Tempita Templating Engine

Author:	Ian Bicking <[email protected]>
source:	https://bitbucket.org/ianb/tempita

Status & License

Tempita is available under a MIT-style license.

It is not actually actively developed, and not an ambitious project. It does not seek to take over the templating world, or adopt many new features. I just wanted a small templating language for cases when % and string.Template weren't enough.

Why Another Templating Language

Surely the world has enough templating languages? So why did I write another.

I initially used Cheetah as the templating language for Paste Script, but this caused quite a few problems. People frequently had problems installing Cheetah because it includes a C extension. Also, the errors and invocation can be a little confusing. This might be okay for something that used Cheetah's features extensively, except that the templating was a very minor feature of the system, and many people didn't even understand or care about where templating came into the system.

At the same time, I was starting to create reusable WSGI components that had some templating in them. Not a lot of templating, but enough that string.Template had become too complicated -- I need if statements and loops.

Given this, I started looking around for a very small templating language, and I didn't like anything I found. Many of them seemed awkward or like toys that were more about the novelty of the implementation than the utility of the language.

So one night when I felt like coding but didn't feel like working on anything I was already working on, I wrote this. It was first called paste.util.template, but I decided it deserved a life of its own, hence Tempita.

The Interface

The interface is intended to look a lot like string.Template. You can create a template object like:

>>> import tempita
>>> tmpl = tempita.Template("""Hello {{name}}""")
>>> tmpl.substitute(name='Bob')
'Hello Bob'

Or if you want to skip the class:

>>> tempita.sub("Hello {{name}}", name='Alice')
'Hello Alice'

Note that the language allows arbitrary Python to be executed, so your templates must be trusted.

You can give a name to your template, which is handy when there is an error (the name will be displayed):

>>> tmpl = tempita.Template('Hi {{name}}', name='tmpl')
>>> tmpl.substitute()
Traceback (most recent call last):
    ...
NameError: name 'name' is not defined at line 1 column 6 in file tmpl

You can also give a namespace to use by default, which .substitute(...) will augment:

>>> tmpl = tempita.Template(
...     'Hi {{upper(name)}}',
...     namespace=dict(upper=lambda s: s.upper()))
>>> tmpl.substitute(name='Joe')
'Hi JOE'

Lastly, you can give a dictionary-like object as the argument to .substitute, like:

>>> name = 'Jane'
>>> tmpl.substitute(locals())
'Hi JANE'

There's also an HTMLTemplate class that is more appropriate for templates that produce HTML.

You can also instantiate a template from a filename with Template.from_filename(filename, namespace={}, encoding=None). This is like calling:

Template(open(filename, 'rb').read().decode(encoding),
         name=filename, namespace=namespace)

Unicode

Tempita tries to handle unicode gracefully, for some value of "graceful". Template objects have a default_encoding attribute. It will try to use that encoding whenever unicode and str objects are mixed in the template. E.g.:

>>> tmpl = tempita.Template(u'Hi {{name}}')
>>> import sys
>>> if sys.version.startswith('2'): # unicode is the default in 3 -> failing test
...     val = tmpl.substitute(name='Jos\xc3\xa9')
...     comparison = val == u'Hi Jos\xe9'
... else:
...     comparison = True
>>> comparison
True
>>> tmpl = tempita.Template('Hi {{name}}')
>>> print (tmpl.substitute(name=u'Jos\xe9'))
Hi Jos爊The default encoding is UTF8.

The Tempita Language

The language is fairly simple; all the constructs look like {{stuff}}.

Substitution

To insert a variable or expression, use {{expression}}. You can't use }} in your expression, but if it comes up just use } } (put a space between them). You can pass your expression through filters with {{expression | filter}}, for instance {{expression | repr}}. This is entirely equivalent to {{repr(expression)}}. But it might look nicer to some people; I took it from Django because I liked it. There's a shared namespace, so repr is just an object in the namespace.

If you want to have {{ or }} in your template, you must use the built-in variables like {{start_braces}} and {{end_braces}}. There's no escape character.

You may also specify the delimiters as an argument to the Template __init__ method:

>>> tempita.Template(content='Hello ${name}', delimiters=('${', '}')).substitute(name='world')
'Hello world'

The delimiters argument must be of length two and both items must be strings.

None, as a special case, is substituted as the empty string.

