rnginline 0.0.2

Flatten multi-file RELAX NG schemas

rnginline is a Python library and command-line tool for loading multi-file RELAX NG schemas from arbitary URLs, and flattening them into a single RELAX NG schema.

Features

• Convert multi-file RNG schemas into one file without breaking or restructuring the schemas

  • Great care is taken to maintain the semantics of the separate schema files in the single output

  • The input documents are changed as little as possible, so the output is as readable as the input

• Load schemas from:

  • The filesystem

  • From a Python package’s data (without unpacking it to the filesystem)

• command-line interface as well as a Python API

• Test suite covering lots of edge cases, e.g. namespace handling

  • 100% line & branch code coverage

Quickstart

Install with pip:

$ pip install rnginline

You can use it from Python like this:

>>> import rnginline
>>> rnginline.inline('my-nested-schema-root.rng')
<lxml.etree.RelaxNG object at ...>

You can load a multi-file schema from a Python package’s data like this:

>>> import rnginline
>>> from rnginline.urlhandlers import pydata
>>> url = pydata.makeurl('rnginline.test',
...                      'data/testcases/external-ref-1/schema.rng')
>>> url
'pydata://rnginline.test/data/testcases/external-ref-1/schema.rng'
>>> rnginline.inline(url)
<lxml.etree.RelaxNG object at ...>

You can use it from the command line like this:

$ rnginline my-nested-schema-root.rng flattened-output.rng

Documentation

Documentation is available at http://rnginline.readthedocs.org/

Motivation

lxml has good support for using RELAX NG schemas, but lacks support for loading multi-file schemas from anywhere other than the filesystem. This is a problem if you wish to bundle a multi-file schema with your Python package/module. You’d have to depend on setuptools being available to use its resource extraction, or use one of the existing RELAX NG merging tools to convert your schema into a single file.

Existing RELAX NG flattening tools

The following projects didn’t quite fit my needs, leading me to write rnginline. They may work for you though.

• rng2srng - Implements full simplification, so the structure of the input schema will be lost

• rng-incelim - A similar project to this, implemented in XSLT. Unfortunately doesn’t handle namespace declarations on <include> elements correctly. XSLT 1.0 doesn’t support creating namespace nodes, so to fix this rng-incelim would have to resolve all QNames in the schema to NCNames with ns attributes, which would be undesirable for me.

Developing

An invoke tasks file contains testing, doc building, linting tasks. The tox config uses invoke to run tests over various Python versions.

To do a full test run, install tox ($ pip install tox) then run $ tox from within the checked-out repo. This will take several minutes on the first run, mainly due to the amount of time lxml takes to build from source.

To run invoke tasks, ensure all the dev dependencies are installed:

$ pip install -r requirements/all.txt
$ invoke -l
Available tasks:

  […]
  coverage    Combine coverage of Python 2 and Python 3 test runs
  docs        Build sphinx documentation
  docs_test   Test the doctests in the Sphinx docs. Must be run with Python 3.
  pep8        Lint code for PEP 8 violations
  readme      Lint the README for reStructuredText syntax issues
  test        Run rnginline test suite
$ invoke test
============================= test session starts ==============================
platform darwin -- Python 2.7.8 -- py-1.4.26 -- pytest-2.6.4
plugins: cov
collected 346 items

[…]

Prebuilding Wheels

To mitigate the time lxml takes to build, you can pre-build all the dependencies as Wheels into ./wheelhouse and have pip use the pre-built wheels on every install. To do so, run:

$ invoke cache_all_requirement_wheels
$ export PIP_FIND_LINKS=$(pwd)/wheelhouse