Features

• Simple and structured way to represent data as a set of named properties within and between software

• Simple type system where data type are specified as a basic type and size Supported basic types includes:

  • binary blob (any size)

  • boolean (system-dependent size, typically 8 bits)

  • integer (8, 16, 32, 64 bits)

  • unsigned integer (8, 16, 32, 64 bits)

  • float (32, 64, [80, 128] bits)

  • fixed string (any size, always NUL-terminated)

  • string pointer (pointer to malloc’ed string, NUL-terminated)

  • relation (subject-predicate-object triplet)

  • dimension (only intended for metadata)

  • property (only intended for metadata)

• Supports units and multi-dimensional arrays

• Fully implemented metadata model as presented by Thomas Hagelien

• Builtin JSON, HDF5, RDF, YAML, PostgreSQL, csv, blob storage plugins (JSON is always available, the other depends on external libraries)

• Plugin system for user-provided storage drivers

• Memory for metadata and instances is reference counted

• Lookup of metadata and instances at pre-defined locations (initiated from the DLITE_STORAGES environment variable)

• Template-based code generation, including templates for C and Fortran

• Plugin system for mappings that maps instances of a set of input metadata to an output instance

• Fortran bindings

• Python bindings

• Embedding of Python allowing writing storage and mapping plugins in Python

• Full provenance in instance evolution; see transactions