Phildb

Bug fix release:
- Fix 12: Passing an empty series no longer results in an index error.

Performance/Bug fix release:
- Performance fix when writing irregular data to handle common case of updating end of series.
- Fixes excess logging of unchanged values

Bug fix release:
- Fix when frequency string doesn't have numeric prefix. Pandas doesn't include a numeric prefix for frequency type 'T' so this fix sets the multiplier to (the implied) one if the entire string consists of 'T'. Thus ensuring that minute data can be stored and accessed.

This release also includes changes to allow building as a conda package. Conda packages now uploaded to Anaconda Cloud: https://anaconda.org/amacd31/phildb

Bug fix release:
- Issue 6: Convert examples to jupyter notebooks
- Fix 9: Changes to avoid numpy FutureWarning
- Bug fix: NaN values were not being updated
- Fix pandas FutureWarning about the use of take_last
- Store 64-bit dates in log files
- Ensure that only new or changed values are written when writing irregular data
- Performance fixes

Bug fix release:
- Fix read_log method on phildb.database.PhilDB.

Version 0.6.0 of PhilDB.
- BREAKING CHANGE: The phildb.database.PhilDB write method has changed, it now takes a pandas.Series instead of a tuple of arrays.
- Closed issue 3: New read_log method for reading a time series as it existed for a given datetime.
- Raise more informative exceptions when attempting to create a duplicate timeseries/attribute.
- Various minor updates to documentation.