Skip to main content

Extract template (a pattern) from strings and parse otherstrings with this pattern.

Project description


Given some strings (or files), this library extracts a common template between them (method learn) – some people call it “reverse templating”. Having your template created, you can parse other strings/files using it - the parse method will return only what changes in this file (the “blanks”). It does something like the opposite of what template libraries (such as Jinja) do. But for now, it only can identify fixed variables (it can’t create for and if blocks, for example).

If you have the template and the “blanks” you can also fill the blanks with the method join - it’ll return a string with the template filled. There are some other features:

  • If you don’t want/need to Templater (the main class) create the template for you, you can pass a pre-processed template (created manually or created before using learn and saved somewhere).

  • You can split the learning and parsing process, since the learning process generally is executed one time and takes a lot of time compared to parsing process. To turn this process handy, Templater has the methods dump, save, load and open, so you can learn and save a template definition for later loading and parsing how many times you want (you can also load, learn more and save).

templater is simple to use, easy to learn and does the hard work for you (for example: part of the learning algorithm is implemented in C for performance). Do you have 5 minutes? So learn with the Examples.


templater is available at PyPI, so installing it is as simple as executing:

pip install templater

Or you can download the latest version and install it using

git clone
cd templater
python build install


There are some definitions/concepts we should explicit here:

  • Template: the whole object (instance of Templater).

  • Document: a string or file that have some kind of pattern. You’ll use documents to make a template object learn and recognize these patterns, so later you can use the template object to parse a document and get only the information that is not “static”.

  • Blocks: the fixed parts of a template. Can change (in number and size) when learn is run.

  • Blanks: also called holes or variables, blanks are the parts in a template that changes between documents with the same template.

  • Template definition: the information stored in a template that defines it (it is a Python list with a very simple grammar that describes how the template is composed).

  • Markers: when you want to save a template, something should be put between blocks to “mark” the blanks (so the template definition can be reconstructed later).

  • Named marker: a marker plus a header is called a named marker. They are handy and more legible since you can access the “blanks” by names instead of indexes.

Doubts? Don’t worry, see the Examples and you’ll get it.


All you need to know is below (and in the examples directory):

>>> from templater import Templater
>>> documents_to_learn = ['<b> spam and eggs </b>', '<b> ham and spam </b>',
                          '<b> white and black </b>'] # list of documents
>>> template = Templater()
>>> for document in documents_to_learn:
...    template.learn(document)

>>> print 'Template created:', template._template # template definition
Template created: [None, '<b> ', None, ' and ', None, ' </b>', None]

>>> document_to_parse = '<b> yellow and blue </b>'
>>> print 'Parsing other document:', template.parse(document_to_parse)
Parsing other document: ['', 'yellow', 'blue', '']

>>> print 'Filling the blanks:', template.join(['', 'red', 'orange', ''])
Filling the blanks: <b> red and orange </b>

You can pass pre-processed templates as a list (blanks are None, blocks are strings):

>>> t2 = Templater(template=[None, 'Music: ', None, ', Band: ', None])
>>> print t2.join(['', 'Welcome to the Jungle', 'Guns and Roses'])
Music: Welcome to the Jungle, Band: Guns and Roses

…or you can pass a string with markers, then Templater will create the list for you:

>>> t3 = Templater(template='language=#,cool=#', marker='#')
>>> print t3.join(['', 'Python', 'YES', ''])

Saving and opening templates is easy:

>>>'my-first-template.html', marker='|||')
>>> # and some time later...
>>> loaded_template ='my-first-template.html', marker='|||')
>>> print loaded_template.parse('<b> Romeo and Juliet </b>')
['', 'Romeo', 'Juliet', '']

The difference between save and dump is that save stores the template string, filling the blanks with a marker and dump saves the whole Templater object with cPickle. The pairs are:

  • save and open (raw template string filled with marker)

  • load and dump (whole object)

Note: save always add a \n to the end of file; load deletes trailing \r\n or \n in the end of file (if any).

Note-2: when passing a pre-processed template (using Templater initializer or make sure it starts and ends with a marker.

If you are getting a lot of blanks you can configure the learning process: just adjust min_block_size - it’s the minimum number of characters permitted to create a new block in template:

>>> str_1 = 'my favorite color is blue'
>>> str_2 = 'my favorite color is violet'
>>> t = Templater() # default min_block_size = 1
>>> t.learn(str_1)
>>> t.learn(str_2)
>>> print t._template
[None, 'my favorite color is ', None, 'l', None, 'e', None]

We don’t want that 'l' and 'e' there, right? So:

>>> t = Templater(min_block_size=2)
>>> t.learn(str_1)
>>> t.learn(str_2)
>>> print t._template
[None, 'my favorite color is ', None]

You can also add “headers” to your template - the headers will be the name of your markers, so you’ll have a template with named markers and parse will return a dict instead of list. It’s more legible than using list indices, let’s see:

>>> import re
>>> # Let's create a regexp that cases with '{{var}}' (it'll be our marker)
>>> regexp_marker = re.compile(r'{{([a-zA-Z0-9_-]*)}}')
>>> template = Templater('{{first-var}}<b>{{second-var}}</b>{{third-var}}',
>>> # The template knows the name of each marker just using the regexp provided
>>> # Passing marker as regexp to specify named markers also work for

>>> print template.parse('This <b> is </b> a test.')
{'second-var': ' is ', 'third-var': ' a test.', 'first-var': 'This '}

>>> # To save the template with named markers we need to provide a Python string.
>>> # Templater will call .format() of this string for each marker with its name
>>>'template-with-named-markers.html', marker='--{}--')
>>> # Will save '--first-var--<b>--second-var--</b>--third-var--\n'

And if you have a template without headers, just add to it with add_headers method:

>>> t = Templater('+<tr><td>+</td><td>+</td></tr>+', marker='+')
>>> t.parse('<tr><td>hello</td><td>world</td></tr>')
['', 'hello', 'world', '']

>>> t.add_headers(['before', 'first-column', 'second-column', 'after'])
>>> t.parse('<tr><td>hello</td><td>world</td></tr>')
{'after': '', 'before': '', 'first-column': 'hello', 'second-column': 'world'}

Note: named markers have a problem: you can’t run learn if you use them.


I really want to know if you are using this project and what is your impression about it. If you have new ideas of features, discovered bugs or just want to say “thank you, I’m using it!”, please contact me at alvarojusten at gmail.

If you want to code some stuff, just fork it on GitHub and create a pull request. Some technical notes for you:


This software is developed by Álvaro Justen aka Turicas.

Many thanks to Adrian Holovaty - he created templatemaker, the project which templater was inspired in/forked from - and to Escola de Matemática Aplicada (Fundação Getúlio Vargas) which gives me interesting problems to solve. :-)


GPL version 2

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

templater-0.4.0.tar.gz (16.5 kB view hashes)

Uploaded Source

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page