Cheetah 2.0rc8

Cheetah is an open source template engine and code generation tool.

It can be used standalone or combined with other tools and frameworks. Web
development is its principle use, but Cheetah is very flexible and is also being
used to generate C++ game code, Java, sql, form emails and even Python code.

Documentation
================================================================================
For a high-level introduction to Cheetah please refer to the User's Guide
at http://cheetahtemplate.org/learn.html

Mailing list
================================================================================
cheetahtemplate-discuss@lists.sourceforge.net
Subscribe at http://lists.sourceforge.net/lists/listinfo/cheetahtemplate-discuss

Credits
================================================================================
http://cheetahtemplate.org/credits.html

Praise
================================================================================
"I'm enamored with Cheetah" - Sam Ruby, senior member of IBM Emerging
Technologies Group & director of Apache Software Foundation

"Give Cheetah a try. You won't regret it. ... Cheetah is a truly powerful
system. ... Cheetah is a serious contender for the 'best of breed' Python
templating." - Alex Martelli

"People with a strong PHP background absolutely love Cheetah for being Smarty,
but much, much better." - Marek Baczynski

"I am using Smarty and I know it very well, but compiled Cheetah Templates with
its inheritance approach is much powerful and easier to use than Smarty." -
Jaroslaw Zabiello

"There is no better solution than Cheetah" - Wilk

"A cheetah template can inherit from a python class, or a cheetah template, and
a Python class can inherit from a cheetah template. This brings the full power
of OO programming facilities to the templating system, and simply blows away
other templating systems" - Mike Meyer

"Cheetah has successfully been introduced as a replacement for the overweight
XSL Templates for code generation. Despite the power of XSL (and notably XPath
expressions), code generation is better suited to Cheetah as templates are much
easier to implement and manage." - The FEAR development team
(http://fear.sourceforge.net/docs/latest/guide/Build.html#id2550573)

"I've used Cheetah quite a bit and it's a very good package" - Kevin Dangoor,
lead developer of TurboGears.

Recent Changes
================================================================================
See http://cheetahtemplate.org/docs/CHANGES for full details.

Please initial your changes (there's a key at bottom) and add a date for each
release
================================================================================

2.0rc8 (April 11, 2007)
!!!THIS RELEASE REQUIRES RECOMPILATION OF ALL COMPILED CHEETAH TEMPLATES!!!
Core Changes: [TR]

- added a '#unicode <encoding>' directive to indicate that the output of the
template should be a unicode string even if the template source is a
normal byte string.

- #unicode and #encoding are mutually exclusive. Use one or the other.
- #unicode must be on a line by itself.
- Strings in embedded code must be explictly marked as unicode if they
contain non-ascii chars:

#unicode latin-1
$f(u"<some non-ascii char>") ## right
$f("<some non-ascii char>") ## wrong

However, this works fine:

#unicode latin-1
blah blah <some non-ascii char> blah blah

- fixed several unicode bugs in the compiler.

- fixed some unicode issues in the standard filters.

- fixed a few minor bugs in code that never gets called. Thanks to
Alejandro Dubrovsky for pointing them out.

- make RawOrEncodedUnicode the baseclass of all filters and remove some
unused/redudant filters

- added new compiler setting 'addTimestampsToCompilerOutput'. See Brian
Bird's post about it. He stores his cheetah generated .py files in
subversion and needed to disable the timestamp code so svn wouldn't care
when he recompiles those .py modules.

- added the #super directive, which calls the method from the parent class
which has the same as the current #def or #block method.

#def foo
... child output
#super ## includes output of super(<CurrentClass>, self).foo()
... child output
#end def


#def bar(arg)
... child output
#super(arg) ## includes output of super(<CurrentClass>, self).bar(arg)
... child output
#end def

- added some unit tests for the new directives


2.0rc7 (July 4, 2006)
!!!THIS RELEASE REQUIRES RECOMPILATION OF ALL COMPILED CHEETAH TEMPLATES!!!
Core Changes: [TR]
- extended the #implements directive so an arguments list can be declared in
the same fashion as #def and #block.