Also there is a command for setting default values in your template:

{{default width = 100}}

You can use this so that the width variable will always have a value in your template (the number 100). If someone calls tmpl.substitute(width=200) then this will have no effect; only if the variable is undefined will this default matter. You can use any expression to the right of the =.

if

You can do an if statement with:

{{if condition}}
  true stuff
{{elif other_condition}}
  other stuff
{{else}}
  final stuff
{{endif}}

Some of the blank lines will be removed when, as in this case, they only contain a single directive. A trailing : is optional (like {{if condition:}}).

for

Loops should be unsurprising:

{{for a, b in items}}
    {{a}} = {{b | repr}}
{{endfor}}

See? Unsurprising. Note that nested tuples (like for a, (b, c) in...) are not supported (patches welcome).

inherit & def

You can do template inheritance. To inherit from another template do:

{{inherit "some_other_file"}}

From Python you must also pass in, to Template, a get_template function; the implementation for Template.from_filename(...) is:

def get_file_template(name, from_template):
    path = os.path.join(os.path.dirname(from_template.name), name)
    return from_template.__class__.from_filename(
        path, namespace=from_template.namespace,
        get_template=from_template.get_template)

You can also pass in a constructor argument default_inherit, which will be the inherited template name when no {{inherit}} is in the template.

The inherited template is executed with a variable self, which includes self.body which is the text of the child template. You can also put in definitions in the child, like:

{{def sidebar}}
  sidebar links...
{{enddef}}

Then in the parent/inherited template:

{{self.sidebar}}

If you want to make the sidebar method optional, in the inherited template use:

{{self.get.sidebar}}

If sidebar is not defined then this will just result in an object that shows up as the empty string (but is also callable).

This can be called like self.sidebar or self.sidebar() -- defs can have arguments (like {{def sidebar(name)}}), but when there are no arguments you can leave off () (in the call and definition).

Python blocks

For anything more complicated, you can use blocks of Python code, like:

{{py:x = 1}}

{{py:
lots of code
}}

The first form allows statements, like an assignment or raising an exception. The second form is for multiple lines. If you have multiple lines, then {{py: must be on a line of its own and the code can't start out indented (but if you have something like def x(): you would indent the body).

These blocks of code can't output any values, but they can calculate values and define functions. So you can do something like:

{{py:
def pad(s):
    return s + ' '*(20-len(s))
}}
{{for name, value in kw.items()}}
{{s | pad}} {{value | repr}}
{{endfor}}

As a last detail {{# comments...}} doesn't do anything at all, because it is a comment.

bunch and looper

There's two kinds of objects provided to help you in your templates. The first is tempita.bunch, which is just a dictionary that also lets you use attributes:

>>> bunch = tempita.bunch(a=1)
>>> bunch.a
1
>>> list(bunch.items())
[('a', 1)]
>>> bunch.default = None
>>> print (bunch.b)
None

This can be nice for passing options into a template.

The other object is for use inside the template, and is part of the default namespace, looper. This can be used in for loops in some convenient ways. You basically use it like:

{{for loop, item in looper(seq)}}
  ...
{{endfor}}

The loop object has a bunch of useful methods and attributes:

.index

The index of the current item (like you'd get with enumerate())

.number

The number: .index + 1

.item

The item you are looking at. Which you probably already have, but it's there if you want it.

.next

The next item in the sequence, or None if it's the last item.

.previous

The previous item in the sequence, or None if it's the first item.

.odd

True if this is an odd item. The first item is even.

.even

True if it's even.

.first

True if this is the first item.

.last

True if this is the last item.

.length

The total length of the sequence.

.first_group(getter=None)

Returns true if this item is the first in the group, where the group is either of equal objects (probably boring), or when you give a getter. getter can be '.attribute', like '.last_name' -- this lets you group people by their last name. Or a method, like '.birth_year()' -- which calls the method. If it's just a string, it is expected to be a key in a dictionary, like 'name' which groups on item['name']. Or you can give a function which returns the value to group on. This always returns true when .first returns true.

.last_group(getter=None)

Like first_group, only returns True when it's the last of the group. This always returns true when .last returns true.

