internal/tools/: chameleon-4.5.4 metadata and description

Homepage Simple index

Fast HTML/XML Template Compiler.

author Malthe Borch
  • Development Status :: 5 - Production/Stable
  • Intended Audience :: Developers
  • Programming Language :: Python
  • Programming Language :: Python :: 3
  • Programming Language :: Python :: 3.9
  • Programming Language :: Python :: 3.10
  • Programming Language :: Python :: 3.11
  • Programming Language :: Python :: 3.12
  • Programming Language :: Python :: 3.13
  • Programming Language :: Python :: Implementation :: CPython
  • Programming Language :: Python :: Implementation :: PyPy
license BSD-like (
  • Documentation,
  • Issue Tracker,
  • Sources,
provides_extras docs
  • importlib-metadata ; python_version < "3.10"
  • Sphinx ; extra == 'docs'
  • sphinx-rtd-theme ; extra == 'docs'
requires_python >=3.9
File Tox results History
200 KB
Python Wheel
176 KB


PyPI - Python Version GitHub Actions Workflow Status

Chameleon is an HTML/XML template engine for Python. It uses the page templates language.

You can use it in any Python application with just about any version of Python (3.9+ and up, plus pypy 3).

Visit the documentation for more information.



4.5.4 (2024-04-08)

  • Fix an issue where $-sign interpolation escaping would not work correctly when more than two such symbols appeared next to each other. (#422)

4.5.3 (2024-04-05)

  • Minor optimization when rendering translations with a static message id (don’t need to test if it’s non-empty).

  • Fix a bug where a macro could not be used correctly to render a translation name. (#419)

4.5.2 (2024-01-29)

  • Fix a regression where a static symbol would not get correctly imported. (#414)

4.5.1 (2024-01-28)

  • Add Python 3.13 classifier.

  • Fix a regression where default_extension was no longer permitted as a positional argument to PageTemplateLoader. (#411)

4.5.0 (2024-01-18)

  • Chameleon now has type annotations!

  • The list of names previously disallowed for use as variables in templates such as “int” and “float” has been trimmed significantly, not because it’s a good idea to use such names but because the list of disallowed names was not exhaustive and complicated the compiler code; and perhaps more importantly, the technical reason for disallowing the names in the first place no longer applies.

  • Fix a regression where generated template code would suboptimal due to incorrect handling of internal variables.

  • Always cook templates in debug mode, even when using CHAMELEON_CACHE option to persist generated code on disk.

  • Parsing the AST back to Python code now uses the built-in ast.unparse function. This change is not directly surfaced but means that the unparsing code is now more correctly tracking changes to the interpreter.

  • Drop support for platforms where AST nodes aren’t weakref-capable (e.g., older PyPy).

4.4.4 (2024-01-18)

  • Fix a PyPy compatibility issue having to do with determining the set of versions when some packages (e.g. cffi) would not carry a version.

4.4.3 (2023-12-30)

  • Remove zope.interface.

  • Fix an issue where auto_reload is enabled and a file is loaded from a package that wasn’t zip-compressed. (#402)

4.4.2 (2023-12-18)

  • Fix brown bag release.

4.4.1 (2023-12-18)

  • Fix brown bag release.

4.4.0 (2023-12-12)

  • Add optional parameter package_name which allows loading a template relative to a package.

  • Drop support for Python 3.7.

  • Fix regression where Chameleon would not load templates correctly on Windows.

  • Fix names of dependencies for importlib_resources and importlib_metadata. (#394)

4.3.0 (2023-12-04)

  • Local variable scope is now iterable, completing dict interface. (#390)

  • Minor optimizations to rendering logic.

  • Implicit translation now provides the translation context, domain, and target language to the translation function (if applicable). Previously, the target language was provided, but this did not respect a change via i18n:target. (#369)

  • Replace pkg_resources with newer and faster importlib.resources and importlib.metadata. Just importing pkg_resources becomes slower and slower the more packages are installed.

4.2.0 (2023-09-25)

  • An XML document provided as a string (i.e. decoded) now correctly has its content encoding parsed.

  • Boolean attributes are now automatically configured for templates in non-XML mode, presuming that we’re being used to generate HTML.

    This means that the same loading mechanism can be used for both XML- and HTML-based templates.

4.1.0 (2023-08-29)

  • Boolean attributes (those configured using the optional boolean_attributes parameter) now work with $-expression interpolation.

    Unlike content and regular attributes, a special check for a falsy value is now done for boolean attributes, such that boolean logic can be used for interpolation expressions (only a truthy value will include the attribute).

    This reverts a change in behavior introduced in 3.8.0.

    Additionally, dynamic attributes now respect the boolean attributes configuration as well.