- made the parser raise ParseError when $*placeholder, $*5*placeholder,
$(placeholder), etc. are found within expressions. They are only valid in
top-level text.

- tweaked the parser so it's possible to place a comment on the same line as
a directive without needing to explicitly close the directive first. This
works regardless of whether or not you added a colon.

self.verify("#if 1:\n$aStr\n#end if\n",
"blarg\n")

self.verify("#if 1: \n$aStr\n#end if\n",
"blarg\n")

self.verify("#if 1: ##comment \n$aStr\n#end if\n",
"blarg\n")

self.verify("#if 1 ##comment \n$aStr\n#end if\n",
"blarg\n")

Previously, that last test would have required an extra # to close the #if
directive before the comment directive started:
self.verify("#if 1 ###comment \n$aStr\n#end if\n",
"blarg\n")

Code that makes use of explicit directive close tokens immediately
followed by
another directive will still work as expected:
#if test##for i in range(10)# foo $i#end for##end if

- safer handling of the baseclass arg to Template.compile(). It now does
the right thing if the user passes in an instance rather than a class.

ImportHooks: [TR]
- made it possible to specify a list of template filename extentions that are
looped through while searching for template modules. E.g.:
import Cheetah.ImportHooks
Cheetah.ImportHooks.install(templateFileExtensions=('.tmpl','.cheetah'))

Core changes by MO:
- Filters are now new-style classes.
- WebSafe and the other optional filters in Filters.py now use
RawOrEncodedUnicode instead of Filter as a base class. This allows them
to work with Unicode values containing non-ASCII characters.
User-written custom filters should inherit from
RawOrEncodedUnicode and call the superclass .filter() instead of str().
str() as of Python 2.4.2 still converts Unicode to string using
ASCII codec, which raises UnicodeEncodeError if it contains non-ASCII
characters.

2.0rc6 (Feb 4, 2006)
!!!THIS RELEASE REQUIRES RECOMPILATION OF ALL COMPILED CHEETAH TEMPLATES!!!
Core Changes: [TR]
- added a Cheetah version dependency check that raises an assertion if a
template was compiled with a previous version of Cheetah whose templates
must be recompiled.

- made the Cheetah compilation metadata accessible via class attributes in
addition to module globals

- major improvement to exception reporting in cases where bad Python syntax
slips past the Cheetah parser:
"""
File "/usr/lib/python2.4/site-packages/Cheetah/Template.py", line 792,
in compile
raise parseError
Cheetah.Parser.ParseError:

Error in the Python code which Cheetah generated for this template:

================================================================================

invalid syntax (DynamicallyCompiledCheetahTemplate.py, line 86)

Line|Python Code
----|-------------------------------------------------------------
84 |
85 | write('\n\n')
86 | for i an range(10): # generated from line 4, col 1
^
87 | _v = i # '$i' on line 5, col 3
88 | if _v is not None: write(_filter(_v, rawExpr='$i')) #
from line 5, col 3.
89 | write('\n')


================================================================================

Here is the corresponding Cheetah code:

Line 4, column 1

Line|Cheetah Code
----|-------------------------------------------------------------
2 |#compiler useNameMapper=False
3 |
4 |#for i an range(10)
^
5 | $i
6 |#end for
7 |
"""

2.0rc5 (Feb 3, 2006)
!!!THIS RELEASE REQUIRES RECOMPILATION OF ALL COMPILED CHEETAH TEMPLATES!!!
Core Changes: [TR]
- fixed a memory leak in Template.compile(), reported by Andrea Arcangeli
- simplified concurrency locking and compile caching in Template.compile()

The command line tool (CheetahWrapper.py):
- added new option --settings for supplying compiler settings
- added new option --templateAPIClass to replace the environment var
CHEETAH_TEMPLATE_CLASS lookup I added in 2.0b1

