Contents
The format parameter can now be set to 'straight' if you do not want any output formatting at all (the default format removes empty lines).
These bugs in the last version have been fixed:
- The Template _filters list is now an instance attribute (otherwise TurboKid may expand that list endlessly).
- In some situations Kid printed misleading error tracebacks.
- There was a small problem with indentation formatting.
- The representation of the Format class was not correct.
Error reporting has been taken a step further: Kid now tries to track down the location of errors to the original Kid template file instead of merely printing the error location in the compiled template module. This has been one of the most urgently requested features.
The dependency on the ElementTree package has been removed completely.
Kid 0.9.3 templates needed to be recompiled in order to run with newer versions. This recompilation is now done automatically in the background whenever there is a new version, in order to ensure compatibility and to benefit from improvements in template creation in newer versions.
The old Kid namespace URL http://naeblis.cx/ns/kid# is not supported any more. Please use the new URL http://purl.org/kid/ns# which has been introduced already in Kid 0.6.
These bugs in the last version have been fixed:
- Combination of both py:layout and py:extends caused a problem.
- Passing parameters to functions with default parameters which were defined with py:def did not work properly.
- The XML encoding declaration in the template was not evaluated.
XML errors from the Expat parser are now shown along with the corresponding line of the erroneous XML code and the exact position of the error.
Kid is now much less picky about py:extends and py:layout expressions and will output much better error messages if they cause an error.
Kid will now also check that you dont' use a Template parameter that conflicts with the name of a member of the BaseTemplate class.
The serialization methods now take an additional attribute format that must be an instance of the new Format class or a string referring to one of several predefined output formats. This class gives you more fine control over how the text content is serialized. You can do things like removing redundant whitespace, adding indentation, word wrapping, using named entities and using typographic characters for quotes etc. You can also plug-in your own text filters here. The chapter on Formatting in the User's Guide lists all available parameters.
The old filter methods of the Serializer class have been replaced by a more powerful general method incorporating the specified format.
As discussed on the mailing list, the HTML serializer now produces tags in lowercase. You can change this behavior with the transpose attribute or using the output method 'HTML' instead of 'html'.
Additional HTML output methods have been defined corresponding to Mozilla's "full standards mode", "almost standards mode" and "quirks mode."
The XHTML Serializer now also injects a meta tag with the content type at the top of the head section. This had been only done by the HTMLSerializer before. The content type is not injected if such a meta tag already exists. This automatism can be controlled with the inject_type parameter.
The rule when variable substitution is applied to XML comments has been slightly changed, as explained in the Language Specification.
The load_template() function now takes additional parameters entity_map and exec_module. This lets you choose a different XML entity map to be used when parsing the template, and gives you more control over the way in which the compiled template code is executed in the newly created module. The entity_map parameter can also be passed to some other functions such as XML() and compile_file().
The enable_import mechanism has been completely rewritten to support both Python 2.5 and Python eggs. The old method based on ihooks is not used any more since ihooks is not compatible with Python eggs; and the old method based on path_hooks is not used any more since in Python 2.5, importers installed via path_hooks do not fall back to the built-in import mechanism any more, breaking the previous implementation.
The new implementation allows enabling imports from specified paths as well as imports from sys.path which is now implemented via meta_path.
Python versions older than Python 2.3 are not supported by Kid any more.
Due to optimizations in kid.template_util, Kid 0.9.3 templates need to be recompiled in order to run with Kid 0.9.4.
The 'html' output method now produces lower-case tags. If you want upper-case tags, you have to use the 'HTML' method now.
The kid.parser and kid.pull modules have been renamed kid.codewriter and kid.parser respectively (in anticipation of a parsing subsystem).
A couple of bugs that had been reported as trac tickets have been fixed.
Using layout templates which are extended by another template did not work properly (as reported on the TurboGears mailing list); this has been fixed.
More unit tests have been added. The testing machinery now supports both py.test and nose, or you can simply use the run_tests.py script.