4.0.1 (2023-06-19)

  • Fix format spec applying for f-strings. (#376)

4.0.0 (2023-03-06)

  • Drop support for Python 2.7, 3.5, 3.6.

  • Add support for set- and dict comprehensions (#367)

  • Remove the following functions resp. modules:

    • .utils.text_()

    • .utils.unescape()

    • .compat

3.10.2 (2022-12-16)

  • Fix handling of eager compilation with subclassing (e.g. Pyramid integration).

3.10.1 (2022-05-17)

  • Fix __str__ method of chameleon.exc.TemplateError (#356)

3.10.0 (2022-04-06)

  • Make scope class dict-like (#305)

  • Work in FIPS enabled environments (#345)

  • Fix tab handling issue (#350)

3.9.1 (2021-05-14)

  • Avoid various traceback reference cycles.

3.9.0 (2021-02-26)

  • Removed compatibility code for unsupported Python versions

  • Dropped support for obsolete Python 3.4

  • Fixed namespace handling when redefining a variable in the same tal:define (#237)

  • Fixed failure computing a template’s digest if template.filename is None (#254)

3.8.1 (2020-07-06)

  • Added code optimization to reduce sequential appends of static text.

  • The default symbol in dynamic attributes is now symbolic. Previously, it was assigned the string value of the default attribute text. A similar change has been made for switch/case expressions.

  • The built-in attrs dictionary of static element attributes now correctly works with tal:define, etc.

  • Fix slice code generation compatibility issue on Python 3.9.

3.8.0 (2020-06-25)

  • Expose default marker as importable symbol chameleon.tales.DEFAULT_MARKER.

  • Removed legacy flag literal_false. To get a similar behavior, use boolean_attributes.

3.7.4 (2020-06-17)

  • Fix brown-bag release.

3.7.3 (2020-06-17)

  • Fix regression introduced in 3.6.2 where the default marker would incorrectly change its value between templates, causing issues in software which depends on the value being treated as a global object.

3.7.2 (2020-05-31)

  • Allow setting a custom value representation function, allowing custom formatting of variables during exception formatting.

3.7.1 (2020-05-10)

  • Fix compatibility issue with Python 3.9.

3.7.0 (2020-03-26)

  • Fixed garbage collection issue with variable scope objects (issue #301).

  • Fixed issue where setting a global variable would not be available locally.

  • A RepeatDict no longer inherits from dict since it does not actually provide that interface in a meaningful way.

  • Added feature gate enable_comment_interpolation which controls whether expression interpolation is enabled inside HTML comments (default is enabled).

  • Added support for Python 3.6+ f-strings (issue #294).

3.6.2 (2019-06-22)

3.6.1 (2019-04-01)

  • Fix limited search expression for illegal double hyphens in HTML comments to fix issue #289.

3.6 (2019-02-19)

  • Exclude RuntimeError (or RecursionError when available) from exception wrapping.

  • Fix double dollar ‘$$’ escaping such that a double dollar is always resolved, either as an interpolation expression, or as an escape where it is substituted by a single dollar symbol. This is now consistent with Zope’s handling of this character.

    Backslash-escaping of dollar-based string interpolation is no longer supported. The documentation has been updated to reflect this change.

    This fixes issue #283. Note that this reverses some of the changes introduced to fix issue #265.

  • Drop support for Python 3.3.

3.5 (2018-10-17)

  • Add support for Python 3.8.

  • Add support for TAL attributes in an XML declaration tag. This fixes issue #269.

  • Add support for custom exception handling for the tal:on-error statement. There is now an option on_error_handler available as a template configuration (issue #266).

  • Fix issue where double ‘$$’ escaping would affect non-interpolation expressions such as the bare ‘$$’ (issue #265).

  • Fix an issue where backslash dollar escaping would leave the backslash character still in place.

3.4 (2018-07-14)


  • Fix regression with translations in case of multiple nodes.

3.3 (2018-05-23)


  • Reset error token when rendering internal macro calls.

  • Fix edge case in exception handler causing recursion. [MatthewWilkes]

3.2 (2017-10-06)


  • Add the automatic variable macroname that’s bound to the name of the executing macro. Fixes

  • A tokenizer can now be configured on the template class. This is useful in the case where the template file input is modified before parsing (for example, where some tags are stripped away) such that token positions need to be offset accordingly for error locations to be rendered correctly.

  • Expression errors now display source marker (previously only filename, line and column was shown).

  • No longer require Python source files to import modules. [mrh1997]


  • Exception tracking now defers metadata allocation to time of error.

3.1 (2017-02-21)


  • Add option restricted_namespace which controls whether to restrict namespaces to those defined and used by the page template language. [hansroh]


  • Fixed attribute HTML entity escaping issue where an entity such as &amp; would be encoded twice.


  • Simplify exception tracking, reducing bytecode size significantly.