2.0rc4 (Jan 31, 2006)
!!!THIS RELEASE REQUIRES RECOMPILATION OF ALL COMPILED CHEETAH TEMPLATES!!!
Core Changes: [TR]
- fixed a typo-bug in the compile hashing code in Template.compile()
- improved the macros framework and made it possible to implement macros in
Python code so they can be shared between templates
- more work on the #i18n directive. It's now a macro directive.
- added new Cheetah.Macros package
- more tests

2.0rc3 (Jan 29, 2006)
!!!THIS RELEASE REQUIRES RECOMPILATION OF ALL COMPILED CHEETAH TEMPLATES!!!
Core Changes: [TR]
- added short-form single line versions of all directives that have an #end
tag, except for #errorCatcher:
#if, #else, #elif, #unless,
#for, #while, #repeat,
#try, #except, #finally,
#cache, #raw
#call, #capture

The #def and #block directives already had single-line versions.
#if cond: foo
#elif cond2: bar
#else: blarg

#for i, val in enumerate(vals): $i-$val

Note that if you accidentally leave a colon at the end of one of these
directives but nothing else follows it, aside from whitespace, the parser
will treat it as a normal multi-line directive.

The first leading space after the colon is discarded. Any additional
spaces will be included in the output.

Also note, if you use the short form versions of #if/#else/#elif you must
it for all three. The following is not valid:
#if cond: foo
#elif cond2
bar
#else: blarg

- added support for $!silentModePlaceholders
This is the same as quiet mode in Velocity:

http://jakarta.apache.org/velocity/docs/user-guide.html#Quiet%20Reference%20Notation

- added support for function/method @decorators. It also works with blocks.
As in vanilla Python, the @decorator statement must be followed with a
function/method definition (i.e. #def or #block).

#from xxx import aDecorator
...
...
#@aDecorator
#def func
foo
#end def

#@aDecorator
#def singleLineShortFormfunc: foo

#@aDecorator
#block func2
bar
#end block

- added a new callback hook 'handlerForExtendsDirective' to the compiler
settings. It
can be used to customize the handling of #extends directives. The
callback can dynamically add import statements or rewrite the baseclass'
name if needed:
baseClassName = handler(compiler, baseClassName)
See the discussion on the mailing list on Jan 25th for more details.

- changed the default filter to the one that doesn't try to encode Unicode
It was 'EncodeUnicode' and is now 'RawOrEncodedUnicode'.

- added optional support for parsing whitespace between the directive start
token (#) and directive names, per Christophe Eymard's request. For the
argument behind this see the mailing list archives for Jan 29th. This is
off by default. You must turn it on using the compiler setting
allowWhitespaceAfterDirectiveStartToken=True

#for $something in $another
# for $somethin2 in $another2
blahblah $something in $something2
# end for
#end for

- made the handling of Template.compile()'s preprocessors arg simpler and
fixed a bug in it.

- fixed attribute name bug in the .compile() method (it affected the feature
that allows generated module files to be cached for better exception
tracebacks)

- refactored the #cache/CacheRegions code to support abitrary backend cache
data stores.

- added MemcachedCacheStore, which allows cache data to be stored in a
memcached backend. See http://www.linuxjournal.com/article/7451 and
http://www.danga.com/memcached/. This is only appropriate for systems
running many Python server processes that need to share cached data to
reduce memory requirements. Don't bother with this unless you actually
need it. If you have a limited number of Python server processes it is
much faster, simpler, and more secure to just cache in the memory of each
process.

KEEP MEMCACHED'S LIMITED SECURITY IN MIND!! It has no authentication or
encryption and will introduce a gaping hole in your defenses unless you
are careful. If you are caching sensitive data you should take measures
to ensure that a) untrusted local system users cannot connect to memcached
server, b) untrusted external servers cannot connect, and c) untrusted
users on trusted external servers cannot connect. Case (a) can be dealt
with via iptable's owner match module for one way to do this: "iptables -A
... -m owner ..." Cases (b) and (c) can be handled by tunnelling
memcached network connections over stunnel and implementing stunnel
authentication with mandatory peer/client certs.

