v0.1.6

Fix for the use of non-named groups in patterns

v0.1.5

• use fullmatch() instead of match() (avoid matching trailing characters)
• allow transformer functions to return histogram file names so outputs can be arbitrary

v0.1.4

Add --defer option to speed up single-shot processing (accumulate all transformation inputs before calling functions)
Add --delaywrite option to allow output histograms to all be published at once at end of job

Fix histogram ordering bug

Fix issue #5 .

v0.1.2

Fix issue preventing writing out histograms in 0.1.1

v0.1.1

Fixes setup.py release number

v0.1.0

Refactors interface to postprocessing functions to follow convention agreed with ATLAS. Specifically, the functions can be called with multiple histograms at once if a pattern appears in the matching regex that is not present in the output histogram name format string. (This is done so that you can e.g. combine myhistogram1 and myhistogram2 into a single output.)

This is handled using a new HistCombinationIterable object which can be iterated over or indexed to obtain a pair (match, list_of_histograms), in which the first entry is a dict that gives the specific pattern match satisfied by this entry, and list_of_histograms is a list with the same number of entries as the Input specification in the yaml file. The patterns will be matched correctly across the histograms in the list using the combinatoric engine.

In order that we don't call the postprocessing functions with partial subsets of the full list of matches, a "warmup" phase has been added. For the ROOT input module, this does a dry run first to figure out what histograms are available, so the regular postprocessing phase knows what to expect.

Various examples of using these matching features have been added to example.yaml.

v0.0.6

Should function for ATLAS