  • Avoid checking if a static string is None during expression interpolation.

3.0 (2016-12-07)


  • Fix issue on Python 2 where an exception was not cleared when using the pipe operator and was thus accessible through sys.exc_info().

  • The “exists” expression no longer leaks error information.

  • Escape ‘$$’ into ‘$’ in both content and string expressions.

  • Fix use of macro definition inside translation block.


  • Allow unquoted attribute values.

  • Wrap attribute error thrown when trying to use a non-macro as a macro as a RenderError to get proper error output.

  • Throw a parse error if ‘–’ (double hyphen) appears in an XML comment.

  • The i18n:target attribute now overrides a default target_language variable and is passed to the translation function.

  • Include filename in the on-disk cache module name. Previously, only the SHA digest in hex representation would be used, making it difficult to see where the module came from. This fixes issue #132.

  • Add support for non-ascii attribute names. [sank]


  • Drop support for Python 2.6, 3.1, and 3.2.

2.25 (2016-09-24)

  • Add explicit support / testing for Python 3.5.

  • Add \r to negative regex matches to the chameleon parser, where \n is used but \r was missing. Fixes a case, where the tag name was parsed into html\r instead of html. Fixes:

2.24 (2015-10-28)

  • Fixed Python 3.5 compatibility.

  • Fixed brown bag release.

2.23 (2015-10-26)

  • Added enable_data_attributes option that allows using HTML5 data attributes as control attributes instead or in addition to XML namespace attributes.

2.22 (2015-02-06)

  • Fix brown bag release.

2.21 (2015-02-06)

  • Added RenderError exception which indicates that an error occurred during the evaluation of an expression.

  • Clean up TemplateError exception implementation.

2.20 (2015-01-12)

  • Pass search_path to template class when loaded using TemplateLoader (or one of the derived classes). [faassen]

2.19 (2015-01-06)

  • Fix logging deprecation.

  • Fix environment-based configuration logging error.

2.18 (2014-11-03)

  • Fix minor compilation error.

2.17 (2014-11-03)

  • Add support for i18n:context. [wiggy]

  • Add missing ‘parity’ repeat property. [voxspox]

  • Don’t modify environment when getting variables from it. [fschulze]

2.16 (2014-05-06)

  • If a repeat expression evaluates to None then it is now equivalent to an empty set.

    This changes a behavior introduced in 2.14.

    This fixes issue #172.

  • Remove fossil test dependency on deprecated distribute.

  • Add explicit support / testing for Python 3.3 / 3.4.

  • Drop explicit support for Python 2.5 (out of maintenance, and no longer supported by tox or Travis-CI).

2.15 (2014-03-11)

  • Add Support for Python 3.4’s NameConstant. [brakhane]

2.14 (2013-11-28)

  • Element repetition using the TAL namespace no longer includes whitespace. This fixes issue #110.

  • Use absolute import for chameleon.interfaces module. This fixes issue #161.

2.13-1 (2013-10-24)

  • Fixing brown bag release.

2.13 (2013-10-21)


  • The template cache mechanism now includes additional configuration settings as part of the cache key such as strict and trim_attribute_space. [ossmkitty]

  • Fix cache issue where sometimes cached templates would not load correctly. [ossmkitty]

  • In debug-mode, correctly remove temporary files when the module loader is garbage-collected (on __del__). [graffic]

  • Fix error message when duplicate i18n:name directives are used in a translation.

  • Using the three-argument form of getattr on a chameleon.tal.RepeatDict no longer raises KeyError, letting the default provided to getattr be used. This fixes attempting to adapt a RepeatDict to a Zope interface under PyPy.

2.12 (2013-03-26)


  • When a tal:case condition succeeds, no other case now will.


  • Implicit translation now correctly extracts and normalizes complete sentences, instead of words. [witsch]

  • The default symbol in a tal:case condition now allows the element only if no other case succeeds.

2.11 (2012-11-15)


  • An issue was resolved where a METAL statement was combined with a tal:on-error handler.

  • Fix minor parser issue with incorrectly formatted processing instructions.

  • Provide proper error handling for Python inline code blocks.


  • The simple translation function now supports the translationstring interface.


  • Minor optimization which correctly detects when an element has no attributes.

2.10 (2012-10-12)


  • The fast_translate function has been deprecated. Instead, the default translation function is now always a function that simply interpolates the mapping onto the message default or id.

    The motivation is that since version 2.9, the context argument is non-trivial: the econtext mapping is passed. This breaks an expectation on the Zope platform that the context parameter is the HTTP request. Previously, with Chameleon this parameter was simply not provided and so that did not cause issues as such.

  • The ast24 module has been renamed to ast25. This should help clear up any confusion that Chameleon 2.x might be support a Python interpreter less than version 2.5 (it does not).


