tectonic | columnar parsing framework JSON/CSV with support | CSV Processing library

tectonic is a Scala library typically used in Utilities, CSV Processing applications. tectonic has no bugs, it has no vulnerabilities and it has low support. You can download it from GitHub.

A columnar fork of Jawn with added backend support for CSV. The distinction between "columnar" and its asymmetric opposite, "row-oriented", is in the orientation of data structures which you are expected to create in response to the event stream. Jawn expects a single, self-contained value with internal recursive structure per row, and its Facade trait is designed around this idea. Tectonic expects many rows to be combined into a much larger batch with a flat internal structure, and the Plate class is designed around this idea. Tectonic is also designed to support multiple backends, making it possible to write a parser for any sort of input stream (e.g. CSV, XML, etc) while driving a single Plate. At present, both CSV and JSON are supported, and we have plans to support XML in the future. Finally, Tectonic implements a form of fast skipping based on signals returned from the user-supplied Plate implementation driving a particular parse. In this way, it is possible to achieve Mison-style projection pushdown into the parse process. At present, skip scans do not compile down to vectorized assembly instructions (SIMD), but it is theoretically possible to achieve this, despite sitting on top of the JVM. All of these differences have led to some relatively meaningful changes within the parser implementation. Despite that, the bulk of the ideas and, in some areas, the vast bulk of the implementation of Tectonic's JSON parser is drawn directly from Jawn. Special heartfelt thanks to Erik Osheim and the rest of the Jawn contributors, without whom this project would not be possible. Tectonic is very likely the optimal JSON parser on the JVM for producing columnar data structures. It is definitely the optimal JSON parser for producing columnar structures when you are able to compute some projections or row filters in advance, allowing skipping. When producing row-oriented structures (such as conventional JSON ASTs), it falls considerably behind Jawn both in terms of performance and usability. Tectonic is also relatively opinionated in that it assumes you will be applying it to variably-framed input stream (corresponding to Jawn's AsyncParser) and does not provide any other operating modes.