Note that there's currently a limitation in the templating language, so you can't do {{for loop, (key, value) in looper(d.items())}}. You'll have to do:

{{for loop, key_value in looper(d.items())}}
  {{py:key, value = key_value}}
  ...
{{endfor}}

HTMLTemplate

In addition to Template there is a template specialized for HTML, HTMLTemplate (and the substitution function sub_html).

The basic thing that it adds is automatic HTML quoting. All values substituted into your template will be quoted unless they are specially marked.

You mark objects as instances of tempita.html. The easiest way is {{some_string | html}}, though you can also use tempita.html(string) in your functions.

An example:

>>> tmpl = tempita.HTMLTemplate('''\
... Hi {{name}}!
... <a href="{{href}}">{{title|html}}</a>''')
>>> name = tempita.html('<img src="bob.jpg">')
>>> href = 'Attack!">'
>>> title = '<i>Homepage</i>'
>>> tmpl.substitute(locals())
'Hi <img src="bob.jpg">!\n<a href="Attack!&quot;&gt;"><i>Homepage</i></a>'

It also adds a couple handy builtins:

html_quote(value):

HTML quotes the value. Turns all unicode values into character references, so it always returns ASCII text. Also it calls str(value) or unicode(value), so you can do things like html_quote(1).

url(value):

Does URL quoting, similar to html_quote().

attr(**kw):

Inserts attributes. Use like:

<div {{attr(width=width, class_=div_class)}}>

Then it'll put in something like width="{{width}}" class={{div_class}}. Any attribute with a value of None is left out entirely.

Extending Tempita

It's not really meant for extension. Instead you should just write Python functions and classes that do what you want, and use them in the template. You can either add the namespace to the constructor, or extend default_namespace in your own subclass.

The extension that HTMLTemplate uses is to subclass and override the _repr(value, pos) function. This is called on each object just before inserting it in the template.

Two other methods you might want to look at are _eval(code, ns, pos) and _exec(code, ns, pos), which evaluate and execute expressions and statements. You could probably make this language safe with appropriate implementations of those methods.

Command-line Use

There's also a command-line version of the program. In Python 2.5+ you can run python -m tempita; in previous versions you must run python path/to/tempita/__init__.py.

The usage:

Usage: __init__.py [OPTIONS] TEMPLATE arg=value

Use py:arg=value to set a Python value; otherwise all values are
strings.


Options:
  --version             show program's version number and exit
  -h, --help            show this help message and exit
  -o FILENAME, --output=FILENAME
                        File to write output to (default stdout)
  --html                Use HTML style filling (including automatic HTML
                        quoting)
  --env                 Put the environment in as top-level variables

So you can use it like:

$ python -m tempita --html mytemplate.tmpl \
>     var1="$var1" var2="$var2" > mytemplate.html

Still To Do

• Currently nested structures in for loop assignments don't work, like for (a, b), c in x. They should.

• There's no way to handle exceptions, except in your py: code. I'm not sure what there should be, if anything.

• Probably I should try to dedent py: code.

• There should be some way of calling a function with a chunk of the template. Maybe like:

  {{call expr}}
    template code...
  {{endcall}}
  

  That would mean {{expr(result_of_template_code)}}. But maybe there should be another assignment form too, if you don't want to immediately put the output in the code ({{x = call}}...{{endcall}}?). For now defs could be used for this, like:

  {{def something}}
    template code...
  {{enddef}}
  {{expr(something())}}
  

News

0.6

• Support for Python 2 has been removed, Tempita is now Python 3 only.
• Tests rewritten to use unittest instead of nose

0.5

• Python 3 compatible.
• Fixed bug where file-relative filenames wouldn't work well.
• Fixed the stripping of empty lines.

0.4

• Added a line_offset constructor argument, which can be used to adjust the line numbers reported in error messages (e.g., if a template is embedded in a file).
• Allow non-dictionary namespace objects (with tmpl.substitute(namespace) (in Python 2.5+).
• Instead of defining __name__ in template namespaces (which has special rules, and must be a module name) the template name is put into __template_name__. This became important in Python 2.5.
• Fix some issues with r

0.3

• Added {{inherit}} and {{def}} for doing template inheritance.
• Make error message annotation slightly more robust.
• Fix whitespace stripping for the beginning and end of lines.

0.2

• Added html_quote to default functions provided in HTMLTemplate.
• HTML literals have an .__html__() method, and the presence of that method is used to determine if values need to be quoted in HTMLTemplate.