0.0.25

Try again to trigger deployment.

0.0.24

Fixed two bugs (PRs #105, #107) and added a MANIFEST.in.

0.0.23

Added Sphinx/Readthedocs documentation and promoted uproot4.behaviors.TBranch.Report from a namedtuple to a regular class (to make it inspectable by autodoc).

0.0.22

@raymondEhlers fixed the interpretation of TH2 and TH3 histograms as NumPy in PR #97. (They were transposed.)

@jpivarski added an interface to offload building of some doubly jagged arrays to Awkward. This is the general plan for anything that CanBeAwkward, but for now, we just have this foot-in-the-door for one specific type. PR #96. (Only works when used with awkward1 0.2.37.) TL;DR this data type can now be read 80× faster.

@jpivarski also added an interface for file-like objects (anything with a read and seek method).

0.0.21

Added docsstrings to everything and fixed handling of lazy arrays for scikit-hep/awkward#428.

0.0.20

Fixed #32, #36, #38 (as a bug, not as a feature), and #85.

0.0.19

Fixed some bugs, dropped the Dask backend, and modified the colon-handling in filenames (long discussion here). The main thing that came out of that was a more fully descriptive error message.

0.0.18

@douglasdavis fixed a bug in which TArrayDs were assumed to be NumPy arrays (with a reshape method).

@nsmith- ported the Uproot 3 "clusters" function, now named common_entry_offsets. He also prompted me to fix my use of units: MB vs MiB.

I revamped the physical layer to match the way it's actually used in the higher-level code. It's about half as many lines as before. (No longer duplicating thread-handling in vector-read and multipart classes; it's all handled by executors now.)

0.0.17

Fixed an HTTP-freeze issue (fallback method wasn't informed that it needs to notify when done) discovered when writing Awkward tutorials.

0.0.16

Fixed scikit-hep/uproot3#513.