- some under-the-hood refactoring of the parser

- made it possible to add custom directives, or customize the
parsing/handling of existing ones, via the compiler settings
'directiveNamesAndParsers' and 'endDirectiveNamesAndHandlers'

- added a compile-time macro facility to Cheetah. These macros are very
similar to macros in Lisp:
http://www.apl.jhu.edu/~hall/Lisp-Notes/Macros.html.

As with Lisp macros, they take source code (Cheetah source) as input and
return source code (again Cheetah source) as output. They are executed at
compile-time, just like in Lisp and C. The resultant code
gets executed at run-time.

The new #defmacro directive allows users to create macros inside the
source of their templates. Macros can also be provided via the compiler
setting 'macroDirectives'. The 'macroDirectives' setting allows you to
share common macros between templates.

The syntax for the opening tag of #defmacro is the same as for #def and
#block. It expects a macro name followed by an optional argument list in
brackets. A `src` argument is automatically added to the beginning of
every macro's argument list. The value of the `src` is the block of
input source code that is provided during a macro call (see below).

#defmacro <macroname>[(argspec)]
<macrobody>
#end defmacro

All of Cheetah's syntax is available for use inside macros, but the
placeholderStartToken is @ instead of $ and the
directiveStartToken/EndToken is % instead of #. Any syntax using the
standard $/# tokens will be treated as plain text and included in the output
of the macro.

Here are some examples:
#defmacro addHeaderFooter
header
@src
footer
#end defmacro

#defmacro addHeaderFooter(header='h', footer='f')
@header
@src
@footer
#end defmacro

There is a single-line short form like for other directives:

#defmacro addHeaderFooter: header @src footer
#defmacro addHeaderFooter(header='h', footer='f'): @header @src @footer

The syntax for calling a macro is similar to the simplest usage of the
#call directive:

#addHeaderFooter
Source $code to wrap
#end addHeaderFooter

#addHeaderFooter: Source $code to wrap

#addHeaderFooter header='header', footer='footer: Source $code to wrap


In Elisp you write
(defmacro inc (var)
(list 'setq var (list '1+ var)))
to define the macro `inc` and write
(inc x)
which expands to
(setq x (1+ x))

In Cheetah you'd write
#defmacro inc: #set @src +=1
#inc: $i
which expands to
#set $i += 1

print Template("""\
#defmacro inc: #set @src +=1
#set i = 1
#inc: $i
$i""").strip()==2

- fixed some bugs related to advanced usage of Template.compile(). These
were found via new unit tests. No one had actually run into them yet.

- added the initial bits of an #i18n directive. It has the same semantics
as
#call self.handleI18n
Some $var cheetah source
#end call
but has a simpler syntax:
#i18n
Some $var cheetah source
#end i18n

## single-line short form:
#i18n: Some $var cheetah source

The method it calls, self.handleI18n, is just a stub at the moment, but it
will soon be a wrapper around gettext. It currently has one required
positional argument `message`. I anticipate supporting the following
optional arguments:

id = msgid in the translation catalog
domain = translation domain
source = source lang
target = a specific target lang
comment = a comment to the translation team

plural = the plural form of the message
n = a sized argument to distinguish between single and plural forms

#i18n is executed at runtime, but it can also be used in conjunction with
a Cheetah preprocessor or macro (see above) to support compile time
translation of strings that don't have to deal with plural forms.

- added Cheetah.Utils.htmlEncode and Cheetah.Utils.htmlDecode

- more docstring text

Unit tests: [TR]
- extended the caching tests
- added tests for the various calling styles of Template.compile()
- added copies of all the SyntaxAndOutput tests that use a template
baseclass other than `Template`. This ensures that all syntax & core
features work with 2.0's support for arbitrary baseclasses.
- added tests for all the new directives and the new single-line short forms

