Programming by Examples

Programming by Examples (PBE) has the potential to revolutionize enduser programming by enabling end users, most of whom are non-programmers, to create scripts for automating repetitive tasks. PBE involves synthesizing intended programs in an underlying domain-specific language (DSL) from example based specifications (Ispec). We formalize the notion of Ispec and discuss some principles behind designing useful DSLs for synthesis. A key technical challenge in PBE is to search for programs that are consistent with the Ispec provided by the user. We present a divide-and-conquer based search paradigm that leverages deductive rules and version space algebras for manipulating sets of programs. Another technical challenge in PBE is to resolve the ambiguity that is inherent in the Ispec. We show how machine learning based ranking techniques can be used to predict an intended program within a set of programs that are consistent with the Ispec. We also present some user interaction models including program navigation and active-learning based conversational clarification that communicate actionable information to the user to help resolve ambiguity in the Ispec. The above-mentioned concepts are illustrated using practical PBE systems for data wrangling (including FlashFill, FlashExtract, FlashRelate), several of which have already been deployed in the real world.