  • The ProxyExpr expression class (and hence the load: expression type) is now a TALES-expression. In practical terms, this means that the expression type (which computes a string result using the standard "${...}" interpolation syntax and proxies the result through a function) now supports fallback using the pipe operator ("|"). This fixes issue #128.

  • An attempt to interpolate using the empty string as the expression (i.e. ${}) now does nothing: the string ${} is simply output as is.

  • Added support for adding, modifying, and removing attributes using a dictionary expression in tal:attributes (analogous to Genshi’s py:attrs directive):

    <div tal:attributes="name value; attrs" />

    In the example above, name is an identifier, while value and attrs are Python expressions. However, attrs must evaluate to a Python dictionary object (more concisely, the value must implement the dictionary API-methods update() and items()).


  • In order to cut down on the size of the compiled function objects, some conversion and quoting statements have been put into functions. In one measurement, the reduction was 35%. The benchmark suite does not report of an increased render time (actually slightly decreased).


  • An exception is now raised if a trivial string is passed for metal:fill-slot. This fixes issue #89.

  • An empty string is now never translated. Not really a bug, but it’s been reported in as an issue (#92) because some translation frameworks handle this case incorrectly.

  • The template module loader (file cache) now correctly encodes generated template source code as UTF-8. This fixes issue #125.

  • Fixed issue where a closure might be reused unsafely in nested template rendering.

  • Fixed markup class __repr__ method. This fixes issue #124.

  • Added missing return statement to fix printing the non-abbreviated filename in case of an exception. [tomo]

2.9.2 (2012-06-06)


  • Fixed a PyPy incompatibility.

  • Fixed issue #109 which caused testing failures on some platforms.

2.9.1 (2012-06-01)


  • Fixed issue #103. The tal:on-error statement now always adds an explicit end-tag to the element, even with a substitution content of nothing.

  • Fixed issue #113. The tal:on-error statement now works correctly also for dynamic attributes. That is, the fallback tag now includes only static attributes.

  • Fixed name error which prevented the benchmark from running correctly.


  • Fixed deprecation warning on Python 3 for zope interface implements declaration. This fixes issue #116.

2.9.0 (2012-05-31)


  • The translation function now gets the econtext argument as the value for context. Note that historically, this was usually an HTTP request which might provide language negotiation data through a dictionary interface. [alvinyue]


  • Fixed import alias issue which would lead to a syntax error in generated Python code. Fixes issue #114.

2.8.5 (2012-05-02)


  • Fixed minor installation issues on Python 2.5 and 3. [ppaez]

  • Ensure output is unicode even when trivial (an empty string).

2.8.4 (2012-04-18)


  • In exception output, long filenames are now truncated to 60 characters of output, preventing line wrap which makes it difficult to scan the exception output.


  • Include filename and location in exception output for exceptions raised during compilation.

  • If a trivial translation substitution variable is given (i.e. an empty string), simply ignore it. This fixes issue #106.

2.8.3 (2012-04-16)


  • Log template source on debug-level before cooking.

  • The target_language argument, if given, is now available as a variable in templates.

2.8.2 (2012-03-30)


  • Temporary caches used in debug mode are cleaned up eagerly, rather than waiting for process termination. [mitchellrj]


  • The index, start and end methods on the TAL repeat object are now callable. This fixes an incompatibility with ZPT.

  • The loader now correctly handles absolute paths on Windows. [rdale]

2.8.1 (2012-03-29)


  • The exception formatter now lists errors in ‘wrapping order’. This means that the innermost, and presumably most relevant exception is shown last.


  • The exception formatter now correctly recognizes nested errors and does not rewrap the dynamically generated exception class.

  • The exception formatter now correctly sets the __module__ attribute to that of the original exception class.

2.8.0 (2012-02-29)


  • Added support for code blocks using the <?python … ?> processing instruction syntax.

    The scope is name assignments is up until the nearest macro definition, or the template itself if macros are not used.


  • Fall back to the exception class’ __new__ method to safely create an exception object that is not implemented in Python.

  • The exception formatter now keeps track of already formatted exceptions, and ignores them from further output.

2.7.4 (2012-02-27)

  • The error handler now invokes the __init__ method of BaseException instead of the possibly overridden method (which may take required arguments). This fixes issue #97. [j23d, malthe]

2.7.3 (2012-01-16)


  • The trim whitespace option now correctly trims actual whitespace to a single character, appearing either to the left or to the right of an element prefix or suffix string.

2.7.2 (2012-01-08)


  • Added option trim_attribute_space that decides whether attribute whitespace is stripped (at most down to a single space). This option exists to provide compatibility with the reference implementation. Fixes issue #85.


