= Markup Templates = The `markup` package provides a template engine that was heavily inspired by Kid (http://kid.lesscode.org) to a certain extent. Common concepts and features: * The same basic set of directives (excluding `py:extends` and `py:layout`) * Inline expressions using the `${expr}` (or just `$expr`) syntax. * Expressions are real Python code, not some “crippled” mini language. * Stream-based processing model making heavy use of Python generators. * Different serialization methods, for example to produce XML or HTML output. Differences include: * No generation of Python code for a template; the template is executed directly. * No support for `` processing instructions * Expressions are evaluated in a more flexible manner, meaning you can use e.g. dotted notation to access items in a dictionary, and the other way around (see the [source:/trunk/markup/eval.py markup.eval] module) * Proper scoping of variables. Names do not leak into outer scopes. * Use of XInclude and match templates instead of Kid's `py:extends` / `py:layout` directives * `py:match` directives use (basic) XPath expressions to match against input nodes, making match templates more powerful while keeping the syntax simple (see the [source:/trunk/markup/path.py markup.path] module) * Real (thread-safe) search path support * No dependency on ElementTree (due to the lack of position reporting) * The original location of parse events is kept throughout the processing pipeline, so that errors can be tracked back to a specific line/column in the template file * In simple tests (see [source:/trunk/examples/basic examples/basic]) the rendering phase (i.e. not including parsing/compilation) runs about twice as fast compared to Kid. Parsing is generally quite a bit faster in Markup, too.