0.5.0

• Access named columns via myarray.at.x, equivalent to myarray["x"].
• Access numeric columns via myarray.at(0), equivalent to myarray["0"].
• All subindexes of jagged indexing is now supported.

0.4.5

• Added astype method to all classes. (#39)
• Corrected use of reduceat in jagged reducers. (#38)
• Bubble up mix-in methods from deeper nesting when selecting through column name.
• Added test_crosscut and test_methods.

0.4.4

Fixed #31 and #33. It is now possible to slice the second dimension of a jagged array:

jaggedarray[:, start:stop]

Jagged arrays that happen to have fixed dimension can be cast as a regular array:

jaggedarray.regular()

And an error was fixed in filtering jagged arrays with jagged masks that affected sublists with more than 256 elements.

0.4.3

Persist Table views, not just base Tables.

0.4.2

Formally migrated to pytest, fixed the infinite loop when converting awkward-arrays to Numpy (#19), fixed a bug in UnionArray ufuncs (probably #15), and added a mechanism to avoid pickle in awkward-array serializations.

0.4.1

To get a Zenodo DOI.

0.4.0

Arrow buffers can now be viewed as awkward arrays, and by extension, Parquet files. Parquet files are read lazily as ChunkedArray(VirtualArray(...)).

Persistence has also been updated, both to accommodate Arrow/Parquet and also to provide a consistent, clean interface.

0.3.0

Persistence: awkward arrays may now be read from and written to pickle, ZIP files, HDF5 files, and anything with a dict-like interface (e.g. shelve). Compression and cross-references/cyclic-references are included.

0.2.1

Added JaggedArray.fromindex (#10) and fixed UnionArray.fromtags (#8).

0.2.0

Finished a basic implementation of all planned awkward array classes.