Semantic code browsing

Resumen

Programmers currently enjoy access to a very high number of code repositories and libraries
of ever increasing size. The ensuing potential for reuse is however hampered by the fact that
searching within all this code becomes an increasingly difficult task. Most code search engines
are based on syntactic techniques such as signature matching or keyword extraction. However,
these techniques are inaccurate (because they basically rely on documentation) and at the
same time do not offer very expressive code query languages. We propose a novel approach
that focuses on querying for semantic characteristics of code obtained automatically from the
code itself. Program units are pre-processed using static analysis techniques, based on abstract
interpretation, obtaining safe semantic approximations. A novel, assertion-based code query
language is used to express desired semantic characteristics of the code as partial specifications.
Relevant code is found by comparing such partial specifications with the inferred semantics for
program elements. Our approach is fully automatic and does not rely on user annotations or
documentation. It is more powerful and flexible than signature matching because it is parametric
on the abstract domain and properties, and does not require type definitions. Also, it reasons
with relations between properties, such as implication and abstraction, rather than just equality.
It is also more resilient to syntactic code differences. We describe the approach and report on a
prototype implementation within the Ciao system.