  • Ignore unhashable builtins when generating a reverse builtin map to quickly look up a builtin value. [malthe]

  • Apply translation mapping even when a translation function is not available. This fixes issue #83. [malthe]

  • Fixed issue #80. The translation domain for a slot is defined by the source document, i.e. the template providing the content for a slot whether it be the default or provided through metal:fill-slot. [jcbrand]

  • In certain circumstances, a Unicode non-breaking space character would cause a define clause to fail to parse.

2.7.1 (2011-12-29)


  • Enable expression interpolation in CDATA.

  • The page template class now implements dictionary access to macros:


    This is a short-hand for:



  • An invalid define clause would be silently ignored; we now raise a language error exception. This fixes issue #79.

  • Fixed regression where ${...} interpolation expressions could not span multiple lines. This fixes issue #77.

2.7.0 (2011-12-13)


  • The load: expression now derives from the string expression such that the ${...} operator can be used for expression interpolation.

  • The load: expression now accepts asset specs; these are resolved by the pkg_resources.resource_filename function:


    An example from the test suite:



  • If an attribute name for translation was not a valid Python identifier, the compiler would generate invalid code. This has been fixed, and the compiler now also throws an exception if an attribute specification contains a comma. (Note that the only valid separator character is the semicolon, when specifying attributes for translation via the i18n:translate statement). This addresses issue #76.

2.6.2 (2011-12-08)


  • Fixed issue where tal:on-error would not respect tal:omit-tag or namespace elements which are omitted by default (such as <tal:block />).

  • Fixed issue where macros attribute would not be available on file-based templates due to incorrect initialization.

  • The TryExcept and TryFinally AST nodes are not available on Python 3.3. These have been aliased to Try. This fixes issue #75.


  • The TAL repeat item now makes a security declaration that grants access to unprotected subobjects on the Zope 2 platform:

    __allow_access_to_unprotected_subobjects__ = True

    This is required for legacy compatibility and does not affect other environments.

  • The template object now has a method write(body) which explicitly decodes and cooks a string input.

  • Added configuration option loader_class which sets the class used to create the template loader object.

    The class (essentially a callable) is created at template construction time.

2.6.1 (2011-11-30)


  • Decode HTML entities in expression interpolation strings. This fixes issue #74.

  • Allow xml and xmlns attributes on TAL, I18N and METAL namespace elements. This fixes issue #73.

2.6.0 (2011-11-24)


  • Added support for implicit translation:

    The implicit_i18n_translate option enables implicit translation of text. The implicit_i18n_attributes enables implicit translation of attributes. The latter must be a set and for an attribute to be implicitly translated, its lowercase string value must be included in the set.

  • Added option strict (enabled by default) which decides whether expressions are required to be valid at compile time. That is, if not set, an exception is only raised for an invalid expression at evaluation time.

  • An expression error now results in an exception only if the expression is attempted evaluated during a rendering.

  • Added a configuration option prepend_relative_search_path which decides whether the path relative to a file-based template is prepended to the load search path. The default is True.

  • Added a configuration option search_path to the file-based template class, which adds additional paths to the template load instance bound to the load: expression. The option takes a string path or an iterable yielding string paths. The default value is the empty set.


  • Exception instances now support pickle/unpickle.

  • An attributes in i18n:attributes no longer needs to match an existing or dynamic attribute in order to appear in the element. This fixes issue #66.

2.5.3 (2011-10-23)


  • Fixed an issue where a nested macro slot definition would fail even though there existed a parent macro definition. This fixes issue #69.

2.5.2 (2011-10-12)


  • Fixed an issue where technically invalid input would result in a compiler error.


  • The markup class now inherits from the unicode string type such that it’s compatible with the string interface.

2.5.1 (2011-09-29)


  • The symbol names “convert”, “decode” and “translate” are now no longer set as read-only compiler internals. This fixes issue #65.

  • Fixed an issue where a macro extension chain nested two levels (a template uses a macro that extends a macro) would lose the middle slot definitions if slots were defined nested.

    The compiler now throws an error if a nested slot definition is used outside a macro extension context.

2.5.0 (2011-09-23)


  • An expression type structure: is now available which wraps the expression result as structure such that it is not escaped on insertion, e.g.:

    <div id="content">
       ${structure: context.body}

    This also means that the structure keyword for tal:content and tal:replace now has an alternative spelling via the expression type structure:.

  • The string-based template constructor now accepts encoded input.

2.4.6 (2011-09-23)


  • The tal:on-error statement should catch all exceptions.

  • Fixed issue that would prevent escaping of interpolation expression values appearing in text.

2.4.5 (2011-09-21)


  • The tal:on-error handler should have a error variable defined that has the value of the exception thrown.

  • The tal:on-error statement is a substitution statement and should support the “text” and “structure” insertion methods.

2.4.4 (2011-09-15)


  • An encoding specified in the XML document preamble is now read and used to decode the template input to unicode. This fixes issue #55.