2.0rc2 (Jan 13th, 2006)
!!!THIS RELEASE REQUIRES RECOMPILATION OF ALL COMPILED CHEETAH TEMPLATES!!!
Core Changes: [TR]
- fixed some python 2.4isms that slipped in. All the tests pass with Python
2.2 now
- added lots more docstring content in the Template class
- made multiline comments gobble whitespace like other directives, per JJ's
request. The rather longwinded compiler setting
gobbleWhitespaceAroundMultiLineComments can be used to go back to the old
non-gobbling behaviour if needed.
- added #capture directive to complement the #call directive.
#call executes a region of Cheetah code and passes its output into a
function call
#capture executes a region of Cheetah code and assigns its output to a
variable
- extended the compile caching code in Template.compile so it works with the
'file' arg.
- added cacheModuleFilesForTracebacks and cacheDirForModuleFiles args to
Template.compile(). See the docstring for details.
- misc internal refactoring in the parser
- improved handling of keyword args in the __init__ method and fixed a
potential clash between the namespaces and searchList args

WWW: [TR]
- added the source for the new Cheetah website layout/content

2.0rc1 (Jan 10, 2006)
!!!THIS RELEASE REQUIRES RECOMPILATION OF ALL COMPILED CHEETAH TEMPLATES!!!
Core Changes: [TR]
- made it possible nest #filter directives
- added lots more docstring content in the Template class
- added Template.subclass() classmethod for quickly creating subclasses of
existing Cheetah template classes. It takes the same args as the
.compile() classmethod and returns a template that is a subclass of the
template .subclass() is called from:
T1 = Template.compile(' foo - $meth1 - bar\n#def meth1: this is T1.meth1')
T2 = T1.subclass('#implements meth1\n this is T2.meth1')

- added baseclass arg to Template.compile(). It simplifies the reuse of
dynamically compiled templates:
# example 1, quickly subclassing a normal Python class and using its
# __init__ call signature:
dictTemplate = Template.compile('hello $name from $caller', baseclass=dict)
print dictTemplate(name='world', caller='me')

# example 2, mixing a Cheetah method into a class definition:
class Foo(dict):
def meth1(self):
return 'foo'
def meth2(self):
return 'bar'
Foo = Template.compile('#implements meth3\nhello $name from $caller',
baseclass=Foo)
print Foo(name='world', caller='me')

A side-benefit is the possibility to use the same Cheetah source with
several baseclass, as the baseclass is orthogonal to the source code,
unlike the #extends directive.

- added 'namespaces' as an alias for 'searchList' in Template.__init__
- made it possible to pass in a single namespace to 'searchList', which will
automatically be converted into a list.
- fixed issue with buffering and use of #call when template is used as a
webkit servlet
- added Cheetah.Utils.htmlEncode and htmlDecode

The command line tool (CheetahWrapper.py):
- changed insertion order for the --env and --pickle options so they match the
commandline UI of the compiled template modules themselves [TR]

2.0b5 (Jan 7, 2006)
!!!THIS RELEASE REQUIRES RECOMPILATION OF ALL COMPILED CHEETAH TEMPLATES!!!
Core Changes: [TR]
- made Cheetah.Template a new-style class by inserting 'object' into its'
inheritance tree. Templates can now use super(), properties and all the
other goodies that come with new-style classes.
- removed the WebInputMixin by placing its one method directly in the
Template class.
- removed the SettingsManager Mixin. It wasn't being used by anything
anymore.
- added a framework for caching the results of compilations in
Template.compile(). This is on by default and protects against bad
performance issues that are due to programmers misguidedly compiling
templates inside tight loops. It also saves on memory usage.
- misc attr name changes to avoid namespace pollution
- more + improved docstrings
- replaced the oldstyle dynamic compile hacks with a wrapper around
Template.compile(). The old usage pattern Template(src) now benefits from
most of the recent changes.
Template(src).__class__ == Template.compile(src)
- removed all the extra imports required by oldstyle dynamic compile hacks
- converted the cheetah #include mechanism to newstyle compilation and made it
more flexible
- made the #include mechanism work with file objects in addition to file names
- made the handling of args to Template.compile() more flexible. You can now
provide defaults via class attributes.
- made preprocessors for Template.compile() work with file arguments
- added support for specifying a __metaclass__ on cheetah template classes
- refactored both the class and instance initialization processes
- improved the handling of __str__ in _assignRequiredMethodsToClass

