Most reuse libraries come with few source-code examples that demonstrate how the library at hand should be used. We have developed a source-code recommendation approach for constructing and delivering relevant code snippets that programmers can use to complete a certain programming task. Our approach is semantic-based; relying on an explicit ontological representation of source-code. We argue that such representation opens new doors for an improved recommendation mechanism that ensures relevancy and accuracy. Current recommendation systems require an existing repository of relevant code samples. However, for many libraries, such a repository does not exist. Therefore, we instead utilize points-to analysis to infer precise type information of library components. We have backed our approach with a tool that has been tested on multiple libraries. The obtained results are promising and demonstrate the effectiveness of our approach.