  • Encoded expression input on Python 3 is now correctly decoded. Previously, the string representation output would be included instead of an actually decoded string.

  • Expression result conversion steps are now correctly included in error handling such that the exception output points to the expression location.

2.4.3 (2011-09-13)


  • When an encoding is provided, pass the ‘ignore’ flag to avoid decoding issues with bad input.


  • Fixed pypy compatibility issue (introduced in previous release).

2.4.2 (2011-09-13)


  • Fixed an issue in the compiler where an internal variable (such as a translation default value) would be cached, resulting in variable scope corruption (see issue #49).

2.4.1 (2011-09-08)


  • Fixed an issue where a default value for an attribute would sometimes spill over into another attribute.

  • Fixed issue where the use of the default name in an attribute interpolation expression would print the attribute value. This is unexpected, because it’s an expression, not a static text suitable for output. An attribute value of default now correctly drops the attribute.

2.4.0 (2011-08-22)


  • Added an option boolean_attributes to evaluate and render a provided set of attributes using a boolean logic: if the attribute is a true value, the value will be the attribute name, otherwise the attribute is dropped.

    In the reference implementation, the following attributes are configured as boolean values when the template is rendered in HTML-mode:

    "compact", "nowrap", "ismap", "declare", "noshade",
    "checked", "disabled", "readonly", "multiple", "selected",
    "noresize", "defer"

    Note that in Chameleon, these attributes must be manually provided.


  • The carriage return character (used on Windows platforms) would incorrectly be included in Python comments.

    It is now replaced with a line break.

    This fixes issue #44.

2.3.8 (2011-08-19)

  • Fixed import error that affected Python 2.5 only.

2.3.7 (2011-08-19)


  • Added an option literal_false that disables the default behavior of dropping an attribute for a value of False (in addition to None). This modified behavior is the behavior exhibited in reference implementation.


  • Undo attribute special HTML attribute behavior (see previous release).

    This turned out not to be a compatible behavior; rather, boolean values should simply be coerced to a string.

    Meanwhile, the reference implementation does support an HTML mode in which the special attribute behavior is exhibited.

    We do not currently support this mode.

2.3.6 (2011-08-18)


  • Certain HTML attribute names now have a special behavior for a attribute value of True (or default if no default is defined). For these attributes, this return value will result in the name being printed as the value:

    <input type="input" tal:attributes="checked True" />

    will be rendered as:

    <input type="input" checked="checked" />

    This behavior is compatible with the reference implementation.

2.3.5 (2011-08-18)


  • Added support for the set operator ({item, item, ...}).


  • If macro is defined on the same element as a translation name, this no longer results in a “translation name not allowed outside translation” error. This fixes issue #43.

  • Attribute fallback to dictionary lookup now works on multiple items (e.g. d1.d2.d2). This fixes issue #42.

2.3.4 (2011-08-16)


  • When inserting content in either attributes or text, a value of True (like False and None) will result in no action.

  • Use statically assigned variables for "attrs" and "default". This change yields a performance improvement of 15-20%.

  • The template loader class now accepts an optional argument default_extension which accepts a filename extension which will be appended to the filename if there’s not already an extension.


  • The default symbol is now True for an attribute if the attribute default is not provided. Note that the result is that the attribute is dropped. This fixes issue #41.

  • Fixed an issue where assignment to a variable "type" would fail. This fixes issue #40.

  • Fixed an issue where an (unsuccessful) assignment for a repeat loop to a compiler internal name would not result in an error.

  • If the translation function returns the identical object, manually coerce it to string. This fixes a compatibility issue with translation functions which do not convert non-string objects to a string value, but simply return them unchanged.

2.3.3 (2011-08-15)


  • The load: expression now passes the initial keyword arguments to its template loader (e.g. auto_reload and encoding).

  • In the exception output, string variable values are now limited to a limited output of characters, single line only.


  • Fixed horizontal alignment of exception location info (i.e. ‘String:’, ‘Filename:’ and ‘Location:’) such that they match the template exception formatter.

2.3.2 (2011-08-11)


  • Fixed issue where i18n:domain would not be inherited through macros and slots. This fixes issue #37.

2.3.1 (2011-08-11)


  • The Builtin node type may now be used to represent any Python local or global name. This allows expression compilers to refer to e.g. get or getitem, or to explicit require a builtin object such as one from the extra_builtins dictionary.


  • Builtins which are not explicitly disallowed may now be redefined and used as variables (e.g. nothing).

  • Fixed compiler issue with circular node annotation loop.

2.3 (2011-08-10)


  • Added support for the following syntax to disable inline evaluation in a comment:

    <!–? comment appears verbatim (no ${…} evaluation) –>

    Note that the initial question mark character (?) will be omitted from output.