The command line tool (CheetahWrapper.py): [TR]
- improved error output in CheetahWrapper
- switched fill command over to new style compile usage

Unit tests: [TR]
- fixed format string bug in unittest_local_copy.py

2.0b4 (Jan 6, 2006)
!!!THIS RELEASE REQUIRES RECOMPILATION OF ALL COMPILED CHEETAH TEMPLATES!!!
Core Changes: [TR]
- fixed up parsing of target lists in for loops. This was previously limited
to fairly simple target lists.
#for ($i, $j) in [('aa','bb'),('cc','dd')]
$i.upper,$j.upper
#end for"
#for (i, j) in [('aa','bb'),('cc','dd')]
$i.upper,$j.upper
#end for"
#for i,(j, k) in enumerate([('aa','bb'),('cc','dd')])
$j.upper,$k.upper
#end for"
- refactored the class initialization process
- improved handling of target lists in #set directive. This was previously
limited to fairly simple target lists.
#set i,j = [1,2] ... #set $i,$j = [1,2]
#set (i,j) = [1,2] ... #set ($i,$j) = [1,2]
#set i, (j,k) = [1,(2,3)] ... #set $i, ($j,$k) = [1,(2,3)]

- made it possible for the expressionFilter hooks to modify the code chunks
they are fed. Also documented the hooks in a docstring. Thus the hooks
can be used as preprocessors for expressions, 'restricted execution', or
even enforcement of style guidelines.

- removed cheetah junk from docstrings and placed it all in comments or
__moduleVars__. Per JJ's suggestion.

- made it possible to nest #cache directives to any level
- made it possible to nest #call directives to any level

Unit Tests [TR]
- extended tests for #for directive
- expanded tests for #set directive
- expanded tests for #call directive
- expanded tests for #cache directive
- added basic tests for the new $placeholder string expressions:
c'text $placeholder text'

2.0b3 (Jan 5, 2006)
!!!THIS RELEASE REQUIRES RECOMPILATION OF ALL COMPILED CHEETAH TEMPLATES!!!
Core Changes: [TR]
- added #yield statement
- added ability to create nested scopes/functions via nested #def statements
- added new #call directive and related #arg directive, per Ian Bicking's
suggestion.
- added new expression syntax c"text $placeholder text"

for those basic function calling cases where you just need to pass in a
small bit of cheetah output as an argument:

c'a string with $placeholders',
c'''a string with $placeholders''',
c"a string with $placeholders",
c"""a string with $placeholders"""

- They can't contain #directives, but accept any valid $placeholder syntax
except caching placeholders. Caching placeholders don't make any sense in
this context.
- They can be used *any* place where a python expression is expected.
- They can be nested to any depth.

$func(c'<li>$var1-$var2</li>')
$func(c'<li>$var1-$var2</li>', doSomething=True)
$func(content=c'<li>$var1-$var2</li>', doSomething=True)
$func(lambda x,y: c'<li>$x-$y</li>')
$func(callback=lambda x,y: c'<li>$x-$y</li>')
$func(lambda x,y: c'<li>$x-$y-$varInSearchList</li>')
$func(c'<li>$var1-$var2-$(var3*10)-$(94.3*58)</li>')
$func(c'<li>$var1-$var2-$func2(c"a nested expr $var99")</li>')
#if $cond then c'<li>$var1-$var2</li>' else c'<p>$var1-$var2</p>'
#def foo(arg1=c'$var1<span class="foo">$var2</span>'): blah $arg1 blah
$foo(c'$var1<i>$var2</i>')

- added preprocessor hooks to Template.compile()
can be used for partial completion or 'compile-time-caching'
... more details and examples coming. It's very useful, but takes a bit
of explaining.
- added '#set module varName = expr' for adding module globals. JJ's suggestion
- improved generated docstring notes about cached vars
- fixed silly bug related to """ in docstring comments and statements like
this '#def foo: $str("""foo""")'. Reported by JJ.
- changed the handling of single-line defs so that
'#def xxx:<just whitespace>\n' will be treated as a multi-line #def.
The same applies to #block. There's a compiler setting to turn this off
if you really need empty single-line #def:'s.
JJ reported that this was causing great confusion with beginners.
- improved error message for unclosed directives, per Mike Orr's suggestion.
- added optional support for passing the trans arg to methods via **KWS rather
than trans=None. See the discussion on the mailing list Jan 4th (JJ's
post) for
details. The purpose is to avoid a positional argument clash that
apparently is very confusing for beginners.

