \""{ignore
               stuff
            }
      )

   *  Fixed zum.azm, which contained bogus begin#1 description.

   *  Fixed small bug; \write#3 would erroneously set \__fnout__.
      Using \__fnout__ as in \write{\__fnout__}{filter}{stuff}
      now works as expected.

Wed, 7 Apr 2004

   *  zoem-04-098 released.

   *  fixed documentation to exclusively use/describe the begin#2 primitive.

   *  The old and ugly \begin{itemize{{foo}{bar}}} syntax now generates
      a syntax error. (use \begin{itemize}{{foo}{bar}}, introduced
      in zoem-04-072).

   *  fixed -e option related bug, changed its implementation.

   *  Introduced \*{'e} syntax, equivalent with \*'e* syntax. The new syntax
      is optionally extendible and more in line with the rest of zoem.
      This move is similar to the inline file name change \=fname= -> \={fname}
      introduced earlier.  The old \*'e* syntax will not be removed but
      if used error messages are issued.

   *  The \formatted#1 tokens \`<` and \`>` were changed to \`[` and \`]` for
      esthetic reasons.

Tue, 30 Mar 2004

   *  zoem-04-090 released.

   *  Fixed some small documentation glitches.

   *  At request of the zoem user community, the faqsec environment
      was moved to a key/value based syntax. The two positional arguments
      <label> (first) and <caption> (second) are now tied to
      values 'ref' and 'cap'. So now you write

         \begin{faqsec}{{ref}{abc}{cap}{Any Balding Clown}}

      which was previously

         \begin{faqsec}{:{abc}{Any Balding Clown}}

Fri, 12 Mar 2004

   *  zoem-04-072 released.

   *  As an exceptional case, a syntax change (fix, really)
      was introduced that requires updating of user-space zoem files that use
      the faqsec environment from the FAQ package faq.zmm .  This is the only
      environment affected by the change in begin#1 parsing described below.

   *  Implemented chunked reads. By default, zoem will read entire files
      before processing them. If -chunk <num> is specified on the command
      line, zoem will try to output a chunk once its size exceeds <num> bytes.
      At that stage it will simply append lines until it finds itself in the
      outermost scope (outside of any block).  So, finally, you will be able
      to throw files of any size at zoem.  Do note the level restriction
      though; if a block (pair of balanced curlies) spans the entire file zoem
      will attempt to read the file in its entirety.

      Perhaps chunked reads will become the default in a later release.

   *  Added begin#2 macro (not a primitive). The vararg part of
      the begin#1 syntax \begin{itemize{{align}{left}}} is placed
      in a separate argument as it should have been from the
      beginning: \begin{itemize}{{align}{left}}. The old syntax
      is deprecated, as it is special-case and not in line with
      general zoem syntax principles.

      The old syntax is still available - in fact, begin#2
      is currently rewritten to the begin#1 syntax.
      The plan is to phase out begin#1 at some point in the future.

   *  By default, the second argument of begin#2 (or the vararg
      part of begin#1 for that matter) is parsed as a sequence of key-value
      pairs which are set in the newly pushed dollar scope.

         \begin{foo}{{bar}{zut}{tim}{eek}}

      thus results in these keys and values:

         \$bar -> zut
         \$tim -> eek
         \$0   -> 2     \: contains the number of key-value pairs.

      The old behaviour (in which pseudo positional parameters
      \$1, \$2 and onwards are set) can be obtained by inserting a colon.

         \begin{foo}{:{bar}{zut}{tim}{eek}}

      results in these keys and values:

         \$1   -> bar
         \$2   -> zut
         \$3   -> tim
         \$4   -> eek
         \$0   -> 4     \: the number of arguments.

      As a bonus, you can now set arbitrarily many pseudo positional
      parameters.

Sat, 6 Mar 2004

   *  zoem-04-066 released.

   *  Changed format parsing so that parts may occur in any order.
      A new part *{..}{..} may be used to specify virtual length.
      A new part ~{..}{..} may be used to specify padding and delimiter.
      A new part @{..}{..} may be used to align on substring/offset.

      <=>            alignment specification; left right centered.
      ~{..}{..}      padding specification
      @{..}{..}      alignment specification; substring + offset.
      *{..}{..}      length macro + glyph string specification.
      <num>

   *  Added \lenght#1 which computes the length of its argument.
      It can be useful in the powerful new format virtual length
      functionality.

   *  Added *{key-name}{target-string} pretense part to format specification.
      Key-name specifies a macro computing a length (a likely candidate
      is \length#1 which is specified simply as length). Target-string will be
      used to compute the length, but the actual argument from the format
      vararg will be inserted. This allows some degree of separation between
      visible width of glyph arrays and their representation.

   *  Added introspective \zoem#1 primitive. The content will be processed
      and if this succeeds output is written in the macro \__zoemput__.  The
      result text is "1" for succesful and "0" for unsuccesful processing.

   *  Added \vanish#1, which processes it content but does not output
      anything. It is an alternative to \formatted#2.
      \vanish#1 allows to do lots of commenting outside definitions.
      \formatted#2 allows to format definitions themselves.

Sun, 29 Feb 2004

   *  zoem-04-061 released.

   *  Added \format#2 primitive.
      It can separately left align, center, or right align different
      arguments from a sequence.  First argument is the format string, second
      argument is a vararg.

