Jinja Changelog
===============

Version 1.3
-----------
(codename to be selected, release around February 2007)

Version 1.2
-----------
(codename to be hatsuyuki, released Nov 17th 2007)

.. admonition:: Backwards Incompatible Changes

    - `call` is a keyword now
    - the internal parser AST changed
    - `extends` must be the first tag in a template
    - the tuple literal yields tuples now, instead of lists.

- environments now have a `translator_factory` parameter that allows
  to change the translator without subclassing the environment.

- fixed bug in buffet plugin regarding the package loader

- once again improved debugger.

- added `groupby` filter.

- added `sameas` test function.

- standalone parser. Jinja does not use the python parser any more and will
  continue having the same semantics in any future python versions. This
  was done in order to simplify updating Jinja to 2.6 and 3.0 and to support
  non python syntax.

- added support for ``expr1 if test else expr2`` (conditional expressions)

- ``foo.0`` as alias for ``foo[0]`` is possible now. This is mainly for
  django compatibility.

- the filter operators has a much higher priority now which makes it
  possible to do ``foo|filter + bar|filter``.

- new AST. the return value of `Environment.parse` is now a Jinja AST and not
  a Jinja-Python AST. This is also the only backwards incompatible change but
  should not affect many users because this feature is more or less
  undocumented and has few use cases.

- tuple syntax returns tuples now and not lists any more.

- the print directive and ``{{ variable }}`` syntax now accepts and implicit
  tuple like the `for` and `cycle` tags. (``{{ 1, 2 }}`` is an implicit alias
  for ``{{ (1, 2) }}` like ``{% for a, b in seq %}`` is for
  ``{% for (a, b) in seq %}``.

- tests called with *one* parameter don't need parentheses. This gives a more
  natural syntax for the `sameas` test and some others:
  ``{{ foo is sameas bar }}`` instead of ``{{ foo is sameas(bar) }}``. If you
  however want to pass more than one argument you have to use parentheses
  because ``{{ foo is sometest bar, baz }}`` is handled as
  ``{{ (foo is sometest(bar), baz) }}``, so as tuple expression.

- removed support for octal character definitions in strings such as
  ``'\14'``, use ``'\x0c'`` now.

- added regular expression literal. ``@/expr/flags`` equals
  ``re.compile(r'(?flags)expr')``. This is useful for the `matching` test and
  probably some others.

- added set literal. We do not use python3's {1, 2} syntax because
  this conflicts with the dict literal. To be compatible with the regex
  literal we use ``@(1, 2)`` instead.

- fixed bug in `get_attribute` that disallowed retreiving attributes of objects
  without a `__class__` such as `_sre.SRE_Pattern`.

- addded `django.contrib.jinja` which provides advanced support for django.
  (thanks Bryan McLemore)

- debugger is now able to rewrite the whole traceback, not only the first
  frame. (requires the optional debugger c module which is compiled
  automatically on installation if possible)

- if the set that is postfixed with a bang (!) it acts like the python 3
  "nonlocal" keyword. This means that you can now override variables
  defined in the outer scope from within a loop.

- ``foo ~ bar`` is now a simpler alternative to ``foo|string + bar|string``

- `PackageLoader` can now work without pkg_resources too

- added `getattribute` and `getitem` filter.

- added support for the `pretty` library.

- changed the way the `MemcachedLoaderMixin` creates the class so that it's
  possible to hook your own client in.


Version 1.1
-----------
(codename: sinka, released Jun 1, 2007)

- blocks now support ``{{ super() }}`` to render the parent output.

- debugging system improved, smaller filesize for the cached files.
  Debugging works now well for any module using linecache.

- ``{{ debug() }}`` can now be used to get a list of filters and
  tags.

- the template lexer keeps not track of brace, parenthesis and
  bracket balance in order to not break variable tags apart if they
  are configured to look like this: ``${expr}``. This also fixes
  the problem with nested dicts in variable expressions.

- it's now possible to configure the variable blocks to look the
  same as the block delimiters. Thus you can set the syntax to something
  like ``{ / }`` for both blocks and variables.

- added whitespace management system for the template designer.

- some small bugfixes.

- improved security system regarding function calls and variable
  assignment in for loops.

- added `lipsum` function to generate random text.

- strings without unicode characters are processed as binary strings now
  to workaround problems with `datetime.strftime` which only accepts
  binary strings.

- it's now possible to use newlines in string literals

- developer friendly traceback is now toggleable

- the variable failure is now pluggable by replacing the undefined
  singleton for an environment instance

- fixed issue with old-style classes not implementing `__getitem__`
  (thanks to Axel Böhm for discovering that bug)

- added a bunch of new docstrings to the Jinja classes. Makes fun now to
  use pydoc :-)

- fixed severe memcaching bug. Formerly it wasn't possible to use memcaching
  without enabling disk cache.

- fixed a bug that allowed users to override the special names `_`, `true` etc.

- added `batch` and `slice` filters for batching or slicing sequences

- added `sum`, `abs`, `round` and `sort` filters. This fixes #238

- added `striptags` and `xmlattr` filters for easier SGML/XML processing

- the trans tag does not need explicit naming for variables with the same
  name any more. You can now use ``{% trans foo %}`` instead of the verbose
  version ``{% trans foo=foo %}``.

- reimplemented Buffet plugin so that it works at least for pylons

- added `Environment.get_translations_for_string`

- fixed a bug in the parser that didn't unescape keyword arguments. (thanks
  to Alexey Melchakov for reporting)

- You can now use the environment to just tokenize a template. This can
  be useful for syntax highlighting or other purposes.

- added optional C-implementation of the context baseclass.

- you can now use optional parentheses around macro defintions. Thus it's
  possible to write ``{% macro foo(a, b, c) %}`` instead of ``{% macro
  foo a, b, c %}``.

- additional macro arguments now end up in `varargs`.

- implemented the `{% call %}` block. `call` and `endcall` can still be used
  as identifiers until Jinja 1.3

- it's now possible to stream templates.

- fixed a corner case when defining a block inside of a condition

- the cached loader mixin is now able to cache multiple templates from
  different loaders in the same cache folder.

- Translatable strings returned by ``_()`` will leave their string formatting
  signs untouched. Thanks to Stefan Ebner for reporting.

- ``{% block name "data" %}`` is now an alias for
  ``{% block name %}data{% endblock %}``. Note that the second argument can
  be an expression. As soon as you specify an expression as second argument
  the closing tag has to be omitted.

- It's now possible to iterate over iterators additionally to sequences.
  If the iterator is inifite it will crash however, so makes sure you don't
  pass something like that to a template!

- added `rendetemplate` to render included templates in an isolated
  environment and get the outout back.

- added `simplefilter` decorator.

- improved ChoiceLoader error reporting (thanks to Bryan McLemore)

- fixed extended slicing

- reworked loader layer. All the cached loaders now have "private" non cached
  baseclasses so that you can easily mix your own caching layers in.

- added `MemcachedLoaderMixin` and `MemcachedFileSystemLoader` contributed
  by Bryan McLemore.


Version 1.0
-----------
(codename: siyutusan, released Mar 23, 2007)

- Initial release