Version 1.5.0 beta 5

Changes

• Dimension links: Dimensions can now be linked to data objects (DataArray or DataFrame). This links a Dimension object to a specific vector of data, which is either a single vector of a DataArray's data or a single column of a DataFrame.
  • This removes the AliasRangeDimension object. The same functionality can be achieved by adding a RangeDimension and then linking it back to the parent DataArray.
  • The FILE FORMAT version is now 1.2.0.
• New script nixio-upgrade: Upgrades files to newest version if there are file format changes.
• Every NIX object holds a reference to the open file object it belongs to and can be accessed via obj.file.
• Storage optimisation: Dataset and property creation was causing large per-object overhead which made files with a large number of small objects take up too much space. We changed the way these are initialised so file sizes should be much smaller going forward.
• MultiTags can have extents specified in the constructor.
• MultiTags can now be created by supplying an array or a list for the positions and extents arguments. This will result in the automatic creation of DataArrays to hold these values.
• DataArray now has an iter_dimensions() method which, when iterated upon, generates a 1-based index along with each dimension of the array.
• Changed the structure of the information returned by the File.validate() method. The information is now organised in a dictionary with two keys, "errors" and "warnings". Each contains dictionary that maps NIX objects to a list of error or warning messages respectively.

Bug fixes

• Fixed the behaviour of the object modification time and the option which disables it.
• Fixed several bugs in the validator module.
• Fixed DataFrame creation using a compound (structured) data type without specifying column names.

Relevant PRs

• #421: Validator part 2
• #423: Validator script
• #432: Add shape when initializing properties to optimize storage
• #434: Fix bug for h5group.root
• #436: Allow supplying np_arrays or list for multi_tags positions
• #438: Add method that returns 1-based index dimension iterators
• #440: Change the search method of index_of() for range_dim
• #442: Enforce unique column names for DataFrame
• #445: Fix bug when create dataframe from compound arrays and add test
• #452: Upgrade script
• #462: File references
• #464: Fixes and simplifications for DataTypes and DataType mappings
• #469: Dimension links