  • The parser now accepts ‘<’ and ‘>’ in attributes. Note that this is invalid markup. Previously, the ‘<’ would not be accepted as a valid attribute value, but this would result in an ‘unexpected end tag’ error elsewhere. This fixes issue #38.

  • The expression compiler now provides methods assign_text and assign_value such that a template engine might configure this value conversion to support e.g. encoded strings.

    Note that currently, the only client for the assign_text method is the string expression type.

  • Enable template loader for string-based template classes. Note that the filename keyword argument may be provided on initialization to identify the template source by filename. This fixes issue #36.

  • Added extra_builtins option to the page template class. These builtins are added to the default builtins dictionary at cook time and may be provided at initialization using the extra_builtins keyword argument.


  • If a translation domain is set for a fill slot, use this setting instead of the macro template domain.

  • The Python expression compiler now correctly decodes HTML entities 'gt' and 'lt'. This fixes issue #32.

  • The string expression compiler now correctly handles encoded text (when support for encoded strings is enabled). This fixes issue #35.

  • Fixed an issue where setting the filename attribute on a file-based template would not automatically cause an invalidation.

  • Exceptions raised by Chameleon can now be copied via copy.copy. This fixes issue #36. [leorochael]

  • If copying the exception fails in the exception handler, simply re-raise the original exception and log a warning.

2.2 (2011-07-28)


  • Added new expression type load: that allows loading a template. Both relative and absolute paths are supported. If the path given is relative, then it will be resolved with respect to the directory of the template.

  • Added support for dynamic evaluation of expressions.

    Note that this is to support legacy applications. It is not currently wired into the provided template classes.

  • Template classes now have a builtins attribute which may be used to define built-in variables always available in the template variable scope.


  • The file-based template class no longer accepts a parameter loader. This parameter would be used to load a template from a relative path, using a find(filename) method. This was however, undocumented, and probably not very useful since we have the TemplateLoader mechanism already.

  • The compiled template module now contains an initialize function which takes values that map to the template builtins. The return value of this function is a dictionary that contains the render functions.


  • The file-based template class no longer verifies the existence of a template file (using os.lstat). This now happens implicitly if eager parsing is enabled, or otherwise when first needed (e.g. at render time).

    This is classified as a bug fix because the previous behavior was probably not what you’d expect, especially if an application initializes a lot of templates without needing to render them immediately.

2.1.1 (2011-07-28)


  • Improved exception display. The expression string is now shown in the context of the original source (if available) with a marker string indicating the location of the expression in the template source.


  • The structure insertion mode now correctly decodes entities for any expression type (including string:). This fixes issue #30.

  • Don’t show internal variables in the exception formatter variable listing.

2.1 (2011-07-25)


  • Expression interpolation (using the ${...} operator and previously also $identifier) now requires braces everywhere except inside the string: expression type.

    This change is motivated by a number of legacy templates in which the interpolation format without braces $identifier appears as text.

2.0.2 (2011-07-25)


  • Don’t use dynamic variable scope for lambda-scoped variables (#27).

  • Avoid duplication of exception class and message in traceback.

  • Fixed issue where a metal:fill-slot would be ignored if a macro was set to be used on the same element (#16).

2.0.1 (2011-07-23)


  • Fixed issue where global variable definition from macro slots would fail (they would instead be local). This also affects error reporting from inside slots because this would be recorded internally as a global.

  • Fixed issue with template cache digest (used for filenames); modules are now invalidated whenever any changes are made to the distribution set available (packages on sys.path).

  • Fixed exception handler to better let exceptions propagate through the renderer.

  • The disk-based module compiler now mangles template source filenames such that the output Python module is valid and at root level (dots and hyphens are replaced by an underscore). This fixes issue #17.

  • Fixed translations (i18n) on Python 2.5.

2.0 (2011-07-14)

  • Point release.

2.0-rc14 (2011-07-13)


  • The tab character (\t) is now parsed correctly when used inside tags.


  • The RepeatDict class now works as a proxy behind a separate dictionary instance.

  • Added template constructor option keep_body which is a flag (also available as a class attribute) that controls whether to save the template body input in the body attribute.

    This is disabled by default, unless debug-mode is enabled.

  • The page template loader class now accepts an optional formats argument which can be used to select an alternative template class.

2.0-rc13 (2011-07-07)


  • The backslash character (followed by optional whitespace and a line break) was not correctly interpreted as a continuation for Python expressions.


  • The Python expression implementation is now more flexible for external subclassing via a new parse method.

2.0-rc12 (2011-07-04)


  • Initial keyword arguments passed to a template now no longer “leak” into the template variable space after a macro call.

  • An unexpected end tag is now an unrecoverable error.


  • Improve exception output.

