Test-driven code search addresses the problem of searching and reusing code. One of the first things a programmer is expected to do when developing new functionality is to find existing code with the same functionality and reuse as much of that code as possible. However, to be useful, existing code has to behave the way programmer expects. This requires systematic approaches for specifying the behavior of a desired functionality and validating it against existing instances; two hard problems yet to be resolved in code search and retrieval. We are seeking to confront these problems by applying a novel approach that exploits the semantic information in the test code.