- Re-applied a patch from ticket [66] that fix a bug where comments caused errors when in base templates.
- Changed all of the lesscode.org links into kid-templating.org
- Added and updated a few tests
- Removed the NamespaceStack.set method and made NamespaceStack.pop return the deleted value.
- Set balanced blocks to be off by default.
- Updated the parser to better handle interpolation of non-string types in comments. Reported in #182
- Updated to current version of ez_setup.py.
- Improved importer.py, resolving tickets #103 (FutureWarnings) and #137 (using new import hooks).
- The testing code can now figure out what testing modules to run dynamically. In addition, the code also determines which functions are tests dynamically. Tests that need pylib are skipped for those that don't have it. If you run 'python test_kid.py' now you should be seeing more tests executed.
- Removed the revision history from the language and and instead include a pointer to the Release Notes.
- Allow the kid command to accept XML piped into it's stdin when '-' is used as the filename.
- Patch from #143. The load_template() function accepts an 'ns' keyword argument to pre-populate the template module namespace with global variables. Thanks!
- Created an API to replace the various ways that configuration options are currently set.
- Allow the XML function to take a new keyword parameter (xmlns) that sets the default namespace for a fragment.
- Added support for xml.etree, which is the ElementTree packaged with Python 2.5. Running 'make test' only checks xml.etree currently.
- The __future__ imports have been moved to the top of the module to play nicely in Python 2.5. I have also added 2.5 to the makefile so it will be tested before each release.
- makefile regression test includes Python 2.5.
The parameters passed to a template with a py:layout were not visible in named template functions or match templates.
A small bug existed in Python 2.3 where the dict.update() method was being called incorrectly. Python 2.4 allows a list of tuples to be passed to update(), whereas Python 2.3 does not.
The logic to apply the template matches has been reworked. This was due to the discovery of some odd behavior when using multiple template inheritence.
A template type or name can now be dynamically passed into a template to be used as the layout template.
There is a new feature in Kid for a template to specify a layout template to which match templates, named template definitions, and template parameters will be applied. This is useful for applying a generic set of headers, menus, footers, etc. to a many pages without duplicating large amounts of code in each page. More information and examples can be found in the py:layout section of the Language Specification.
Kid Template instances now provide convenience funcitons defined(name) and value_of(name).
XML comments starting with a ! character will not appear in the serialized output of a template.
On Windows, the kid and kidc console commands can now be used as conveniently as on Unix, since the Kid installer creates kid.exe and kidc.exe launchers in the Python Scripts directory.
There has been significant change in version 0.6. This includes enhancements and modifications to the template language and python interface.
The following changes are likely to impact existing code and templates. Where possible, we have tried to maintain backward compatibility but that wasn't possible in all cases.
The Upgrade Script can be used to bring 0.5 templates up to 0.6 syntax.
The Kid namespace has changed from http://naeblis.cx/ns/kid# to http://purl.org/kid/ns#. The naeblis.cx domain is privately owned and could expire some time in the future. purl.org is a system for establishing and maintaining "persistent" URIs.
A temporary hack has been put in place to substitute references to the old namespace URI with the new namespace URI. A warning is output when this occurs. This will be removed in a couple of months so it is recommended that templates be upgraded as soon as possible.
Due to initial confusion many experienced with the name py:omit, it has been renamed py:strip. The term "omit" was often read as "omit the element and all descendants". The new term "strip" seems to better indicate the semantic: "strip the start and end tag but process descendants."
The syntax of brace expansions has been modified match more closely with existing Python substitution syntax. In 0.5 python expressions enclosed in curly braces ({}) were evaluated and their results substituted. In 0.6, the rules have changed as follows:
- $$ is an escape; it is replaced with a single $.
- $name substitutes a variable value.
- ${expr} substitutes the result of evaluating any python expression.
See Python Expression Substitution in the Language Reference.