2.0-rc11 (2011-05-26)


  • Fixed issue where variable names that begin with an underscore were seemingly allowed, but their use resulted in a compiler error.


  • Template variable names are now allowed to be prefixed with a single underscore, but not two or more (reserved for internal use).

    Examples of valid names:

  • Added support for Genshi’s comment “drop” syntax:

    <!--! This comment will be dropped -->

    Note the additional exclamation (!) character.

    This fixes addresses issue #10.

2.0-rc10 (2011-05-24)


  • The tal:attributes statement now correctly operates case-insensitive. The attribute name given in the statement will replace an existing attribute with the same name, without respect to case.


  • Added meta:interpolation statement to control expression interpolation setting.

    Strings that disable the setting: "off" and "false". Strings that enable the setting: "on" and "true".

  • Expression interpolation now works inside XML comments.

2.0-rc9 (2011-05-05)


  • Better debugging support for string decode and conversion. If a naive join fails, each element in the output will now be attempted coerced to unicode to try and trigger the failure near to the bad string.

2.0-rc8 (2011-04-11)


  • If a macro defines two slots with the same name, a caller will now fill both with a single usage.

  • If a valid of None is provided as the translation function argument, we now fall back to the class default.

2.0-rc7 (2011-03-29)


  • Fixed issue with Python 2.5 compatibility AST. This affected at least PyPy 1.4.


  • The auto_reload setting now defaults to the class value; the base template class gives a default value of chameleon.config.AUTO_RELOAD. This change allows a subclass to provide a custom default value (such as an application-specific debug mode setting).

2.0-rc6 (2011-03-19)


  • Added support for target_language keyword argument to render method. If provided, the argument will be curried onto the translation function.


  • The HTML entities ‘lt’, ‘gt’ and ‘quot’ appearing inside content subtition expressions are now translated into their native character values. This fixes an issue where you could not dynamically create elements using the structure (which is possible in ZPT). The need to create such structure stems from the lack of an expression interpolation operator in ZPT.

  • Fixed duplicate file pointer issue with test suite (affected Windows platforms only). This fixes issue #9. [oliora]

  • Use already open file using os.fdopen when trying to write out the module source. This fixes LP #731803.

2.0-rc5 (2011-03-07)


  • Fixed a number of issues concerning the escaping of attribute values:

    1. Static attribute values are now included as they appear in the source.

      This means that invalid attribute values such as "true && false" are now left alone. It’s not the job of the template engine to correct such markup, at least not in the default mode of operation.

    2. The string expression compiler no longer unescapes values. Instead, this is left to each expression compiler. Currently only the Python expression compiler unescapes its input.

    3. The dynamic escape code sequence now correctly only replaces ampersands that are part of an HTML escape format.


  • The page template classes and the loader class can now be imported directly from the chameleon module.


  • If a custom template loader is not provided, relative paths are now resolved using os.abspath (i.e. to the current working directory).

  • Absolute paths are normalized using os.path.normpath and os.path.expanduser. This ensures that all paths are kept in their “canonical” form.

2.0-rc4 (2011-03-03)


  • Fixed an issue where the output of an end-to-end string expression would raise an exception if the expression evaluated to None (it should simply output nothing).

  • The convert function (which is configurable on the template class level) now defaults to the translate function (at run-time).

    This fixes an issue where message objects were not translated (and thus converted to a string) using the a provided translate function.

  • Fixed string interpolation issue where an expression immediately succeeded by a right curly bracket would not parse.

    This fixes issue #5.

  • Fixed error where tal:condition would be evaluated after tal:repeat.


  • Python expression is now a TALES expression. That means that the pipe operator can be used to chain two or more expressions in a try-except sequence.

    This behavior was ported from the 1.x series. Note that while it’s still possible to use the pipe character (“|”) in an expression, it must now be escaped.

  • The template cache can now be shared by multiple processes.

2.0-rc3 (2011-03-02)


  • Fixed atexit handler.

    This fixes issue #3.

  • If a cache directory is specified, it will now be used even when not in debug mode.

  • Allow “comment” attribute in the TAL namespace.

    This fixes an issue in the sense that the reference engine allows any attribute within the TAL namespace. However, only “comment” is in common use.

  • The template constructor now accepts a flag debug which puts the template instance into debug-mode regardless of the global setting.

    This fixes issue #1.


  • Added exception handler for exceptions raised while evaluating an expression.

    This handler raises (or attempts to) a new exception of the type RenderError, with an additional base class of the original exception class. The string value of the exception is a formatted error message which includes the expression that caused the exception.

    If we are unable to create the exception class, the original exception is re-raised.

2.0-rc2 (2011-02-28)

  • Fixed upload issue.

2.0-rc1 (2011-02-28)

  • Initial public release. See documentation for what’s new in this series.