Repository to interface and parse hydrographic and scientific sonar files
Formats in development: - KMALL - S7K - MBCDF
Parsely is a library for parsing multibeam echosounder (MBES) datagrams and files. Currently, the package acts as a file interface that reads the binary datafiles to python objects that the user can interface with. In addition to the ability to parse data, the library has a format converter to translate MBES data from proprietary format to an open source format built on the netCDF4 format (eventually).
Most manufacturers package their data in a compact binary representation stored in relatively chronological order with packets of data from various sensors interleaved.
In order to read these packets and files, this project adopts a dataclass approach using the attrs package. If a datagram has sub compenents (The substructures in KMALL, the DRF/RTH/RD... of Reson S7k) these are treated as child dataclass objects which are constructed first and inserted into the parent datagram dataclass (dependency injection). For documentation purposes, references to dataclass reference an attrs class.
A dataclass contains the structure of the data, and where applicable relevant metadata and initial values. A dataclass instance can be constructed using the class's factory method which parses the data chunk related to that datagram. Constructing a dataclass instance without the factory method is possible, but requires the user to unpack the data and pass the attributes (preferably by keyword) directly, ensuring data types match.
The datagram definitions are defined independently of the readers allowing the user to create their own readers if they choose. However, the library includes a reader which serves as an interface to the file. As a result, the MBES file_reader object behaves similar to the standard library file object. A file_reader is associated with a single file, supports context management using the with command, should be destroyed when no longer in use.
The python standard library has support for dataclasses through the dataclass module. Initial development was done using this package. However , development has switched to attrs for the following reasons: - Maturity: attrs has been in development much longer than dataclass - Validation: It is possible to validate data inputs to a datagram class, which is useful when creating mock objects for testing - Slots: The biggest reason, attrs objects are slotted by default when fully typed, which reduces computational time and memory. While dataclass does provide the same support, this is a new feature only available in Python 3.10, of which Pydro does not support yet.