Note that any existing client code that passing the trans arg in
positionally rather than as a keyword will break as a result. WebKit
does this with the .respond method so I've kept the old style there.
You can also turn this new behaviour off by either manually including
the trans arg in your method signature (see the example below) or by
using the compiler setting 'useKWsDictArgForPassingTrans'=False.

#def manualOverride(arg1, trans=None)
foo $arg1
#end def

ImportHooks:
- made the ImportHook more robust against compilation errors during import [TR]

Install scripts: [TR]
- added optional support for pje's setuptools
- added cheeseshop classifiers
- removed out of date install instructions in __init__.py

Servlet Base Class For Webkit: [TR]
- disabled assignment of self.application (was a webware hack)

Unit Tests: [TR]
- unit tests for most of the new syntax elements
- tidied up some old tests
- misc refactoring

2.0b2 (Dec 30, 2005)
!!!THIS RELEASE REQUIRES RECOMPILATION OF ALL COMPILED CHEETAH TEMPLATES!!!

Core Changes:
- In previous versions of Cheetah tracebacks from exceptions that were raised
inside dynamically compiled Cheetah templates were opaque because
Python didn't have access to a python source file to use in the traceback:

File "xxxx.py", line 192, in getTextiledContent
content = str(template(searchList=searchList))
File "cheetah_yyyy.py", line 202, in __str__
File "cheetah_yyyy.py", line 187, in respond
File "cheetah_yyyy.py", line 139, in writeBody
ZeroDivisionError: integer division or modulo by zero

It is now possible to keep the generated source code from the python
classes returned by Template.compile() in a cache dir. Having these files
around allows Python to include the actual source lines in tracebacks and
makes them much easier to understand:

File "/usr/local/unsnarl/lib/python/us/ui/views/WikiPageRenderer.py",
line 192, in getTextiledContent
content = str(template(searchList=searchList))
File "/tmp/CheetahCacheDir/cheetah_yyyy.py", line 202, in __str__
def __str__(self): return self.respond()
File "/tmp/CheetahCacheDir/cheetah_yyyy.py", line 187, in respond
self.writeBody(trans=trans)
File "/tmp/CheetahCacheDir/cheetah_yyyy.py", line 139, in writeBody
__v = 0/0 # $(0/0)
ZeroDivisionError: integer division or modulo by zero

This is turned off by default. To turn it on, do this:

class NiceTracebackTemplate(Template):
_CHEETAH_cacheModuleFilesForTracebacks = True
_CHEETAH_cacheDirForModuleFiles = '/tmp/CheetahCacheDir' # change to
a dirname

templateClass = NiceTracebackTemplate.compile(src)

# or
templateClass = Template.compile(src,
cacheModuleFilesForTracebacks=True,
cacheDirForModuleFiles='/tmp/CheetahCacheDir')


This only works with the new Template.compile(src) usage style!

Note, Cheetah generated modules that are compiled on the command line have
never been affected by this issue. [TR]

- added an extra comment per $placeholder to generated python code so it is
easier to grok. [TR]

2.0b1 (Dec 29, 2005)
!!!THIS RELEASE REQUIRES RECOMPILATION OF ALL COMPILED CHEETAH TEMPLATES!!!

