The task of content-type detection—which entails identifying the data encoded in an arbitrary byte sequence—is critical for operating systems, development, reverse engineering environments, and a variety of security applications. In this paper, we introduce Magika, a novel AI-powered content-type detection tool. Under the hood, Magika employs a deep learning model that can execute on a single CPU with just 1MB of memory to store the model's weights. We show that Magika achieves an average F1 score of 99% across over a hundred content types and a test set of more than 1M files, outperforming all existing content-type detection tools today. To foster adoption and improvements, we open source Magika under an Apache 2 license on GitHub and we make our model and training pipeline publicly available. Our tool has already seen adoption by Gmail and Google Drive for attachment scanning, by VirusTotal to aid with malware analysis, and by prominent open-source projects such as Apache Tika. While this paper focuses on the initial version, Magika continues to evolve with support for over 200 content types now available. The latest developments can be found at https://github.com/google/magika.

Erasable coercions in System F-eta, also known as retyping functions, are well-typed eta-expansions of the identity. They may change the type of terms without changing their behavior and can thus be erased before reduction. Coercions in F-eta can model subtyping of known types and some displacement of quantifiers, but not subtyping assumptions nor certain forms of delayed type instantiation. We generalize F-eta by allowing abstraction over retyping functions. We follow a general approach where computing with coercions can be seen as computing in the lambda-calculus but keeping track of which parts of terms are coercions. We obtain a language where coercions do not contribute to the reduction but may block it and are thus not erasable. We recover erasable coercions by choosing a weak reduction strategy and restricting coercion abstraction to value-forms or by restricting abstraction to coercions that are polymorphic in their domain or codomain. The latter variant subsumes F-eta, F-sub, and MLF in a unified framework.

Bidirectional programming languages are a practical approach to the view update problem. Programs in these languages, called lenses, define both a view and an update policy - i.e., every program can be read as a function mapping sources to views as well as one mapping updated views back to updated sources.