asn1ate 0.4

ASN.1 translation library.

asn1ate – ASN.1 translation library. Copyright 2013 Schneider Electric Buildings AB

Introduction

asn1ate is a Python library for translating ASN.1 into other forms. It is intended for code generation from formal ASN.1 definitions, and a code generator for pyasn1 is included.

asn1ate is released under a 3-clause BSD license. For details, see LICENSE.txt.

Caveat

This is very much an alpha-quality prototype. Things that need doing:

• Regression test suite

• HACK/TODO/BUGs need to be fixed

• ASN.1 grammar is very incomplete in some places. Known issues:

  • Constraint syntax is currently limited to simple value range constraints (ftp://ftp.rsasecurity.com/pub/pkcs/pkcs-12/pkcs-12.asn, line 53)

  • Reference syntax is not accepted at all (ftp://ftp.rsasecurity.com/pub/pkcs/pkcs-12/pkcs-12.asn, line 74)

• Improve parser error handling/reporting

• Allow for semantic analysis, e.g. validity check, warnings for problematic constructs, etc.

Usage

The immediate use of asn1ate is to generate pyasn1 definitions from ASN.1 definitions. The command to do this is:

$ python .../asn1ate/pyasn1gen.py source.asn1

It will print the pyasn1 equivalent of source.asn1 to stdout.

Dependencies

The only third-party dependency is pyparsing.

Although asn1ate was initially developed on Python 3.2, it has been tested with Python 2.7.3 and should port to older Python versions easily.

Design notes

The asn1ate package is designed along the same lines as a compiler with a driver, a parser, a semantic model and a convention for code generators.

• parser.py – a tokenizing parser for ASN.1 per X.680. It currently recognizes a naive sub-set of X.680

• sema.py – a semantic ASN.1 object model, which can be constructed from the AST generated by parser.py

• support/pygen.py – a support library for generating Python code.

• pyasn1gen.py – a code generator to transform a semantic model into pyasn1 syntax. This can be used as a script in which case it will dump output to stdout.

The ASN.1 parser is very ad-hoc, I’ve experimented with the grammar until I found something that accepted our proprietary ASN.1 definition. It’s based on pyparsing but sets up parse actions to build an annotated AST. Every node of interest is annotated with a string denoting its type, e.g. Identifier, TypeAssignment, etc. I’ve tried to stay with token types as named in X.680, but added custom ones or suppressed others, as necessary to get the AST in a useful shape.

Annotated tokens are represented by a simple class containing the type name and a list of children (called elements) which may be annotated tokens, lists or simple values. This gives a very discoverable tree structure, but there are probably cleaner AST representations we could use. Patches welcome.

asn1ate.sema is an object model that represents ASN.1 constructs. It describes everything from type assignments to default values and tags, but still only the parts of ASN.1 we happen to use here. Most of the logic revolves around transforming the AST produced by asn1ate.parser into a more semantic model with proper Python objects.

Codegen is designed to be extensible. In-house we have a set of code generators to build an entire protocol stack based on an ASN.1 source, but asn1ate only includes the generally useful one, asn1ate.pyasn1gen.

The most notable members of asn1ate.support are probably the PythonWriter and PythonFragment classes, which simplify generation of correctly indented Python code.