Kid now uses cElementTree if it is available. Preliminary tests show moderate performance increases. In most cases, we're seeing template parse and execution time increase by about 15%. The poor increase (relative to other cET/ET numbers) is due to the fact that we're not using cElementTree's native parser as it doesn't support comments or processing instructions. The plan is to lobby the effbot organization to add these features (hint, hint: send patches) so that we can get the huge increases people are seeing elsewhere.
Kid automatically determines whether cElementTree is available and uses it if so. If cElementTree is not available, Kid falls back on Python ElementTree. If you want to turn off use of cElementTree, you can set the environment variable KID_NOCET to 1.
In versions of Kid prior to 0.6, the first line of an embedded Python code block had to be a Python comment (#). This was due to Python's whitespace semantics. Christoph determined a process for establishing the correct indent levels without requiring a comment as the first line.
Starting in Kid 0.6, a comment is no longer required to be the first line in a <?python?> processing instruction. It is also possible to have single line code blocks:
<?python x = 10 ?>
The Python interfaces have been reworked significantly and now are very similar to Cheetah's. There are two preferred methods for accessing a template.
The first method existed in 0.5 but was not documented well. If you have enabled the kid import hooks, then you can import a template and create an instance of the template by accessing the Template class exposed by the module:
import kid ; kid.enable_import() import mytemplate template = mytemplate.Template(foo='bar', bling=1) print template.serialize()
The primary difference from 0.5 is that template variables are passed to the Template constructor instead of to the individual execution methods (serialize, generate, write, pull).
It is also possible to set template variables after the template instance is created by simply assigning to template object instance:
template = mytemplate.Template() template.foo = 'bar' template.bling = 1 print str(template)
Here we see another small addition: template instances implement __str__ and __unicode__ built-ins. These methods are equivalent to calling serialize(encoding='utf-8') and serialize(encoding='utf-16'), respectively.
The kid.Template function works much like Template class constructors but takes an additional parameter that allows the template to be loaded from a file, string, or module name. It is sometimes easier to manage templates as files on disk rather than as python modules.
Example:
from kid import Template # create a template from file template = Template(file='mytemplate.kid', foo='bar', bling=1) # create a template from string template = Template(source="<p>${foo}</p>", foo='bar') # create a template from a python module name template = Template(name='templates.mytemplate', foo='bar')
This last form is sometimes useful because it doesn't require the kid import hook to be enabled and it also allows template names to be specified at run-time.
See kid.Template function in the User's Guide for more info.
Match Templates are a cross between XSLT's match templates and JSP tag libraries. They allow a set of filters to be put in place that matches infoset items generated by a template so that output can be modified.
While match templates provide a general purpose mechanism for transforming XML content, it is especially useful in a couple of situations which have driven the design:
See Match Templates in the Language Reference for more information.
Templates now support multiple inheritance of template functions (py:def) and match templates (py:match). A template indicates that it extends one or more other templates by setting the py:extends attribute on the root element:
<?xml version='1.0' encoding='utf-8'?> <html py:extends="'common.kid', 'forms.kid'" ...
py:extends may contain template modules, Template classes, or strings specifying template paths relative to current template file.
See Template Reuse in the Language Reference for more information.
A new py:attrs attribute has been added that allows attributes to be specified using a dictionary.
See Dynamic Attributes in the Language Reference for more information.
Due to the amount of changes in template syntax, a migration script is provided that can upgrade kid 0.5 templates to 0.6 syntax. This includes changing the namespace, py:strip, and new expression substitution syntax.
The script can be found in the source distribution as misc/upgrade-0.6.py. The script can take multiple file names and upgrades each in-place while preserving a backup. For instance:
$ python upgrade-0.6.py path/to/template.kid Upgraded: template.kid...
On posix systems, you can upgrade a bunch of kid templates under the current working directory with the following command:
$ find . -name '*.kid' | xargs python upgrade-0.6.py Upgraded: template1.kid... Upgraded: template2.kid... Upgraded: template3.kid... Upgraded: template4.kid...