   *  \writeto#1 (which can redirect the default output stream), if
      invoked more than once for the same file name, will append
      from the second time onwards, rather than overwrite.

   *  Fixed bug in switch#2 and branch#1; in the absence of a default-
      or else-type block, the last block would be evaluated regardless
      of the truth value of its pivot block.

   *  Fixed various documentation flaws pointed out by Joost van Baal (thx!).

   *  The obscure \%% syntax (providing freeing of multidimensional data)
      was changed to \%free, the syntax \%%% (providing dumping of
      multidimensional data) was changed to \%dump.

   *  Implemented --allow=cmd1[:cmd2]* option, allowing explicit
      specification of system commands that can be executed without asking.

Mon, 27 Oct 2003

   *  zoem-03-300 released. We are spending day 300 in 2003 people!

   *  added -tl k [tablength] option for regulating (optional) indent.

   *  Introduced zoemzoem manual page which lists the available
      zoem documentation.

   *  Fixed DOCTYPE declaration for html output, it contained stupid
      whitespace/case errors.

   *  Fixed some XML/HTML syntactic sugar bugs.

   *  Introduced the concept of alpha-stage new syntax.
      alpha-stage syntax is volatile and subject to change.

   *  \: is now generally described as introducing a pre-processing sequence;
      Its main use (and with syntax obeying the principle of Huffman
      encoding) is that of introducing comments.

   *  Retro-actively declared the recently introduced \:\ syntax
      to be alpha-stage, and removed it.

      Subtle errors could arise when commenting out
      macro sequences \foo as \:\foo (removing the newline as well).
      \:\ will never be extended preprocessing syntax again.

      The functionality previously provided by \:\ is now provided by \:/
      (introducing a comment up till and including the newline).

   *  Added alpha-stage \<foobar!> syntactic sugar for writing HTML-style
      single tags such as <br>. Now you can write
      \<br!> rather than \@{<br>} -- both result in <br>.
      It's alpha stage because perhaps another character might be better.
      Opinions, anyone?

   *  Added \:! preprocessing sequence, and the following rule:
      \: followed by whitespace or alphanumerical character or backslash will
      always be a regular comment up till and excluding the newline.

      \:! will be replaced during preprocessing time by a '\' (backslash).
      One use is this:
      \verbatim{\protect{\foo
         \bar
         \:!:  this will end up as a comment.
      }}

Sun, 12 Oct 2003

   *  zoem-03-285 released.

   *  Disabled (by default) \system#3, implemented the --unsafe,
      --unsafe-silent, and --system-honor options.

      With --unsafe, the user is prompted for each \system#3 invocation.
         If prompting is not possible (due to STDIN redirection),
         the primitive fails.
      With --unsafe-silent, the user is not prompted.
      With --system-honor, any \system#3 failure (for whatever reason,
         including safe behaviour) is regarded as a zoem failure.
         Without --system-honor (i.e. by default), \system#3 failures are
         ignored in all modes: safe, unsafe, and unsafe-silent.

   *  applied/fixed a number of fixes/issues brought forth by Andreas Kahari.

   *  added \dowhile#2 primitive.

   *  added \branch#1 primitive, which employs a vararg.
      \branch{
         {condition1}{branch1}
         {condition2}{branch2}
           ...
         {conditionN}{branchN}
         {else branch}     \: optional
      }

   *  fixed inspect#4 segfaulting bug, and inspect#4 functional bug.

   *  documented the man macros \genkvp#2 \usekvp#2 \defkvp#3 in man_zmm(7).

   *  Added \"man::cat" which can optionally be used to set
      the category heading for manual pages. If not set, a default
      is derived from the \"man::section" macro.

Mon, 22 Sep 2003

   *  Introduced NAME#2 macro; using this will ensure that the NAME
      section is apropos compatible.

   *  groff-base 1.18.1-10 has an entirely new idea of what a reasonable
      right margin is (no less than 2 positions wide). Since zoem adjusted
      the line lenght a little, this caused previous zoem output
      to have no right margin at all. Consequently, zoem now decreases
      the output line length by 2.

   *  \:\ now introduces a comment that will be stripped up to and including
      the newline. All other \: sequences introduce a comment that will
      be stripped up to and excluding the newline.

   *  Eliminated irregular inline file syntax. It now behaves the
      same as other primitives taking an argument (the argument
      is delimited by balanced curlies).

Tue, 12 Aug 2003

   *  zoem-03-224 released

   *  Removed \%snt% syntax in \formatted; replaced by
      fully similar \`snt` syntax. It looks better (no :?) and the % syntax
      was needed for grape keys (the new name for multi-dimensional data).

   *  Replaced low-level at directives \# \^ \_ by single \+{} directive.

   *  Fixed bug in formerly \dfree#1, now \%%%.

   *  Removed dget#1 dset#2 dsetx#2 dfree#1 dprint#1
      Their syntax was assimilated by set#2 def#2 setx#2 defx#2
      so that more functional behaviour was obtained with less primitives.
      Assigning:

      \set{%{a}{b}}{d}.

      retrieval now done using
         \%{a}{b}

      removal done with
         \%%{a}{b} [ \%% for the full tree ]

      dumping done with
         \%%%{a}{b} [ \%%% for the full tree ]

   *  added \"html::charset" to the generic macros.

Tue, 03 Jun 2003

   *  zoem-03-166 released.

   *  began tracking changes in ChangeLog.