Core Changes:
- enabled use of any expression in ${placeholders}. See the examples I posted to
the email list on Dec 12th. All use cases of the #echo directive can now
be handled with ${placeholders}. This came from a suggestion by Mike
Orr. [TR]

- made it possible for templates to #extend (aka subclass) any arbitrary
baseclass, including Python's new style classes. You must either compile
your classes on the command line or use the new classmethod
Template.compile() as described below. The old Template(src) interface
still works, provided you don't try to use this new arbitrary baseclass
stuff. See my messages to the email list for more details. [TR]

- made it possible to create template classes dynamically, rather than just
instances. See the new classmethod Template.compile(). See my messages
to the email list for more details. [TR]

klass = Template.compile(src)

- made it easier to work with custom compiler settings, particularly from
the command line tool. You can now define a subclass of Template which
will compile your templates using custom compilerSettings, or even a
custom compiler class, without requiring you to manually pass in your
compilerSettings each time or define them in the template src itself via
the #compiler directive. You can make the command line tool use your
subclass by defining the environment variable CHEETAH_TEMPLATE_CLASS. It
should be in the form 'package.module:class'. See my messages
to the email list for more details. [TR]

- made it possible to pass the searchList in as an argument to #def'ined
methods. This makes all lookup that occur within the scope of that method
use the provided searchList rather than self._searchList. This does not
carry over to other methods called within the top method, unless they
explicitly accept the searchList in their signature AND you pass it to
them when calling them. This behaviour can be turned off with the
corresponding compilerSetting 'allowSearchListAsMethArg' [TR]

- added hooks for filtering / restricting dangerous stuff in cheetah source
code at compile time. These hooks can be used to enable Cheetah template
authoring by untrusted users. See my messages to the email list for more
details. Note, it filters expressions at parse/compile time, unlike Python's
old rexec module which restricted the Python environment at runtime. [TR]

# Here are the relevant compiler settings:
# use lower case keys here!!
'disabledDirectives':[], # list of directive keys, without the start token
'enabledDirectives':[], # list of directive keys, without the start token

'disabledDirectiveHooks':[], # callable(parser, directiveKey),
# called when a disabled directive is found, prior to raising an
exception

'preparseDirectiveHooks':[], # callable(parser, directiveKey)
'postparseDirectiveHooks':[], # callable(parser, directiveKey)

'preparsePlaceholderHooks':[], # callable(parser)
'postparsePlaceholderHooks':[], # callable(parser)

'expressionFilterHooks':[],
# callable(parser, expr, exprType, rawExpr=None, startPos=None)
# exprType is the name of the directive, 'psp', or 'placeholder'.
#all lowercase

- added support for a short EOLSlurpToken to supplement the #slurp
directive. It's currently re.compile('#\s*\n') (i.e # followed by
arbitrary whitespace and a new line), but this is not set in stone. One
other suggestion was the backslash char, but I believe Python's own
interpretation of backslashes will lead to confusion. The compiler
setting 'EOLSlurpToken' controls this. You can turn it off completely by
setting 'EOLSlurpToken' to None. See the email list for more details. [TR]

- added '_CHEETAH_' prefix to all instance attribute names in compiled
templates. This is related to the arbitrary baseclass change. [TR]

- shifted instance attribute setup to _initCheetahAttributes() method. This
is related to the arbitrary baseclass change. [TR]

- made it possible to use full expressions in the #extends directive, rather
than just dotted names. This allows you to do things like this:

#from xx.TemplateRepository import getTemplateClass
#extends getTemplateClass('someName')

I don't expect this to be used much. I needed it for a wiki system in
which the baseclasses for the templates are dynamically compiled at run
time and are not available via simple imports. [TR]

- added compiler setting autoImportForExtendDirective=True, so this existing
default behaviour can be turned off when needed. [TR]

- fixed a bug in the parsing of single-line #def's and #block's when they
are enclosed within #if ... #end if. Reported by Marcin Gajda [TR]

- tweak to remove needless write('') calls in generated code [TR]

The command line tool (CheetahWrapper.py):
- added code to cleanup trailing slashes on path arguments (code originally
from Mike Orr) [TR]
- turned on the ImportHooks by default for the 'cheetah fill' command. See the
discussion on the email list [TR]

ImportHooks:
- fixed a name error bug in the ImportHooks [TR]